Martin Brown wrote:> On 26/10/2022 20:29, Phil Hobbs wrote: >> John Larkin wrote: >>> On Wed, 26 Oct 2022 08:56:04 +0100, Martin Brown >>> <'''newspam'''@nonad.co.uk> wrote: >>> >>>> On 26/10/2022 04:41, John Larkin wrote: >>>>> On Tue, 25 Oct 2022 21:52:46 +0100, Martin Brown >>>>> <'''newspam'''@nonad.co.uk> wrote: >>>>> >>>> >>>>>> Chances are one or more of the equations is stiff and the >>>>>> time step is becoming infinitessimal on one of the rapid >>>>>> transitions. Adding a bit of spurious dissipation 1M to >>>>>> ground here and there might take the edge off whatever is >>>>>> making it so stiff. >>>>> >>>>> We are running TI's Cadence sim and TI's encrypted switcher >>>>> chip models. >>>>> >>>>> We just discovered that their TPS562208 model runs about 50x >>>>> faster than the TPS54302 model. They are very similar chips. >>>>> We have both in our power supply design, one 54302 >>>>> pre-regulating for three of the 562208's. >>>> >>>> It might be worth building one to see if it really is inclined >>>> to squeg in real circuits. The sim could be telling you >>>> something important. >>> >>> If we can't reasonably sim it, we'll build it. I can imagine the >>> three secondary switchers, negative impedance loads, making the >>> first one oscillate. That's not a risk we want on our new delay >>> generator. >>> >>> Each reg needs output caps of unknown value, and feedforward caps >>> in its fedback divider. May as well get all that right. > > What might break the deadlock for computation is add a small series > resistance to the capacitance after the first regulator to low pass > filter it. The trick will be to find something modest enough to not > affect the predictions much but sufficient to compute it more > easily. > > You might have to do something like that IRL too. > > I expect they never expected you to daisy chain them back to back > like that and the output of the first one really doesn't like facing > the negative impedance dynamic load. Classic way to make an > oscillator. > >>>> Parasitic things with ridiculously high Q can ring if provoked >>>> which is why adding the odd spurious dissipative resistor to >>>> certain key nodes sometimes helps instil good behaviour. Jeroen >>>> has made the same point. >>> >>> I wish TI knew all that. >>> >> >> I suspect they give their modelers tight budgets for how many hours >> they can spend per model. > > And that they never considered the output of one driving the inputs > of several others.Yup. The misbehaviour there happens at the timescale of the loop bandwidth, so it shouldn't run pathologically slowly on that account. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
favorite Spice speedups
Started by ●October 24, 2022
Reply by ●October 27, 20222022-10-27
Reply by ●October 27, 20222022-10-27
On Thu, 27 Oct 2022 13:46:47 +0100, Martin Brown <'''newspam'''@nonad.co.uk> wrote:>On 26/10/2022 20:29, Phil Hobbs wrote: >> John Larkin wrote: >>> On Wed, 26 Oct 2022 08:56:04 +0100, Martin Brown >>> <'''newspam'''@nonad.co.uk> wrote: >>> >>>> On 26/10/2022 04:41, John Larkin wrote: >>>>> On Tue, 25 Oct 2022 21:52:46 +0100, Martin Brown >>>>> <'''newspam'''@nonad.co.uk> wrote: >>>>> >>>> >>>>>> Chances are one or more of the equations is stiff and the time step is >>>>>> becoming infinitessimal on one of the rapid transitions. Adding a >>>>>> bit of >>>>>> spurious dissipation 1M to ground here and there might take the >>>>>> edge off >>>>>> whatever is making it so stiff. >>>>> >>>>> We are running TI's Cadence sim and TI's encrypted switcher chip >>>>> models. >>>>> >>>>> We just discovered that their TPS562208 model runs about 50x faster >>>>> than the TPS54302 model. They are very similar chips. We have both in >>>>> our power supply design, one 54302 pre-regulating for three of the >>>>> 562208's. >>>> >>>> It might be worth building one to see if it really is inclined to squeg >>>> in real circuits. The sim could be telling you something important. >>> >>> If we can't reasonably sim it, we'll build it. I can imagine the three >>> secondary switchers, negative impedance loads, making the first one >>> oscillate. That's not a risk we want on our new delay