• Micro$oft -- what's inside

    From Sergey Kubushyn@3:633/10 to All on Saturday, November 14, 2026 22:18:31
    and why the entire company, not just Azure and stuff, should've been long
    shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It
    is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It
    is terminal and uncurable. The giant is doomed.

    ---
    ******************************************************************
    * KSI@home KOI8 Net < > The impossible we do immediately. *
    * Las Vegas NV, USA < > Miracles require 24-hour notice. * ******************************************************************

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Don Y@3:633/10 to All on Saturday, November 14, 2026 17:15:31
    On 11/14/2026 3:18 PM, Sergey Kubushyn wrote:
    and why the entire company, not just Azure and stuff, should've been long shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It
    is terminal and uncurable. The giant is doomed.
    MS has never been a "tech powerhouse". Their skill has been in cornering
    a market.

    A market that has now slipped away from them -- alternative OSs (esp
    FOSS offerings) and platforms.

    Their hardware has always been laughable (though some of the surface
    offerings are tolerable but don't really stand out among similar
    offerings in that market segment)

    Microsoft phones?? PDAs??

    They have spent so much time tied to the past that they have missed out on the future.

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Saturday, November 14, 2026 16:56:22
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    and why the entire company, not just Azure and stuff, should've been long >shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It >is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It
    is terminal and uncurable. The giant is doomed.

    ---
    ******************************************************************
    * KSI@home KOI8 Net < > The impossible we do immediately. *
    * Las Vegas NV, USA < > Miracles require 24-hour notice. * >******************************************************************

    Nightmare. It's probably not fixable.


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Monday, November 16, 2026 00:19:14
    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    and why the entire company, not just Azure and stuff, should've been long
    shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It >> is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It >> is terminal and uncurable. The giant is doomed.

    ---
    ******************************************************************
    * KSI@home KOI8 Net < > The impossible we do immediately. *
    * Las Vegas NV, USA < > Miracles require 24-hour notice. *
    ******************************************************************

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more
    difficult. The US hasn't got as big a talent pool as China to dredge for people who might be able to fix the problems, and while both countries
    are educating most of the rising generation, rich and powerful parents
    in both countries have ways of ensuring that their kids get the best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their graduate students weren't great at presenting new ideas - taking teachers too
    seriously can be as much of problem as not taking them seriously enough.

    --
    Bill Sloman, Sydney

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Sunday, November 15, 2026 07:34:56
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    and why the entire company, not just Azure and stuff, should've been long >>> shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion >>>
    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It
    is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It >>> is terminal and uncurable. The giant is doomed.

    ---
    ******************************************************************
    * KSI@home KOI8 Net < > The impossible we do immediately. *
    * Las Vegas NV, USA < > Miracles require 24-hour notice. *
    ******************************************************************

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >difficult. The US hasn't got as big a talent pool as China to dredge for >people who might be able to fix the problems, and while both countries
    are educating most of the rising generation, rich and powerful parents
    in both countries have ways of ensuring that their kids get the best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their graduate >students weren't great at presenting new ideas - taking teachers too >seriously can be as much of problem as not taking them seriously enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    And somehow, after four years of that, they dress up in silly caps and
    gowns [1] and get a piece of paper that magically makes them able to
    think.

    I estimate that about 2% of the CE/EE grads understand electricity and
    have a chance of eventually designing electronics, or at least
    assisting.

    [1] I skipped my ceremony. Who wants to stand for hours in the New
    Orleans heat and humidity, in the sun, in a rented black robe,
    listening to boring speeches?


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From legg@3:633/10 to All on Sunday, November 15, 2026 11:35:18
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    and why the entire company, not just Azure and stuff, should've been long >shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It >is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It
    is terminal and uncurable. The giant is doomed.


    Question arises - why did the author stick around?

    After two weeks, he should have been hitting the road,
    looking forsomething that wouldn'twaste his time.

    I avoid this by researching and reviewing immediate
    project goals andresource asignments, before signing on.
    This is offered as a free design review under an NDA . . .

    If there are serious issues, the time isn't wasted.

    RL

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Don Y@3:633/10 to All on Sunday, November 15, 2026 11:27:18
    On 11/15/2026 9:35 AM, legg wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    and why the entire company, not just Azure and stuff, should've been long
    shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It >> is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It >> is terminal and uncurable. The giant is doomed.

    Question arises - why did the author stick around?

    Agreed. Hero-syndrome? (Surely *he* can fix this mess)

    Or, a desire to stick with something he knew instead of risk
    something new/novel?

    Or, The Money? (he likely lived in the area, already, so
    would prefer a "local" job)

    After two weeks, he should have been hitting the road,
    looking forsomething that wouldn'twaste his time.

    Some people don't put much value on their time -- as long
    as they are being compensated.

    I was asked to bid on a job many years ago -- with an
    "inside track" ("yours if you want it"). I listened to
    the primary stakeholder explain what he wanted, quizzed him
    about issues that needed fleshing out and went away to
    think on it.

    After a few hours, realized it would never get to "Finished",
    despite the monies they were going to throw at it.

    So, I "no bid" the job (politely as it had been intended as
    a gift to me).

    Didn't stop someone else from bidding on the job and issuing
    weekly invoices for the better part of a year...
    ... until the stakeholder was fired and his department
    folded into another.

    I'm sure the "consultant" laughed all the way to the bank.
    (despite having nothing to show for all of that time)

    I avoid this by researching and reviewing immediate
    project goals andresource asignments, before signing on.
    This is offered as a free design review under an NDA . . .

    If there are serious issues, the time isn't wasted.
    Some clients can be educated and their path altered to
    one that is more successful. But, that assumes you aren't
    fighting a shitload of design inertia/technical debt.

    The author likely FEELS vindicated -- despite having missed
    an opportunity to do something useful with those years of
    his life. (hint: you don't get them back, even if you are
    proven right!)

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From someone@3:633/10 to All on Sunday, November 15, 2026 20:30:01
    Microsoft draws from a global talent pool. The locals just do non-discretionary grunt work.

    Sounds like the writer was basically in the grunt work pits. The masterminds don't waste their time talking to that type.

    --
    For full context, visit https://www.electrondepot.com/electrodesign/micro-oft-what-s-inside-4423912-.htm


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From someone@3:633/10 to All on Sunday, November 15, 2026 20:30:02
    There was an IEEE statistic, at one time, to the effect that 70% of them are not "working" in engineering after 5 years.
    I don't put that on the students; they're getting massively ripped off. It's the criminal faculty and university administrations. And to think some of them are graduating with $$$200K student debt.

    --
    For full context, visit https://www.electrondepot.com/electrodesign/micro-oft-what-s-inside-4423912-.htm


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Monday, November 16, 2026 15:09:21
    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more
    difficult. The US hasn't got as big a talent pool as China to dredge for
    people who might be able to fix the problems, and while both countries
    are educating most of the rising generation, rich and powerful parents
    in both countries have ways of ensuring that their kids get the best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their graduate
    students weren't great at presenting new ideas - taking teachers too
    seriously can be as much of problem as not taking them seriously enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the
    dross and get the good stuff up to date.

    It does require the kind of active participation that would have been
    beyond you.

    And somehow, after four years of that, they dress up in silly caps and
    gowns [1] and get a piece of paper that magically makes them able to
    think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with
    black and white film in his camera so the scarlet Ph.D. robe didn't get recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and
    have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that
    they can communicate at the dumbed-down level that you can handle.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics,
    but you can't talk about it all that well, which is where your 2%
    probably comes from.

    [1] I skipped my ceremony. Who wants to stand for hours in the New
    Orleans heat and humidity, in the sun, in a rented black robe,
    listening to boring speeches?

    Family.

    --
    Bill Sloman, Sydney


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Monday, November 16, 2026 08:05:02
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more
    difficult. The US hasn't got as big a talent pool as China to dredge for >>> people who might be able to fix the problems, and while both countries
    are educating most of the rising generation, rich and powerful parents
    in both countries have ways of ensuring that their kids get the best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their graduate >>> students weren't great at presenting new ideas - taking teachers too
    seriously can be as much of problem as not taking them seriously enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the
    dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students
    can. That's obviously absurd.

    It does require the kind of active participation that would have been
    beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and
    gowns [1] and get a piece of paper that magically makes them able to
    think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I >rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with
    black and white film in his camera so the scarlet Ph.D. robe didn't get >recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have >legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and
    have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that
    they can communicate at the dumbed-down level that you can handle.


    I think that three or four of the various capstone team members may be
    keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.



    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of >communicating those various sorts of insight. You can do electronics,
    but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulaton in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.



    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Tuesday, November 17, 2026 16:34:23
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more
    difficult. The US hasn't got as big a talent pool as China to dredge for >>>> people who might be able to fix the problems, and while both countries >>>> are educating most of the rising generation, rich and powerful parents >>>> in both countries have ways of ensuring that their kids get the best of >>>> the opportunities around.

    Finland - in theory - does better, but my wife found that their graduate >>>> students weren't great at presenting new ideas - taking teachers too
    seriously can be as much of problem as not taking them seriously enough. >>>
    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the
    dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students
    can. That's obviously absurd.

    It does require the kind of active participation that would have been
    beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and
    gowns [1] and get a piece of paper that magically makes them able to
    think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with
    black and white film in his camera so the scarlet Ph.D. robe didn't get
    recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and
    have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that
    they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin
    they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics,
    but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind
    of simulation you'd run in an FPGA should be paralleled. Any FPGA can
    run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.

    --
    Bill Sloman, Sydney


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Jeroen Belleman@3:633/10 to All on Tuesday, November 17, 2026 09:25:33
    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org> >>>> wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more
    difficult. The US hasn't got as big a talent pool as China to
    dredge for
    people who might be able to fix the problems, and while both countries >>>>> are educating most of the rising generation, rich and powerful parents >>>>> in both countries have ways of ensuring that their kids get the
    best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their
    graduate
    students weren't great at presenting new ideas - taking teachers too >>>>> seriously can be as much of problem as not taking them seriously
    enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the
    dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students
    can. That's obviously absurd.

    It does require the kind of active participation that would have been
    beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and >>>> gowns [1] and get a piece of paper that magically makes them able to
    think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I >>> rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with
    black and white film in his camera so the scarlet Ph.D. robe didn't get
    recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have >>> legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and >>>> have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that
    they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be
    keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics,
    but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind
    of simulation you'd run in an FPGA should be paralleled. Any FPGA can
    run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.


    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    Jeroen Belleman

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Wednesday, November 18, 2026 00:07:45
    On 17/11/2026 7:25 pm, Jeroen Belleman wrote:
    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>> wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>> difficult. The US hasn't got as big a talent pool as China to
    dredge for
    people who might be able to fix the problems, and while both
    countries
    are educating most of the rising generation, rich and powerful
    parents
    in both countries have ways of ensuring that their kids get the
    best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their
    graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>> seriously can be as much of problem as not taking them seriously
    enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the
    dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students
    can. That's obviously absurd.

    It does require the kind of active participation that would have been
    beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and >>>>> gowns [1] and get a piece of paper that magically makes them able to >>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc,
    if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with
    black and white film in his camera so the scarlet Ph.D. robe didn't get >>>> recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does
    have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and >>>>> have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that
    they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be
    keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin
    they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics,
    but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The
    kind of simulation you'd run in an FPGA should be paralleled. Any FPGA
    can run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.

    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    It would probably work well, but it you went to the trouble of working
    out what could be executed in parallel it could probably be run quite a
    bit faster.

    --
    Bill Sloman, Sydney


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Tuesday, November 17, 2026 07:43:30
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>> wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>> difficult. The US hasn't got as big a talent pool as China to
    dredge for
    people who might be able to fix the problems, and while both countries >>>>>> are educating most of the rising generation, rich and powerful parents >>>>>> in both countries have ways of ensuring that their kids get the
    best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their
    graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>> seriously can be as much of problem as not taking them seriously
    enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU.

    Times haven't changed much since I was an undergrad: listen to the
    lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the
    dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students
    can. That's obviously absurd.

    It does require the kind of active participation that would have been
    beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and >>>>> gowns [1] and get a piece of paper that magically makes them able to >>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I >>>> rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with
    black and white film in his camera so the scarlet Ph.D. robe didn't get >>>> recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have >>>> legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and >>>>> have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that
    they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be
    keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin
    they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics,
    but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind
    of simulation you'd run in an FPGA should be paralleled. Any FPGA can
    run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.


    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    Jeroen Belleman

    LT Spice is a compiler that maps the schematic to x86 machine code,
    but that's still a procedural language that executes sequentially. It
    has to do all sorts of math tricks to be stable at, equivalently, low
    compute clock rates.

    An FPGA can run massively parallel, a computation every clock for
    every node in the circuit. That wouldn't need the fancy convergence
    tricks.

    I have Spice sims that run PPMs of real time, sometimes nanoseconds or
    even picoseconds per second... erratically That can be really
    annoying. An FPGA sim would flip that over and run thousands of times
    faster than real time.

    LT Spice allows one to set a maximum time step, but not a minimum,
    much less a fixed one.

    I've proposed using a DDS, or at least an NCO, to program the
    execution rate in our FPGA sims.


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Jeroen Belleman@3:633/10 to All on Tuesday, November 17, 2026 18:49:28
    On 11/17/26 14:07, Bill Sloman wrote:
    On 17/11/2026 7:25 pm, Jeroen Belleman wrote:
    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman
    <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>>> difficult. The US hasn't got as big a talent pool as China to
    dredge for
    people who might be able to fix the problems, and while both
    countries
    are educating most of the rising generation, rich and powerful
    parents
    in both countries have ways of ensuring that their kids get the >>>>>>> best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their >>>>>>> graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>>> seriously can be as much of problem as not taking them seriously >>>>>>> enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU. >>>>>>
    Times haven't changed much since I was an undergrad: listen to the >>>>>> lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the >>>>> dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students >>>> can. That's obviously absurd.

    It does require the kind of active participation that would have been >>>>> beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps >>>>>> and
    gowns [1] and get a piece of paper that magically makes them able to >>>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc, >>>>> if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with >>>>> black and white film in his camera so the scarlet Ph.D. robe didn't >>>>> get
    recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does >>>>> have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity >>>>>> and
    have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that >>>>> they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be >>>> keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John
    Larkin
    they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics, >>>>> but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The
    kind of simulation you'd run in an FPGA should be paralleled. Any
    FPGA can run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.

    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    It would probably work well, but it you went to the trouble of working
    out what could be executed in parallel it could probably be run quite a
    bit faster.


    You're still thinking in terms of parallelizing a single-thread
    sequential program. Forget that. Circuits quite naturally divide
    up into parallel components and multiple nodes, and FPGAs are good
    at running all that at the same time.

    A circuit is basically a bunch of simultaneous differential
    equations. FPGAs can do that very well. I had this PLL that I
    translated into a collection of discrete-time discrete-value
    differential equations that got implemented in a Xilinx FPGA.
    It worked very well. In fact, I had three per FPGA, all running
    at once.

    Jeroen Belleman


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Nioclásán Cóilín Glostéir@3:633/10 to All on Wednesday, November 18, 2026 00:58:55
    Dear Sergey Kubushyn:

    That webpage interetingly has:
    "How Microsoft Vaporized a Trillion Dollars
    Inside the complacency and decisions that eroded trust in Azure?from a
    former Azure Core engineer.
    AXEL RIETSCHIN".

    Thanks.
    (S. HTTP://Gloucester.Insomnia247.NL/ fuer Kontaktdaten!)

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Nioclásán Cóilín Glostéir@3:633/10 to All on Wednesday, November 18, 2026 01:22:34
    Someone wrote: |---------------------------------------------------------------------|
    |"There was an IEEE statistic, at one time, to the effect that 70% of |
    |them are not "working" in engineering after 5 years. |
    |I don't put that on the students; they're getting massively ripped |
    |off. It's the criminal faculty and university administrations. And to|
    |think some of them are graduating with $$$200K student debt." | |---------------------------------------------------------------------|

    Cf. HTTP://Gloucester.Insomnia247.NL/The_New-York_Times/2026-11-11a_Degrees_are_too_expensive.png
    (S. HTTP://Gloucester.Insomnia247.NL/ fuer Kontaktdaten!)