generator. >>> >>> Each reg needs output caps of unknown value, and feedforward caps in >>> its fedback divider. May as well get all that right. > >What might break the deadlock for computation is add a small series >resistance to the capacitance after the first regulator to low pass >filter it. The trick will be to find something modest enough to not >affect the predictions much but sufficient to compute it more easily. > >You might have to do something like that IRL too. > >I expect they never expected you to daisy chain them back to back like >that and the output of the first one really doesn't like facing the >negative impedance dynamic load. Classic way to make an oscillator. > >>>> Parasitic things with ridiculously high Q can ring if provoked which is >>>> why adding the odd spurious dissipative resistor to certain key nodes >>>> sometimes helps instil good behaviour. Jeroen has made the same point. >>> >>> I wish TI knew all that. >>> >> >> I suspect they give their modelers tight budgets for how many hours they >> can spend per model. > >And that they never considered the output of one driving the inputs of >several others.Here's what I have so far. The first switcher sees a complex load, some of which is negative impedance. We'll split the board into noisy and quiet "Q" halves with ferrite beads bridging the power pours. That makes the impedances yet more interesting. https://www.dropbox.com/s/pklyt3mrlvf3ewm/28S662A7_power.pdf?dl=0 I hope it's not that ratty in real life, and is somewhat more efficient. This being 2022, the design is dominated by what we can get. We have managed to get the TI switchers to work in LT Spice at bearable speeds, tens of minutes per run instead of overnight. I don't trust the results, especially for the first switcher, the TPS54302. https://www.dropbox.com/s/sd5640vv8et0jb1/TPS54302_Sim_JL1.jpg?raw=1 The sim log file is full of complaints about bad curly brackets and such. I didn't draw any curly brackets. Switchers driving switchers is common. Nobody likes going from 24 volts to 1 volt, even if you can get the parts. Go to the TI or ADI web site and look at New Products. It's mind boggling. How can they support, and make good models for, all those parts? How long will they keep making them?
Reply by ●October 27, 20222022-10-27
33On a sunny day (Thu, 27 Oct 2022 08:09:08 -0700) it happened John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote in <166llh58sb6stk1rrg9au34otiomeuu6me@4ax.com>:>Here's what I have so far. The first switcher sees a complex load, >some of which is negative impedance. We'll split the board into noisy >and quiet "Q" halves with ferrite beads bridging the power pours. That >makes the impedances yet more interesting. > >https://www.dropbox.com/s/pklyt3mrlvf3ewm/28S662A7_power.pdf?dl=0 > >I hope it's not that ratty in real life, and is somewhat more >efficient. > >This being 2022, the design is dominated by what we can get. > >We have managed to get the TI switchers to work in LT Spice at >bearable speeds, tens of minutes per run instead of overnight. I don't >trust the results, especially for the first switcher, the TPS54302. > >https://www.dropbox.com/s/sd5640vv8et0jb1/TPS54302_Sim_JL1.jpg?raw=1 > >The sim log file is full of complaints about bad curly brackets and >such. I didn't draw any curly brackets.The sim has C7 1 nF to ground, not in the circuit diagram?>Switchers driving switchers is common. Nobody likes going from 24 >volts to 1 volt, even if you can get the parts.Well walwarts go from 230 V to 6 V etc, ratio even more So do my Meanwells from 230 V to 7 V Maybe the word is 'transformer' here.. Ringcores are cool: http://panteltje.com/pub/drone_power_small_core_test_detail_IMG_6115.JPG But maybe too big for that thing you have? OTOH BAD things must happen for those parts not to be around anymore That was to keep my drone airborne via a thin coax carrying high voltage transformed back to 7V at the drone: http://panteltje.com/pub/drone_power_small_core_test_IMG_6114.JPG http://panteltje.com/pub/h501s_drone_remote_power_flight_test_1_IMG_6274.JPG http://panteltje.com/pub/h501s_drone_remote_power_test_ground_control_1_IMG_6276.JPG