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Wednesday, November 18, 2026 14:48:24
    On 18/11/2026 4:49 am, Jeroen Belleman wrote:
    On 11/17/26 14:07, Bill Sloman wrote:
    On 17/11/2026 7:25 pm, Jeroen Belleman wrote:
    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman
    <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>>>> difficult. The US hasn't got as big a talent pool as China to >>>>>>>> dredge for
    people who might be able to fix the problems, and while both
    countries
    are educating most of the rising generation, rich and powerful >>>>>>>> parents
    in both countries have ways of ensuring that their kids get the >>>>>>>> best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their >>>>>>>> graduate
    students weren't great at presenting new ideas - taking teachers >>>>>>>> too
    seriously can be as much of problem as not taking them seriously >>>>>>>> enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU. >>>>>>>
    Times haven't changed much since I was an undergrad: listen to the >>>>>>> lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the >>>>>> dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students >>>>> can. That's obviously absurd.

    It does require the kind of active participation that would have been >>>>>> beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH >>>>> ships, pipeline supervisory controls, modems, flight hardware for the >>>>> C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly
    caps and
    gowns [1] and get a piece of paper that magically makes them able to >>>>>>> think.

    The university of Melbourne let me graduate in absentia as an
    M.Sc, if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation >>>>>> ceremony, which my parents enjoyed watching. My father showed up with >>>>>> black and white film in his camera so the scarlet Ph.D. robe
    didn't get
    recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It
    does have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand
    electricity and
    have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that >>>>>> they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be >>>>> keepers. Two seem to be geniuses. Student projects are a great way to >>>>> do, basically, extended interviews.

    But if you are interviewing the candidatates to work out how like John >>>> Larkin they are there's not a lot point in an extended interview.

    Thinking about electronics in a productive way can take place in a >>>>>> variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics, >>>>>> but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real >>>>> time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The
    kind of simulation you'd run in an FPGA should be paralleled. Any
    FPGA can run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just
    have been difficult to explain to an unsophisticated audience.

    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    It would probably work well, but it you went to the trouble of working
    out what could be executed in parallel it could probably be run quite
    a bit faster.


    You're still thinking in terms of parallelizing a single-thread
    sequential program. Forget that. Circuits quite naturally divide
    up into parallel components and multiple nodes, and FPGAs are good
    at running all that at the same time.

    A circuit is basically a bunch of simultaneous differential
    equations.

    Well put.

    FPGAs can do that very well. I had this PLL that I
    translated into a collection of discrete-time discrete-value
    differential equations that got implemented in a Xilinx FPGA.
    It worked very well. In fact, I had three per FPGA, all running
    at once.

    Sadly Spice uses a finite difference approach to integrating the
    differential equations in a single thread sequential program, and all I
    was pointing out that this wasn't a good place from which to start
    developing a multi-threaded solution.

    --
    Bill Sloman, Sydney

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Wednesday, November 18, 2026 15:08:26
    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>>> wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>>> difficult. The US hasn't got as big a talent pool as China to
    dredge for
    people who might be able to fix the problems, and while both countries >>>>>>> are educating most of the rising generation, rich and powerful parents >>>>>>> in both countries have ways of ensuring that their kids get the
    best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their
    graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>>> seriously can be as much of problem as not taking them seriously >>>>>>> enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU. >>>>>>
    Times haven't changed much since I was an undergrad: listen to the >>>>>> lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the >>>>> dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students >>>> can. That's obviously absurd.

    It does require the kind of active participation that would have been >>>>> beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH
    ships, pipeline supervisory controls, modems, flight hardware for the
    C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and >>>>>> gowns [1] and get a piece of paper that magically makes them able to >>>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I >>>>> rent the Ph.D. gown and sat on the podium for the Ph.D. graduation
    ceremony, which my parents enjoyed watching. My father showed up with >>>>> black and white film in his camera so the scarlet Ph.D. robe didn't get >>>>> recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have >>>>> legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and >>>>>> have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that >>>>> they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be >>>> keepers. Two seem to be geniuses. Student projects are a great way to
    do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin >>> they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a
    variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics, >>>>> but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real
    time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind >>> of simulation you'd run in an FPGA should be paralleled. Any FPGA can
    run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.


    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    LT Spice is a compiler that maps the schematic to x86 machine code,
    but that's still a procedural language that executes sequentially. It
    has to do all sorts of math tricks to be stable at, equivalently, low
    compute clock rates.

    An FPGA can run massively parallel, a computation every clock for
    every node in the circuit. That wouldn't need the fancy convergence
    tricks.

    The crucial point is that Spice is a finite difference implementation of numerical integration. The "fancy convergence" tricks just get around
    the fact that numerical integration is unstable if the step size gets
    too big.

    The instability isn't in the sequential operation, but rather in the
    finite differences (which can be too big to allow stable numerical
    integration for short time constant differential equations).

    I have Spice sims that run PPMs of real time, sometimes nanoseconds or
    even picoseconds per second... erratically That can be really
    annoying. An FPGA sim would flip that over and run thousands of times
    faster than real time.

    It wouldn't. The instability is a feature of the finite difference
    approach, not the sequential operation.

    LT Spice allows one to set a maximum time step, but not a minimum,
    much less a fixed one.

    "Numerical integration is unstable if the step size gets too big."

    I've proposed using a DDS, or at least an NCO, to program the
    execution rate in our FPGA sims.

    If you want to control the timing of the signals that come out of your
    sim, that is essential. An NCO ought to be infinitely flexible up to it maximum clock rate, but it is harder to get a precisely defined clock
    rate out of it.

    --
    Bill Sloman, Sydney

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Tuesday, November 17, 2026 21:57:51
    On Wed, 18 Nov 2026 15:08:26 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>>>> wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>>>> difficult. The US hasn't got as big a talent pool as China to
    dredge for
    people who might be able to fix the problems, and while both countries >>>>>>>> are educating most of the rising generation, rich and powerful parents >>>>>>>> in both countries have ways of ensuring that their kids get the >>>>>>>> best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their >>>>>>>> graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>>>> seriously can be as much of problem as not taking them seriously >>>>>>>> enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU. >>>>>>>
    Times haven't changed much since I was an undergrad: listen to the >>>>>>> lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the >>>>>> dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students >>>>> can. That's obviously absurd.

    It does require the kind of active participation that would have been >>>>>> beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH >>>>> ships, pipeline supervisory controls, modems, flight hardware for the >>>>> C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and >>>>>>> gowns [1] and get a piece of paper that magically makes them able to >>>>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I >>>>>> rent the Ph.D. gown and sat on the podium for the Ph.D. graduation >>>>>> ceremony, which my parents enjoyed watching. My father showed up with >>>>>> black and white film in his camera so the scarlet Ph.D. robe didn't get >>>>>> recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have >>>>>> legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and >>>>>>> have a chance of eventually designing electronics, or at least
    assisting.

    2% of them may had have their perceptions crippled to the point that >>>>>> they can communicate at the dumbed-down level that you can handle.

    I think that three or four of the various capstone team members may be >>>>> keepers. Two seem to be geniuses. Student projects are a great way to >>>>> do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin >>>> they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a >>>>>> variety of way. Education is more about letting you find ways of
    communicating those various sorts of insight. You can do electronics, >>>>>> but you can't talk about it all that well, which is where your 2%
    probably comes from.

    There's an interesting technical issue with one current project. It
    involves executing, basically, a Spice simulation in an FPGA in real >>>>> time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind >>>> of simulation you'd run in an FPGA should be paralleled. Any FPGA can
    run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have
    been difficult to explain to an unsophisticated audience.


    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    LT Spice is a compiler that maps the schematic to x86 machine code,
    but that's still a procedural language that executes sequentially. It
    has to do all sorts of math tricks to be stable at, equivalently, low
    compute clock rates.

    An FPGA can run massively parallel, a computation every clock for
    every node in the circuit. That wouldn't need the fancy convergence
    tricks.

    The crucial point is that Spice is a finite difference implementation of >numerical integration. The "fancy convergence" tricks just get around
    the fact that numerical integration is unstable if the step size gets
    too big.

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.


    The instability isn't in the sequential operation, but rather in the
    finite differences (which can be too big to allow stable numerical >integration for short time constant differential equations).

    I have Spice sims that run PPMs of real time, sometimes nanoseconds or
    even picoseconds per second... erratically That can be really
    annoying. An FPGA sim would flip that over and run thousands of times
    faster than real time.

    It wouldn't. The instability is a feature of the finite difference
    approach, not the sequential operation.

    A massively parallel process or can take small stable steps, but run
    billions of them a second. As a sim has more nodes, we'd just use more
    of the FPGA but not slow down.


    LT Spice allows one to set a maximum time step, but not a minimum,
    much less a fixed one.

    "Numerical integration is unstable if the step size gets too big."

    Sure, so keep the steps small. But run them fast.


    I've proposed using a DDS, or at least an NCO, to program the
    execution rate in our FPGA sims.

    If you want to control the timing of the signals that come out of your
    sim, that is essential. An NCO ought to be infinitely flexible up to it >maximum clock rate, but it is harder to get a precisely defined clock
    rate out of it.

    It averages right.

    We'll start with SPI ADCs and DACs and overlap them at 1 MHz. We can
    run the sim algorithms in parallel at any speed, 100 MHz or 100 Hz.

    I've been simulating some. It's surprising how a stepsize of 0.1 tau
    simulates an RC very well, even with dumb rectangular integration and
    no tricks. 0.02, 50 steps, is fabulously accurate.

    I've wondered why LT Spice doesn't use the graphics processor that
    most computers have. That would speed up sims by hundreds of times.

    I run some sims that take 20 minutes. That's annoying and bad for
    learning.



    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From JM@3:633/10 to All on Wednesday, November 18, 2026 06:33:45
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once. It is hard to effectively
    parallelise opperations on the sparse matrices, and you certainly
    don't get an X speedup by using X cores.

    There's plenty of good work been done on this in industry. It's all unpublished.

    Academia have been playing about with this for decades. This is an
    example paper :

    https://eprints.soton.ac.uk/id/eprint/364713/1/paper.pdf

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Thursday, November 19, 2026 01:27:54
    On 18/11/2026 4:57 pm, john larkin wrote:
    On Wed, 18 Nov 2026 15:08:26 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>>>>> wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn
    <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>>>>> difficult. The US hasn't got as big a talent pool as China to >>>>>>>>> dredge for
    people who might be able to fix the problems, and while both countries
    are educating most of the rising generation, rich and powerful parents
    in both countries have ways of ensuring that their kids get the >>>>>>>>> best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their >>>>>>>>> graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>>>>> seriously can be as much of problem as not taking them seriously >>>>>>>>> enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU. >>>>>>>>
    Times haven't changed much since I was an undergrad: listen to the >>>>>>>> lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the >>>>>>> dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students >>>>>> can. That's obviously absurd.

    It does require the kind of active participation that would have been >>>>>>> beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH >>>>>> ships, pipeline supervisory controls, modems, flight hardware for the >>>>>> C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and >>>>>>>> gowns [1] and get a piece of paper that magically makes them able to >>>>>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation >>>>>>> ceremony, which my parents enjoyed watching. My father showed up with >>>>>>> black and white film in his camera so the scarlet Ph.D. robe didn't get >>>>>>> recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and >>>>>>>> have a chance of eventually designing electronics, or at least >>>>>>>> assisting.

    2% of them may had have their perceptions crippled to the point that >>>>>>> they can communicate at the dumbed-down level that you can handle. >>>>>>
    I think that three or four of the various capstone team members may be >>>>>> keepers. Two seem to be geniuses. Student projects are a great way to >>>>>> do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin >>>>> they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a >>>>>>> variety of way. Education is more about letting you find ways of >>>>>>> communicating those various sorts of insight. You can do electronics, >>>>>>> but you can't talk about it all that well, which is where your 2% >>>>>>> probably comes from.

    There's an interesting technical issue with one current project. It >>>>>> involves executing, basically, a Spice simulation in an FPGA in real >>>>>> time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind >>>>> of simulation you'd run in an FPGA should be paralleled. Any FPGA can >>>>> run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have >>>>> been difficult to explain to an unsophisticated audience.


    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    LT Spice is a compiler that maps the schematic to x86 machine code,
    but that's still a procedural language that executes sequentially. It
    has to do all sorts of math tricks to be stable at, equivalently, low
    compute clock rates.

    An FPGA can run massively parallel, a computation every clock for
    every node in the circuit. That wouldn't need the fancy convergence
    tricks.

    The crucial point is that Spice is a finite difference implementation of
    numerical integration. The "fancy convergence" tricks just get around
    the fact that numerical integration is unstable if the step size gets
    too big.

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    It can't be too big - if it is the numerical integration routine goes
    unstable and produces meaningless results.

    The instability isn't in the sequential operation, but rather in the
    finite differences (which can be too big to allow stable numerical
    integration for short time constant differential equations).

    I have Spice sims that run PPMs of real time, sometimes nanoseconds or
    even picoseconds per second... erratically That can be really
    annoying. An FPGA sim would flip that over and run thousands of times
    faster than real time.

    It wouldn't. The instability is a feature of the finite difference
    approach, not the sequential operation.

    A massively parallel process or can take small stable steps, but run
    billions of them a second. As a sim has more nodes, we'd just use more
    of the FPGA but not slow down.

    That's the theory, and always has been. Getting it to work in practice
    doesn't seem to be quite as easy as you seem to think.

    LT Spice allows one to set a maximum time step, but not a minimum,
    much less a fixed one.

    "Numerical integration is unstable if the step size gets too big."

    Sure, so keep the steps small. But run them fast.

    I've proposed using a DDS, or at least an NCO, to program the
    execution rate in our FPGA sims.

    If you want to control the timing of the signals that come out of your
    sim, that is essential. An NCO ought to be infinitely flexible up to it
    maximum clock rate, but it is harder to get a precisely defined clock
    rate out of it.

    It averages right.

    Only if you regulate it carefully to get that result.

    We'll start with SPI ADCs and DACs and overlap them at 1 MHz. We can
    run the sim algorithms in parallel at any speed, 100 MHz or 100 Hz.

    If I remember rightly, there's a matrix inversion after every step of a
    Spice simulation. For my Ph.D. work I used numerical integration to
    simulate the chemical reaction I was looking at. The chemical reaction
    also produced temperature changes and mass flows which were coupled, and
    I set up a numerical integration program to simulate all three effects.

    It ran ten times more slowly than the simulation of just the chemical reaction, and the matrix inversion stage didn't help. It was still a lot simpler than a Spice circuit simulation and was looking at different differential equations

    But since you don't seem to know what you are doing, it's unlikely to do
    what you want it to.

    I've been simulating some. It's surprising how a stepsize of 0.1 tau simulates an RC very well, even with dumb rectangular integration and
    no tricks. 0.02, 50 steps, is fabulously accurate.

    There are lots of different tau values inside a Spice simulation, it's
    the smallest smallest time - the fastest loop -that constrains the step
    size.

    I've wondered why LT Spice doesn't use the graphics processor that
    most computers have. That would speed up sims by hundreds of times.

    If you could program the calculations to run in the graphics processor.
    Graphic processors have been around for quite a while. Somebody ought to
    have tried it.

    I run some sims that take 20 minutes. That's annoying and bad for
    learning.

    Tough. Get a faster processor, or simulate a more tractable process.

    --
    Bill Sloman, Sydney



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Wednesday, November 18, 2026 07:59:03
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:

    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously>


    It is hard to effectively
    parallelise opperations on the sparse matrices, and you certainly
    don't get an X speedup by using X cores.

    Adding threads in LT Spice doesn't seem to help much.



    There's plenty of good work been done on this in industry. It's all >unpublished.

    Academia have been playing about with this for decades. This is an
    example paper :

    https://eprints.soton.ac.uk/id/eprint/364713/1/paper.pdf

    Yes, acdemics have verbally abused me for my primitive simulations of
    shipboard propulsion systems. But they work.

    Creative title, that. Paper.pdf

    My other favorites are datasheet.pdf and resume.final.pdf.