Reply by ●October 27, 20222022-10-27
On Thu, 27 Oct 2022 15:56:30 GMT, Jan Panteltje <pNaonStpealmtje@yahoo.com> wrote:>33On a sunny day (Thu, 27 Oct 2022 08:09:08 -0700) it happened John Larkin ><jlarkin@highlandSNIPMEtechnology.com> wrote in ><166llh58sb6stk1rrg9au34otiomeuu6me@4ax.com>: > >>Here's what I have so far. The first switcher sees a complex load, >>some of which is negative impedance. We'll split the board into noisy >>and quiet "Q" halves with ferrite beads bridging the power pours. That >>makes the impedances yet more interesting. >> >>https://www.dropbox.com/s/pklyt3mrlvf3ewm/28S662A7_power.pdf?dl=0 >> >>I hope it's not that ratty in real life, and is somewhat more >>efficient. >> >>This being 2022, the design is dominated by what we can get. >> >>We have managed to get the TI switchers to work in LT Spice at >>bearable speeds, tens of minutes per run instead of overnight. I don't >>trust the results, especially for the first switcher, the TPS54302. >> >>https://www.dropbox.com/s/sd5640vv8et0jb1/TPS54302_Sim_JL1.jpg?raw=1 >> >>The sim log file is full of complaints about bad curly brackets and >>such. I didn't draw any curly brackets. > >The sim has C7 1 nF to ground, not in the circuit diagram?I think that is just to get the sim un-stalled. As is the 100m ESR of the bootstrap cap.> > >>Switchers driving switchers is common. Nobody likes going from 24 >>volts to 1 volt, even if you can get the parts. > >Well walwarts go from 230 V to 6 V etc, ratio even more >So do my Meanwells from 230 V to 7 V >Maybe the word is 'transformer' here..Yes. High ratio transformers help.> >Ringcores are cool: > http://panteltje.com/pub/drone_power_small_core_test_detail_IMG_6115.JPG > >But maybe too big for that thing you have?I am struggling for every square mm of board area.>OTOH BAD things must happen for those parts not to be around anymoreParts seem to be available more lately, especially direct from TI.> >That was to keep my drone airborne via a thin coax carrying high voltage >transformed back to 7V at the drone: > http://panteltje.com/pub/drone_power_small_core_test_IMG_6114.JPG > http://panteltje.com/pub/h501s_drone_remote_power_flight_test_1_IMG_6274.JPG > http://panteltje.com/pub/h501s_drone_remote_power_test_ground_control_1_IMG_6276.JPGI tested a 30 ga twisted pair of magnet wire. It failed at 1400 volts. Wire-wrap wire must be 10 KV-ish. You'd need a clever HV downconverter to use really light wire. How about a single wire to power a drone?
Reply by ●October 27, 20222022-10-27
"Martin Brown" wrote in message news:tjap6l$4vg$1@gioia.aioe.org... On 26/10/2022 04:41, John Larkin wrote:> On Tue, 25 Oct 2022 21:52:46 +0100, Martin Brown > <'''newspam'''@nonad.co.uk> wrote: >>> Chances are one or more of the equations is stiff and the time step is >> becoming infinitessimal on one of the rapid transitions. Adding a bit of >> spurious dissipation 1M to ground here and there might take the edge off >> whatever is making it so stiff. > > We are running TI's Cadence sim and TI's encrypted switcher chip > models. > > We just discovered that their TPS562208 model runs about 50x faster > than the TPS54302 model. They are very similar chips. We have both in > our power supply design, one 54302 pre-regulating for three of the > 562208's.It might be worth building one to see if it really is inclined to squeg in real circuits. The sim could be telling you something important.>> We once did some awkward PDE's in a computer solution and on a Tektronix >> vector display monitor they looked fine so sent them off to the plotter. >> The job came back part done with an apologetic note from the sysop "your >> job was cancelled because the red pen began to work loose". >> >> Careful examination of the plot file showed some of the vector steps >> were just Angstroms long! > >> Sometimes a Spice sim takes femtosecond steps all afternoon. For some >> reason the Spice programs allow us to set the max time step but not >> the min.>Usually if a simulation goes to insanely short steps it is because it >cannot achieve the specified accuracy any other way (or is buggy).Yes.