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Wednesday, November 18, 2026 08:03:27
    On Thu, 19 Nov 2026 01:27:54 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 4:57 pm, john larkin wrote:
    On Wed, 18 Nov 2026 15:08:26 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn >>>>>>>>>>> <ksi@koi8.net> wrote:

    <snip>

    Nightmare. It's probably not fixable.

    Being the industry leader is difficult, and staying there even more >>>>>>>>>> difficult. The US hasn't got as big a talent pool as China to >>>>>>>>>> dredge for
    people who might be able to fix the problems, and while both countries
    are educating most of the rising generation, rich and powerful parents
    in both countries have ways of ensuring that their kids get the >>>>>>>>>> best of
    the opportunities around.

    Finland - in theory - does better, but my wife found that their >>>>>>>>>> graduate
    students weren't great at presenting new ideas - taking teachers too >>>>>>>>>> seriously can be as much of problem as not taking them seriously >>>>>>>>>> enough.

    I'm sponsoring three senior capstone projects, at CalPoly and SFSU. >>>>>>>>>
    Times haven't changed much since I was an undergrad: listen to the >>>>>>>>> lectures, memorize the formulas, execute the formulas.

    Graduate students have to read and review the literature, reject the >>>>>>>> dross and get the good stuff up to date.


    That's the academic party line: undergrads don't design, grad students >>>>>>> can. That's obviously absurd.

    It does require the kind of active participation that would have been >>>>>>>> beyond you.

    As an undergrad, I designed throttle and boiler controls for the LASH >>>>>>> ships, pipeline supervisory controls, modems, flight hardware for the >>>>>>> C5A.

    How about you?


    And somehow, after four years of that, they dress up in silly caps and
    gowns [1] and get a piece of paper that magically makes them able to >>>>>>>>> think.

    The university of Melbourne let me graduate in absentia as an M.Sc, if I
    rent the Ph.D. gown and sat on the podium for the Ph.D. graduation >>>>>>>> ceremony, which my parents enjoyed watching. My father showed up with >>>>>>>> black and white film in his camera so the scarlet Ph.D. robe didn't get
    recorded in its full glory.

    Nobody sane thinks that the ceremony has any actual effect. It does have
    legal significance, which is presumably magical.
    I estimate that about 2% of the CE/EE grads understand electricity and
    have a chance of eventually designing electronics, or at least >>>>>>>>> assisting.

    2% of them may had have their perceptions crippled to the point that >>>>>>>> they can communicate at the dumbed-down level that you can handle. >>>>>>>
    I think that three or four of the various capstone team members may be >>>>>>> keepers. Two seem to be geniuses. Student projects are a great way to >>>>>>> do, basically, extended interviews.

    But if you are interviewing the candidatates to work how like John Larkin
    they are there's not a lot point in an extended intrerview.

    Thinking about electronics in a productive way can take place in a >>>>>>>> variety of way. Education is more about letting you find ways of >>>>>>>> communicating those various sorts of insight. You can do electronics, >>>>>>>> but you can't talk about it all that well, which is where your 2% >>>>>>>> probably comes from.

    There's an interesting technical issue with one current project. It >>>>>>> involves executing, basically, a Spice simulation in an FPGA in real >>>>>>> time. Solving that, making it actually work before May, will be
    educational for everyone involved.

    Spice is written in a high level language for serial execution. The kind >>>>>> of simulation you'd run in an FPGA should be paralleled. Any FPGA can >>>>>> run quite a few threads in parallel.

    It sounds as if the project is ill-conceived, though it may just have >>>>>> been difficult to explain to an unsophisticated audience.


    Au contraire! I think that translating a Spice circuit description
    into FPGA code should work very well indeed. I haven't searched,
    but the software to do so very likely already exists somewhere.

    LT Spice is a compiler that maps the schematic to x86 machine code,
    but that's still a procedural language that executes sequentially. It
    has to do all sorts of math tricks to be stable at, equivalently, low
    compute clock rates.

    An FPGA can run massively parallel, a computation every clock for
    every node in the circuit. That wouldn't need the fancy convergence
    tricks.

    The crucial point is that Spice is a finite difference implementation of >>> numerical integration. The "fancy convergence" tricks just get around
    the fact that numerical integration is unstable if the step size gets
    too big.

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    It can't be too big - if it is the numerical integration routine goes >unstable and produces meaningless results.

    The instability isn't in the sequential operation, but rather in the
    finite differences (which can be too big to allow stable numerical
    integration for short time constant differential equations).

    I have Spice sims that run PPMs of real time, sometimes nanoseconds or >>>> even picoseconds per second... erratically That can be really
    annoying. An FPGA sim would flip that over and run thousands of times
    faster than real time.

    It wouldn't. The instability is a feature of the finite difference
    approach, not the sequential operation.

    A massively parallel process or can take small stable steps, but run
    billions of them a second. As a sim has more nodes, we'd just use more
    of the FPGA but not slow down.

    That's the theory, and always has been. Getting it to work in practice >doesn't seem to be quite as easy as you seem to think.

    LT Spice allows one to set a maximum time step, but not a minimum,
    much less a fixed one.

    "Numerical integration is unstable if the step size gets too big."

    Sure, so keep the steps small. But run them fast.

    I've proposed using a DDS, or at least an NCO, to program the
    execution rate in our FPGA sims.

    If you want to control the timing of the signals that come out of your
    sim, that is essential. An NCO ought to be infinitely flexible up to it
    maximum clock rate, but it is harder to get a precisely defined clock
    rate out of it.

    It averages right.

    Only if you regulate it carefully to get that result.


    I'm an engineer. We have to be careful or things will crash and people
    will get hurt.



    We'll start with SPI ADCs and DACs and overlap them at 1 MHz. We can
    run the sim algorithms in parallel at any speed, 100 MHz or 100 Hz.

    If I remember rightly, there's a matrix inversion after every step of a >Spice simulation. For my Ph.D. work I used numerical integration to
    simulate the chemical reaction I was looking at. The chemical reaction
    also produced temperature changes and mass flows which were coupled, and
    I set up a numerical integration program to simulate all three effects.

    It ran ten times more slowly than the simulation of just the chemical >reaction, and the matrix inversion stage didn't help. It was still a lot >simpler than a Spice circuit simulation and was looking at different >differential equations

    But since you don't seem to know what you are doing, it's unlikely to do >what you want it to.


    See the screen shot that I posted in a thread below. Fun. Works great.



    I've been simulating some. It's surprising how a stepsize of 0.1 tau
    simulates an RC very well, even with dumb rectangular integration and
    no tricks. 0.02, 50 steps, is fabulously accurate.

    There are lots of different tau values inside a Spice simulation, it's
    the smallest smallest time - the fastest loop -that constrains the step >size.

    I've wondered why LT Spice doesn't use the graphics processor that
    most computers have. That would speed up sims by hundreds of times.

    If you could program the calculations to run in the graphics processor. >Graphic processors have been around for quite a while. Somebody ought to >have tried it.

    I run some sims that take 20 minutes. That's annoying and bad for
    learning.

    Tough. Get a faster processor, or simulate a more tractable process.


    Get a job.

    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Jeroen Belleman@3:633/10 to All on Wednesday, November 18, 2026 18:42:17
    On 11/18/26 07:33, JM wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once. It is hard to effectively
    parallelise opperations on the sparse matrices, and you certainly
    don't get an X speedup by using X cores.

    There's plenty of good work been done on this in industry. It's all unpublished.

    Academia have been playing about with this for decades. This is an
    example paper :

    https://eprints.soton.ac.uk/id/eprint/364713/1/paper.pdf

    Hmmm. Maybe I'm naive, but I wouldn't try to directly run
    a sparsely populated matrix in an FPGA. It's wasteful.
    I'd rather divide the circuit into lots of little
    interconnected matrices, dense rather than sparse, each
    with its own little solver. After all, a circuit divides
    quite naturally into an interconnected set of components.

    It makes no sense to first build a sparse matrix describing
    the whole thing, only to try and parallelize it again after.
    That's hard, no surprise.

    Jeroen Belleman


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Thursday, November 19, 2026 18:56:48
    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be
    treated as matrix, which has to be inverted before you can go to
    calculate the next step.
    It is hard to effectively
    parallelise opperations on the sparse matrices, and you certainly
    don't get an X speedup by using X cores.

    Adding threads in LT Spice doesn't seem to help much.

    The compiler is unlikely to have done it for you.

    There's plenty of good work been done on this in industry. It's all
    unpublished.

    Academia have been playing about with this for decades. This is an
    example paper :

    https://eprints.soton.ac.uk/id/eprint/364713/1/paper.pdf

    Yes, acdemics have verbally abused me for my primitive simulations of shipboard propulsion systems. But they work.

    They presumably worked well enough for the job you wrote them to do. You probably didn't explain what you'd done all that clearly.

    Creative title, that. Paper.pdf

    My other favorites are datasheet.pdf and resume.final.pdf.

    You have to write a lot of papers before unambiguous and informative
    titles get to be worth extra effort.

    --
    Bill Sloman, Sydney


    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Thursday, November 19, 2026 20:58:36
    On 19/11/2026 3:03 am, john larkin wrote:
    On Thu, 19 Nov 2026 01:27:54 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 4:57 pm, john larkin wrote:
    On Wed, 18 Nov 2026 15:08:26 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>>>>> wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn >>>>>>>>>>>> <ksi@koi8.net> wrote:

    <snip>

    If you want to control the timing of the signals that come out of your >>>> sim, that is essential. An NCO ought to be infinitely flexible up to it >>>> maximum clock rate, but it is harder to get a precisely defined clock
    rate out of it.

    It averages right.

    Only if you regulate it carefully to get that result.

    I'm an engineer. We have to be careful or things will crash and people
    will get hurt.

    You do keep on claiming that you are an engineer. Being careful does
    depend on you understanding what you are doing, and you claim that too.

    We'll start with SPI ADCs and DACs and overlap them at 1 MHz. We can
    run the sim algorithms in parallel at any speed, 100 MHz or 100 Hz.

    If I remember rightly, there's a matrix inversion after every step of a
    Spice simulation. For my Ph.D. work I used numerical integration to
    simulate the chemical reaction I was looking at. The chemical reaction
    also produced temperature changes and mass flows which were coupled, and
    I set up a numerical integration program to simulate all three effects.

    It ran ten times more slowly than the simulation of just the chemical
    reaction, and the matrix inversion stage didn't help. It was still a lot
    simpler than a Spice circuit simulation and was looking at different
    differential equations

    But since you don't seem to know what you are doing, it's unlikely to do
    what you want it to.

    See the screen shot that I posted in a thread below. Fun. Works great.

    To your complete satifaction. The point about Spice is that it simulates coupled processes. Simulating a single RC and going to the trouble of
    posting the result does suggest that you haven't got the first clue
    about what you ought to be trying to do.

    I've been simulating some. It's surprising how a stepsize of 0.1 tau
    simulates an RC very well, even with dumb rectangular integration and
    no tricks. 0.02, 50 steps, is fabulously accurate.

    There are lots of different tau values inside a Spice simulation, it's
    the smallest smallest time - the fastest loop -that constrains the step
    size.

    I've wondered why LT Spice doesn't use the graphics processor that
    most computers have. That would speed up sims by hundreds of times.

    If you could program the calculations to run in the graphics processor.
    Graphic processors have been around for quite a while. Somebody ought to
    have tried it.

    I run some sims that take 20 minutes. That's annoying and bad for
    learning.

    Tough. Get a faster processor, or simulate a more tractable process.

    Get a job.

    Not easy, at age 83. I've been working at it for years. Early on I got
    some job interviews, but I suspect it was more to make the interview statistics look good than out of any expectation that they might hire me.

    --
    Bill Sloman, Sydney





    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Thursday, November 19, 2026 07:40:13
    On Thu, 19 Nov 2026 18:56:48 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be >treated as matrix, which has to be inverted before you can go to
    calculate the next step.

    I never did that fancy stuff. The ships worked just like the sims.


    It is hard to effectively
    parallelise opperations on the sparse matrices, and you certainly
    don't get an X speedup by using X cores.

    Adding threads in LT Spice doesn't seem to help much.

    The compiler is unlikely to have done it for you.

    The number of threads is a menu option in LT Spice, but doesn't seem
    to make much difference. The choice of solver matters a lot more. Some
    of my sims need the gear solver to not hang up.



    There's plenty of good work been done on this in industry. It's all
    unpublished.

    Academia have been playing about with this for decades. This is an
    example paper :

    https://eprints.soton.ac.uk/id/eprint/364713/1/paper.pdf

    Yes, acdemics have verbally abused me for my primitive simulations of
    shipboard propulsion systems. But they work.

    They presumably worked well enough for the job you wrote them to do. You >probably didn't explain what you'd done all that clearly.

    Sure I explained it, which is why experts complained.

    In a control system, one has, essentially, zero-output-impedance
    blocks here and there, unilateral elements, which don't need fancy
    matrix algorithms to work.

    Modeling the propeller and the hull dynamics was more like having
    interacting nodes, but the time steps could be 1/1000 of the time
    constants, so that just worked. 120 RPM was flat out on LASH ships and
    DDs.


    Creative title, that. Paper.pdf

    My other favorites are datasheet.pdf and resume.final.pdf.

    You have to write a lot of papers before unambiguous and informative
    titles get to be worth extra effort.

    Like your name in the resume? Or the part number in the data sheet?

    TI must encrypt part numbers into pdf file names somehow.


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Thursday, November 19, 2026 07:44:44
    On Thu, 19 Nov 2026 20:58:36 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 19/11/2026 3:03 am, john larkin wrote:
    On Thu, 19 Nov 2026 01:27:54 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 4:57 pm, john larkin wrote:
    On Wed, 18 Nov 2026 15:08:26 +1100, Bill Sloman <bill.sloman@ieee.org> >>>> wrote:

    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn >>>>>>>>>>>>> <ksi@koi8.net> wrote:

    <snip>

    If you want to control the timing of the signals that come out of your >>>>> sim, that is essential. An NCO ought to be infinitely flexible up to it >>>>> maximum clock rate, but it is harder to get a precisely defined clock >>>>> rate out of it.

    It averages right.

    Only if you regulate it carefully to get that result.

    I'm an engineer. We have to be careful or things will crash and people
    will get hurt.

    You do keep on claiming that you are an engineer. Being careful does
    depend on you understanding what you are doing, and you claim that too.

    We'll start with SPI ADCs and DACs and overlap them at 1 MHz. We can
    run the sim algorithms in parallel at any speed, 100 MHz or 100 Hz.

    If I remember rightly, there's a matrix inversion after every step of a
    Spice simulation. For my Ph.D. work I used numerical integration to
    simulate the chemical reaction I was looking at. The chemical reaction
    also produced temperature changes and mass flows which were coupled, and >>> I set up a numerical integration program to simulate all three effects.

    It ran ten times more slowly than the simulation of just the chemical
    reaction, and the matrix inversion stage didn't help. It was still a lot >>> simpler than a Spice circuit simulation and was looking at different
    differential equations

    But since you don't seem to know what you are doing, it's unlikely to do >>> what you want it to.

    See the screen shot that I posted in a thread below. Fun. Works great.

    To your complete satifaction. The point about Spice is that it simulates >coupled processes. Simulating a single RC and going to the trouble of >posting the result does suggest that you haven't got the first clue
    about what you ought to be trying to do.

    I'm working with a team of five college seniors to design and build a
    box that simulates aerospace sensors electrically, in real time. We're basically moving Spice type algorithms into FPGA hardware.

    I'm actually more interested in evaluating the kids than the
    algorithms.


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Lasse Langwadt@3:633/10 to All on Thursday, November 19, 2026 17:12:12
    On 11/15/26 01:15, Don Y wrote:
    On 11/14/2026 3:18 PM, Sergey Kubushyn wrote:
    and why the entire company, not just Azure and stuff, should've been long
    shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad
    accident. It
    is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying
    giant. It
    is terminal and uncurable. The giant is doomed.
    MS has never been a "tech powerhouse".ÿ Their skill has been in cornering
    a market.

    A market that has now slipped away from them -- alternative OSs (esp
    FOSS offerings) and platforms.

    Their hardware has always been laughable (though some of the surface offerings are tolerable but don't really stand out among similar
    offerings in that market segment)

    Microsoft phones??ÿ PDAs??