>That sort of behaviour is characteristic of stiff equations where it is >fighting hard to prevent the solution diverging in some bad way.I have to disagree on the claim about stiff equations.That's not often the reason in my experience. Pretty much all practical circuits are stiff, and sim ok. Very typical reasons for insanely short time steps are that the modal derivative is discontinuous. The math algorithm thus simply fails. "If else" type constructs will do this. One should use functions such as y = v(1).tanh(gain.v(test)) + v(2).(1-tanh(gain.(v(test)) to select between one voltage or the other smoothly with a continuous derivative, gain sets how sharp the tanh "switches" One need to make sure that there are no fast edges going directly into large caps with no series resistor, resulting in extremely large currents. Put a resistor across inductors. All switching edges should be RC filtered. There is a list of dos and don't such as only span 12 orders for resistance, The problem is that, and lets face the reality, most "pro" model makers don't know diddly squat about them.... The design of a good model is as hard as a good analog circuit design is. There aren't many good analog designers.... :-) Model making requires some knowledge of how spice calculates to avoid the obvious problems. I haven't looked at the model in question, but I do know that I have a basic functional current mode controller (CurrentModeController.sss) in SuperSpice running in around 5 secs, thus I find it difficult to hold that some more advanced controller model has to take many hours to run. One of longest running examples in SuperSpice is SuperHetRadioFrontEnd.sss. It takes 2 minutes on my laptop which does around 90 GFlops. For those of you that know about 60s/70s 6 transistor radios. It has an AM input driving an oscillating mixer into 3 stages of IF at 455Khz to its demodulated output. The "stiff" nature of the circuit in that it needs to have the 1 MHz running many cycles and deal with the LF modulation. Kevin Aylward https://www.kevinaylward.co.uk/gr/index.html http://www.anasoft.co.uk/ SuperSpice - Freeware - complete http://www.kevinaylward.co.uk/ee/index.html
Reply by ●October 27, 20222022-10-27
"Phil Hobbs" wrote in message news:4f5595a2-ab59-a6d9-ee64-a4eb8835295a@electrooptical.net... Joe Gwinn wrote:> On Tue, 25 Oct 2022 07:56:38 -0700, John Larkin>>> The guy running this for me has a pretty good, fairly new PC. But the >>> sim takes hours to simulate 10s of milliseconds, so we don't want a >>> modest speedup. >> >>> TI software, TI models, runs for hours. That's silly. > >> My recollection from my power-system colleagues is that this can be >> caused by often parasitic sub circuits with very short time constants >> (compared to the core circuit), so the approach was to model only the >> core circuit at first, then start to decorate it. > >> Joe Gwinn > >The Gear integrator is specifically designed for problems like that. >("stiff systems").>Cheers>Phil HobbsWell... :-) ...... I have truly run millions of sims.......so..... (Cadence, ASIC design, not SuperSpice) Gear v Trap in practice really don't have much between them in terms of speed or convergence statistically. I might use trap when there is the characteristic triangular numerical error. Generally always use trap for oscillators to avoid correct damping. The main determiner of simulation time for a given circuit is reltol, tratol max, step time. If these are not set "appropriately", results can be totally garbled. The Spice3 default for tratol is 7. To ensure that off the bat, SMPS sim correctly trtol needs to be set to 1, however trtol=7 can run 3 times as fast as trtol=1. Tests need to be done on the circuit to see if intermediate values can be reliably used. The fundamental problem in these models is generally complete le lack of model design knowledge. A model needs to be designed just as a real circuit. On has to know what causes slowdowns due to convergence problems. Kevin Aylward https://www.kevinaylward.co.uk/gr/index.html http://www.anasoft.co.uk/ SuperSpice Freeware http://www.kevinaylward.co.uk/ee/index.html
Reply by ●October 27, 20222022-10-27