    They have spent so much time tied to the past that they have missed out
    on the future.

    majority of MS revenue is cloud and a majority of the VMs in Azure is
    running linux

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Don Y@3:633/10 to All on Thursday, November 19, 2026 12:01:59
    On 11/19/2026 9:12 AM, Lasse Langwadt wrote:
    On 11/15/26 01:15, Don Y wrote:
    On 11/14/2026 3:18 PM, Sergey Kubushyn wrote:
    and why the entire company, not just Azure and stuff, should've been long >>> shot dead to put it out of its misery.

    https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion >>>
    Higly recommended reading. Long, in 6 parts, but very informative.

    And clearly telling that this massive failure is not just a bad accident. It
    is just a beginning of the fall of the empire.

    There is NOTHING that could be done to fix it and cure the dying giant. It >>> is terminal and uncurable. The giant is doomed.
    MS has never been a "tech powerhouse".ÿ Their skill has been in cornering
    a market.

    A market that has now slipped away from them -- alternative OSs (esp
    FOSS offerings) and platforms.

    Their hardware has always been laughable (though some of the surface
    offerings are tolerable but don't really stand out among similar
    offerings in that market segment)

    Microsoft phones??ÿ PDAs??

    They have spent so much time tied to the past that they have missed out on >> the future.

    majority of MS revenue is cloud and a majority of the VMs in Azure is running
    linux

    And, if you read the article, you would see why they're headed in the wrong direction.

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Friday, November 20, 2026 15:57:13
    On 20/11/2026 2:44 am, john larkin wrote:
    On Thu, 19 Nov 2026 20:58:36 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 19/11/2026 3:03 am, john larkin wrote:
    On Thu, 19 Nov 2026 01:27:54 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 18/11/2026 4:57 pm, john larkin wrote:
    On Wed, 18 Nov 2026 15:08:26 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>> wrote:

    On 18/11/2026 2:43 am, john larkin wrote:
    On Tue, 17 Nov 2026 09:25:33 +0100, Jeroen Belleman
    <jeroen@nospam.please> wrote:

    On 11/17/26 06:34, Bill Sloman wrote:
    On 17/11/2026 3:05 am, john larkin wrote:
    On Mon, 16 Nov 2026 15:09:21 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 16/11/2026 2:34 am, john larkin wrote:
    On Mon, 16 Nov 2026 00:19:14 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 15/11/2026 11:56 am, john larkin wrote:
    On Sat, 14 Nov 2026 22:18:31 -0000 (UTC), Sergey Kubushyn >>>>>>>>>>>>>> <ksi@koi8.net> wrote:

    <snip>

    If you want to control the timing of the signals that come out of your >>>>>> sim, that is essential. An NCO ought to be infinitely flexible up to it >>>>>> maximum clock rate, but it is harder to get a precisely defined clock >>>>>> rate out of it.

    It averages right.

    Only if you regulate it carefully to get that result.

    I'm an engineer. We have to be careful or things will crash and people
    will get hurt.

    You do keep on claiming that you are an engineer. Being careful does
    depend on you understanding what you are doing, and you claim that too.

    We'll start with SPI ADCs and DACs and overlap them at 1 MHz. We can >>>>> run the sim algorithms in parallel at any speed, 100 MHz or 100 Hz.

    If I remember rightly, there's a matrix inversion after every step of a >>>> Spice simulation. For my Ph.D. work I used numerical integration to
    simulate the chemical reaction I was looking at. The chemical reaction >>>> also produced temperature changes and mass flows which were coupled, and >>>> I set up a numerical integration program to simulate all three effects. >>>>
    It ran ten times more slowly than the simulation of just the chemical
    reaction, and the matrix inversion stage didn't help. It was still a lot >>>> simpler than a Spice circuit simulation and was looking at different
    differential equations

    But since you don't seem to know what you are doing, it's unlikely to do >>>> what you want it to.

    See the screen shot that I posted in a thread below. Fun. Works great.

    To your complete satisfaction. The point about Spice is that it simulates
    coupled processes. Simulating a single RC and going to the trouble of
    posting the result does suggest that you haven't got the first clue
    about what you ought to be trying to do.

    I'm working with a team of five college seniors to design and build a
    box that simulates aerospace sensors electrically, in real time. We're basically moving Spice type algorithms into FPGA hardware.

    I'm actually more interested in evaluating the kids than the
    algorithms.

    You may need a better grasp of what they are trying to do. If you don't actually understand what they are trying to, you may have a problem
    working out how successful they have been at doing it.

    Spice is a very specific sort of numerical integration program,
    integrating a bunch of coupled differential equations. You don't seem to understand what it does or the detail of how it does it.

    See if you can hire Mike Engelhart of LTSpice fame - he may not be a mathematician, as back in the early 1990s he worked on the Schlumberger electron beam tester, but he may have learned enough math to be useful.

    --
    Bill Sloman, Sydney


    --
    Bill Sloman, Sydney

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Friday, November 20, 2026 16:58:35
    On 20/11/2026 2:40 am, john larkin wrote:
    On Thu, 19 Nov 2026 18:56:48 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    Right. And a procedural program runs slow, crunching one node at a
    time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be
    treated as matrix, which has to be inverted before you can go to
    calculate the next step.

    I never did that fancy stuff. The ships worked just like the sims.

    My Ph.D. work did rely on integrating a single differential equation and
    all the experimental results written up in the thesis reflect this.

    I did go to the trouble of setting up system of three differential
    equations to track the heat flows and the mass flows, and it produced
    pretty much the same results at one tenth of the speed, so I didn't
    bother using it on every experiment run.

    It is hard to effectively
    parallelise opperations on the sparse matrices, and you certainly
    don't get an X speedup by using X cores.

    Adding threads in LT Spice doesn't seem to help much.

    The compiler is unlikely to have done it for you.

    The number of threads is a menu option in LT Spice, but doesn't seem
    to make much difference. The choice of solver matters a lot more. Some
    of my sims need the gear solver to not hang up.

    It happens.

    There's plenty of good work been done on this in industry. It's all
    unpublished.

    Academia have been playing about with this for decades. This is an
    example paper :

    https://eprints.soton.ac.uk/id/eprint/364713/1/paper.pdf

    Yes, acdemics have verbally abused me for my primitive simulations of
    shipboard propulsion systems. But they work.

    They presumably worked well enough for the job you wrote them to do. You
    probably didn't explain what you'd done all that clearly.

    Sure I explained it, which is why experts complained.

    They would complain if you didn't explain it clearly enough. Academic
    are teachers, and their stock in trade is correct explanations which are
    easy for the students to understand.

    In a control system, one has, essentially, zero-output-impedance
    blocks here and there, unilateral elements, which don't need fancy
    matrix algorithms to work.

    But they can have stability problems when the blocks interact in
    inconvenient ways. You got lucky.

    Modeling the propeller and the hull dynamics was more like having
    interacting nodes, but the time steps could be 1/1000 of the time
    constants, so that just worked. 120 RPM was flat out on LASH ships and
    DDs.


    Creative title, that. Paper.pdf

    My other favorites are datasheet.pdf and resume.final.pdf.

    You have to write a lot of papers before unambiguous and informative
    titles get to be worth extra effort.

    Like your name in the resume? Or the part number in the data sheet?

    The titles of my various resumes on this computer include the date but
    not my name.

    TI must encrypt part numbers into .pdf file names somehow.

    Every manufacturer does. "T1G4020036-FL Datasheet.pdf" is what I got
    from Quorvo for their dual GaN transistor.

    You need a name long enough to let you distinguish between various .pdf
    files. The more you have, the longer the name has to be.

    --
    Bill Sloman, Sydney



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Friday, November 20, 2026 10:20:44
    On Fri, 20 Nov 2026 16:58:35 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 20/11/2026 2:40 am, john larkin wrote:
    On Thu, 19 Nov 2026 18:56:48 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com> >>>>> wrote:

    Right. And a procedural program runs slow, crunching one node at a >>>>>> time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be
    treated as matrix, which has to be inverted before you can go to
    calculate the next step.

    I never did that fancy stuff. The ships worked just like the sims.

    My Ph.D. work did rely on integrating a single differential equation and
    all the experimental results written up in the thesis reflect this.