On Thu, 27 Oct 2022 08:09:08 -0700, John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote:>On Thu, 27 Oct 2022 13:46:47 +0100, Martin Brown ><'''newspam'''@nonad.co.uk> wrote: > >>On 26/10/2022 20:29, Phil Hobbs wrote: >>> John Larkin wrote: >>>> On Wed, 26 Oct 2022 08:56:04 +0100, Martin Brown >>>> <'''newspam'''@nonad.co.uk> wrote: >>>> >>>>> On 26/10/2022 04:41, John Larkin wrote: >>>>>> On Tue, 25 Oct 2022 21:52:46 +0100, Martin Brown >>>>>> <'''newspam'''@nonad.co.uk> wrote: >>>>>> >>>>> >>>>>>> Chances are one or more of the equations is stiff and the time step is >>>>>>> becoming infinitessimal on one of the rapid transitions. Adding a >>>>>>> bit of >>>>>>> spurious dissipation 1M to ground here and there might take the >>>>>>> edge off >>>>>>> whatever is making it so stiff. >>>>>> >>>>>> We are running TI's Cadence sim and TI's encrypted switcher chip >>>>>> models. >>>>>> >>>>>> We just discovered that their TPS562208 model runs about 50x faster >>>>>> than the TPS54302 model. They are very similar chips. We have both in >>>>>> our power supply design, one 54302 pre-regulating for three of the >>>>>> 562208's. >>>>> >>>>> It might be worth building one to see if it really is inclined to squeg >>>>> in real circuits. The sim could be telling you something important. >>>> >>>> If we can't reasonably sim it, we'll build it. I can imagine the three >>>> secondary switchers, negative impedance loads, making the first one >>>> oscillate. That's not a risk we want on our new delay generator. >>>> >>>> Each reg needs output caps of unknown value, and feedforward caps in >>>> its fedback divider. May as well get all that right. >> >>What might break the deadlock for computation is add a small series >>resistance to the capacitance after the first regulator to low pass >>filter it. The trick will be to find something modest enough to not >>affect the predictions much but sufficient to compute it more easily. >> >>You might have to do something like that IRL too. >> >>I expect they never expected you to daisy chain them back to back like >>that and the output of the first one really doesn't like facing the >>negative impedance dynamic load. Classic way to make an oscillator. >> >>>>> Parasitic things with ridiculously high Q can ring if provoked which is >>>>> why adding the odd spurious dissipative resistor to certain key nodes >>>>> sometimes helps instil good behaviour. Jeroen has made the same point. >>>> >>>> I wish TI knew all that. >>>> >>> >>> I suspect they give their modelers tight budgets for how many hours they >>> can spend per model. >> >>And that they never considered the output of one driving the inputs of >>several others. > >Here's what I have so far. The first switcher sees a complex load, >some of which is negative impedance. We'll split the board into noisy >and quiet "Q" halves with ferrite beads bridging the power pours. That >makes the impedances yet more interesting. > >https://www.dropbox.com/s/pklyt3mrlvf3ewm/28S662A7_power.pdf?dl=0 > >I hope it's not that ratty in real life, and is somewhat more >efficient.The power-supply guys I knew had no problem getting cascaded switchers (and linear final regulators) to work. They simulated the whole combined assembly of voltage regulators, and paid a lot of attention to loop filters and cascade filters. Not to mention cables between, if present. The whole circuit. All this being simulated using LTspice, but with proprietary locally developed models.>This being 2022, the design is dominated by what we can get. > >We have managed to get the TI switchers to work in LT Spice at >bearable speeds, tens of minutes per run instead of overnight. I don't >trust the results, especially for the first switcher, the TPS54302. > >https://www.dropbox.com/s/sd5640vv8et0jb1/TPS54302_Sim_JL1.jpg?raw=1 > >The sim log file is full of complaints about bad curly brackets and >such. I didn't draw any curly brackets. > >Switchers driving switchers is common. Nobody likes going from 24 >volts to 1 volt, even if you can get the parts. > >Go to the TI or ADI web site and look at New Products. It's mind >boggling. How can they support, and make good models for, all those >parts? How long will they keep making them?By extensive cut-and-paste reuse? Joe Gwinn
Reply by ●October 27, 20222022-10-27
Kevin Aylward wrote:> > > > > "Phil Hobbs" wrote in message > news:4f5595a2-ab59-a6d9-ee64-a4eb8835295a@electrooptical.net... > > Joe Gwinn wrote: >> On Tue, 25 Oct 2022 07:56:38 -0700, John Larkin > >>>> The guy running this for me has a pretty good, fairly new PC. But the >>>> sim takes hours to simulate 10s of milliseconds, so we don't want a >>>> modest speedup. >>> >>>> TI software, TI models, runs for hours. That's silly. >> >>> My recollection from my power-system colleagues is that this can be >>> caused by often parasitic sub circuits with very short time constants >>> (compared to the core circuit), so the approach was to model only the >>> core circuit at first, then start to decorate it. >> >>> Joe Gwinn >> >> The Gear integrator is specifically designed for problems like that. >> ("stiff systems"). > >> Cheers > >> Phil Hobbs > > Well... :-) ...... > > I have truly run millions of sims.......so..... (Cadence, ASIC design, > not SuperSpice) > > Gear v Trap in practice really don't have much between them in terms of > speed or convergence statistically.Stiff systems are only one variety of ill-conditioned simulations, sure. As the wise man said, "you can't make anything foolproof because fools are so ingenious." ;) Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Reply by ●October 27, 20222022-10-27
On Thu, 27 Oct 2022 20:02:03 +0100, "Kevin Aylward" <kevinRemoveandReplaceATkevinaylward.co.uk> wrote:> >"Martin Brown" wrote in message news:tjap6l$4vg$1@gioia.aioe.org... > >On 26/10/2022 04:41, John Larkin wrote: >> On Tue, 25 Oct 2022 21:52:46 +0100, Martin Brown >> <'''newspam'''@nonad.co.uk> wrote: >> > >>> Chances are one or more of the equations is stiff and the time step is >>> becoming infinitessimal on one of the rapid transitions. Adding a bit of >>> spurious dissipation 1M to ground here and there might take the edge off >>> whatever is making it so stiff. >> >> We are running TI's Cadence sim and TI's encrypted switcher chip >> models. >> >> We just discovered that their TPS562208 model runs about 50x faster >> than the TPS54302 model. They are very similar chips. We have both in >> our power supply design, one 54302 pre-regulating for three of the >> 562208's. > >It might be worth building one to see if it really is inclined to squeg >in real circuits. The sim could be telling you something important.Yes. We just decided to do a quick-turn 4-layer board with all four switchers. TI's spice models are insane and admittedly don't fully model the devices. I just need a 12 volt 9 tera-amp bench supply to run it. https://www.dropbox.com/s/sd5640vv8et0jb1/TPS54302_Sim_JL1.jpg?raw=1
Reply by ●October 28, 20222022-10-28
On a sunny day (Thu, 27 Oct 2022 09:33:12 -0700) it happened John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote in <99cllhp645bv1v1qsk0hb2nccifpbecgvn@4ax.com>:>I am struggling for every square mm of board area. > >>OTOH BAD things must happen for those parts not to be around anymore > >Parts seem to be available more lately, especially direct from TI. > > > >> >>That was to keep my drone airborne via a thin coax carrying high voltage >>transformed back to 7V at the drone: >> http://panteltje.com/pub/drone_power_small_core_test_IMG_6114.JPG >> http://panteltje.com/pub/h501s_drone_remote_power_flight_test_1_IMG_6274.JPG >> http://panteltje.com/pub/h501s_drone_remote_power_test_ground_control_1_IMG_6276.JPG > >I tested a 30 ga twisted pair of magnet wire. It failed at 1400 volts. >Wire-wrap wire must be 10 KV-ish. You'd need a clever HV downconverter >to use really light wire. > >How about a single wire to power a drone?How would that work? Maybe optical fiber for a high power laser? Never tried that, do have some optical fiber here.. These coaxes I use have teflon based insulation and are really good for some hundreds of volts, very light at that, The idea was to make as little RF as possible (am using about 100 kHz), put a small antenna on the drone and send the amplified received RF down the same coax (many MHz), or even use the whole coax as shortwave antenna, balloon would work too of course, just playing and investigating :-)