    I did go to the trouble of setting up system of three differential
    equations to track the heat flows and the mass flows, and it produced
    pretty much the same results at one tenth of the speed, so I didn't
    bother using it on every experiment run.

    Heat flow is easy. Just divide the thing into points or cubes and keep averaging every element's temperature with its neighbors.

    ATLC does the same sort of thing with voltages and conductors and
    insulators, to simulate complex transmission line geometries.

    There is probably a similar sheet-resistance calculator somewhere. I
    need to do that now and then, which I do with a sheet of copperclad
    FR4 and some test equipment.



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Saturday, November 21, 2026 15:39:53
    On 21/11/2026 5:20 am, john larkin wrote:
    On Fri, 20 Nov 2026 16:58:35 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 20/11/2026 2:40 am, john larkin wrote:
    On Thu, 19 Nov 2026 18:56:48 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com> >>>>>> wrote:

    Right. And a procedural program runs slow, crunching one node at a >>>>>>> time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be
    treated as matrix, which has to be inverted before you can go to
    calculate the next step.

    I never did that fancy stuff. The ships worked just like the sims.

    My Ph.D. work did rely on integrating a single differential equation and
    all the experimental results written up in the thesis reflect this.

    I did go to the trouble of setting up system of three differential
    equations to track the heat flows and the mass flows, and it produced
    pretty much the same results at one tenth of the speed, so I didn't
    bother using it on every experiment run.

    Heat flow is easy. Just divide the thing into points or cubes and keep averaging every element's temperature with its neighbors.

    ATLC does the same sort of thing with voltages and conductors and
    insulators, to simulate complex transmission line geometries.

    There is probably a similar sheet-resistance calculator somewhere. I
    need to do that now and then, which I do with a sheet of copper-clad
    FR4 and some test equipment.

    You miss the point that the differential equations involved are coupled,
    and simulating the whole system creates the same sorts of instability
    problems that Spice has to deal with.

    As usual, you snipped a lot of what I'd posted, and didn't mark the
    snip. Since you don't seem to understand the point I'm trying to make,
    this would have been an attractive option, if a bit silly.

    --
    Bill Sloman, Sydney

    --
    Bill Sloman, Sydney



    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From john larkin@3:633/10 to All on Saturday, November 21, 2026 07:19:35
    On Sat, 21 Nov 2026 15:39:53 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 21/11/2026 5:20 am, john larkin wrote:
    On Fri, 20 Nov 2026 16:58:35 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 20/11/2026 2:40 am, john larkin wrote:
    On Thu, 19 Nov 2026 18:56:48 +1100, Bill Sloman <bill.sloman@ieee.org> >>>> wrote:

    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com> >>>>>>> wrote:

    Right. And a procedural program runs slow, crunching one node at a >>>>>>>> time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be >>>>> treated as matrix, which has to be inverted before you can go to
    calculate the next step.

    I never did that fancy stuff. The ships worked just like the sims.

    My Ph.D. work did rely on integrating a single differential equation and >>> all the experimental results written up in the thesis reflect this.

    I did go to the trouble of setting up system of three differential
    equations to track the heat flows and the mass flows, and it produced
    pretty much the same results at one tenth of the speed, so I didn't
    bother using it on every experiment run.

    Heat flow is easy. Just divide the thing into points or cubes and keep
    averaging every element's temperature with its neighbors.

    ATLC does the same sort of thing with voltages and conductors and
    insulators, to simulate complex transmission line geometries.

    There is probably a similar sheet-resistance calculator somewhere. I
    need to do that now and then, which I do with a sheet of copper-clad
    FR4 and some test equipment.

    You miss the point that the differential equations involved are coupled,
    and simulating the whole system creates the same sorts of instability >problems that Spice has to deal with.

    The thermal model is a 3d mesh of 1st order RCs. The simple iteration
    settles down nicely.

    It's interesting (well, to some of us) that thermal systems never
    oscillate, never ring.

    ATLC is wonderful. It was written by an amateur who didn't know much
    theory.


    John Larkin
    Highland Tech Glen Canyon Design Center
    Lunatic Fringe Electronics

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From JM@3:633/10 to All on Sunday, November 22, 2026 04:07:24
    On Sat, 21 Nov 2026 07:19:35 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    ATLC is wonderful. It was written by an amateur who didn't know much
    theory.

    Really? Did Dave get his three engineering degrees from some mail
    order university?

    Not that he needed any theory, the original version of atlc was just a
    C port of the example Fortran inplementation of the Gauss-Seidel
    procedure listed Dworsky's transmission line book.

    Must be at least 30 lines of code.

    You're just as quick inputing the 2-D outline of the line conductors
    as fixed potentials in cells on a spreadsheet and doing the
    calculations there.

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Sunday, November 22, 2026 15:22:44
    On 22/11/2026 2:19 am, john larkin wrote:
    On Sat, 21 Nov 2026 15:39:53 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 21/11/2026 5:20 am, john larkin wrote:
    On Fri, 20 Nov 2026 16:58:35 +1100, Bill Sloman <bill.sloman@ieee.org>
    wrote:

    On 20/11/2026 2:40 am, john larkin wrote:
    On Thu, 19 Nov 2026 18:56:48 +1100, Bill Sloman <bill.sloman@ieee.org> >>>>> wrote:

    On 19/11/2026 2:59 am, john larkin wrote:
    On Wed, 18 Nov 2026 06:33:45 +0000, JM
    <sunaecoNoChoppedPork@gmail.com> wrote:
    On Tue, 17 Nov 2026 21:57:51 -0800, john larkin <jl@glen--canyon.com> >>>>>>>> wrote:

    Right. And a procedural program runs slow, crunching one node at a >>>>>>>>> time, so step size has to be big or it runs all afternoon.

    No, all nodes are calculated at once.

    How can an X86, or any other CPU, crunch a hundred node calcs
    simultaneously.

    Lots of them in parallel can. But the results of each step have to be >>>>>> treated as matrix, which has to be inverted before you can go to
    calculate the next step.

    I never did that fancy stuff. The ships worked just like the sims.

    My Ph.D. work did rely on integrating a single differential equation and >>>> all the experimental results written up in the thesis reflect this.

    I did go to the trouble of setting up system of three differential
    equations to track the heat flows and the mass flows, and it produced
    pretty much the same results at one tenth of the speed, so I didn't
    bother using it on every experiment run.

    Heat flow is easy. Just divide the thing into points or cubes and keep
    averaging every element's temperature with its neighbors.

    ATLC does the same sort of thing with voltages and conductors and
    insulators, to simulate complex transmission line geometries.

    There is probably a similar sheet-resistance calculator somewhere. I
    need to do that now and then, which I do with a sheet of copper-clad
    FR4 and some test equipment.

    You miss the point that the differential equations involved are coupled,
    and simulating the whole system creates the same sorts of instability
    problems that Spice has to deal with.

    The thermal model is a 3d mesh of 1st order RCs. The simple iteration
    settles down nicely.

    It's interesting (well, to some of us) that thermal systems never
    oscillate, never ring.

    They don't have gain.

    <snip>

    --
    Bill Sloman, Sydney

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Bill Sloman@3:633/10 to All on Sunday, November 22, 2026 17:03:53
    On 22/11/2026 3:07 pm, JM wrote:
    On Sat, 21 Nov 2026 07:19:35 -0800, john larkin <jl@glen--canyon.com>
    wrote:

    ATLC is wonderful. It was written by an amateur who didn't know much
    theory.

    Really? Did Dave get his three engineering degrees from some mail
    order university?

    Not that he needed any theory, the original version of atlc was just a
    C port of the example Fortran inplementation of the Gauss-Seidel
    procedure listed Dworsky's transmission line book.

    Must be at least 30 lines of code.

    That has to be one of the devastating responses posted here.

    John Larkin doesn't seem to realise quite how remarkably ignorant he can
    be. He may be a able to shrug off this comment too, but he'd be wise not to.

    You're just as quick inputing the 2-D outline of the line conductors
    as fixed potentials in cells on a spreadsheet and doing the
    calculations there.

    The user interface is a bit clumsy. The computer manager at Haffmans BV
    in Venlo was a bit parnoid about downloading special purpose software,
    and I did a lot of complicated stuff in Excel. It got tedious.

    --
    Bill Sloman, Sydney

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)