Electronics-Related.com
Forums

little switching supply

Started by John Larkin October 31, 2023
On Wed, 1 Nov 2023 09:21:52 -0700 (PDT), Eddy Lee
<eddy711lee@gmail.com> wrote:

>On Wednesday, November 1, 2023 at 8:57:05?AM UTC-7, John Larkin wrote: >> On Wed, 1 Nov 2023 15:15:15 +0000, piglet <erichp...@hotmail.com> >> wrote: >> >> >On 01/11/2023 14:43, John Larkin wrote: >> >> On Wed, 1 Nov 2023 11:03:34 +0000, piglet <erichp...@hotmail.com> >> >> wrote: >> >> >> >>> On 01/11/2023 12:58 am, John Larkin wrote: >> >>>> This is a DAC-programmed power supply, 48v in and 0-36 out maybe. >> >>>> >> >>>> I want to use the TI switcher, but I don't know how to wedge it into >> >>>> LT Spice. One of my guys can run the TI simulator, so I'll let him >> >>>> tune it with the TI part. >> >>>> >> >>>> We might do programmable current limiting in an FPGA, based on the >> >>>> current shunt measurement. >> >>>> >> >>>> Version 4 >> >>>> SHEET 1 1872 756 >> >>>> WIRE 96 -112 16 -112 >> >>>> WIRE 192 -112 96 -112 >> >>>> WIRE 368 -112 192 -112 >> >>>> WIRE 16 -48 16 -112 >> >>>> WIRE 368 -48 368 -112 >> >>>> WIRE 192 0 192 -112 >> >>>> WIRE 224 0 192 0 >> >>>> WIRE 16 80 16 32 >> >>>> WIRE 656 192 512 192 >> >>>> WIRE 848 192 736 192 >> >>>> WIRE 976 192 848 192 >> >>>> WIRE 1040 192 976 192 >> >>>> WIRE 1104 192 1040 192 >> >>>> WIRE 1168 192 1104 192 >> >>>> WIRE 1312 192 1248 192 >> >>>> WIRE 1376 192 1312 192 >> >>>> WIRE 1520 192 1456 192 >> >>>> WIRE 1648 192 1520 192 >> >>>> WIRE 1712 192 1648 192 >> >>>> WIRE 1776 192 1712 192 >> >>>> WIRE 976 208 976 192 >> >>>> WIRE 1648 256 1648 192 >> >>>> WIRE 1104 272 1104 192 >> >>>> WIRE 1312 272 1312 192 >> >>>> WIRE 1520 272 1520 192 >> >>>> WIRE 96 288 16 288 >> >>>> WIRE 224 288 160 288 >> >>>> WIRE 688 288 512 288 >> >>>> WIRE 1776 288 1776 192 >> >>>> WIRE 976 304 976 272 >> >>>> WIRE 848 320 848 192 >> >>>> WIRE 16 384 16 288 >> >>>> WIRE 96 384 16 384 >> >>>> WIRE 224 384 176 384 >> >>>> WIRE 560 384 512 384 >> >>>> WIRE 976 416 976 384 >> >>>> WIRE 1104 416 1104 336 >> >>>> WIRE 1312 416 1312 336 >> >>>> WIRE 1520 416 1520 336 >> >>>> WIRE 1648 416 1648 336 >> >>>> WIRE 1776 416 1776 352 >> >>>> WIRE 688 432 688 288 >> >>>> WIRE 736 432 688 432 >> >>>> WIRE 848 432 848 384 >> >>>> WIRE 848 432 800 432 >> >>>> WIRE 16 448 16 384 >> >>>> WIRE 560 448 560 384 >> >>>> WIRE 384 560 320 560 >> >>>> WIRE 560 560 464 560 >> >>>> WIRE 688 560 688 432 >> >>>> WIRE 688 560 640 560 >> >>>> WIRE 736 560 688 560 >> >>>> WIRE 848 560 848 432 >> >>>> WIRE 848 560 816 560 >> >>>> WIRE 944 560 848 560 >> >>>> WIRE 1072 560 1024 560 >> >>>> WIRE 1104 560 1072 560 >> >>>> WIRE 320 624 320 560 >> >>>> FLAG 16 80 0 >> >>>> FLAG 976 416 0 >> >>>> FLAG 1648 416 0 >> >>>> FLAG 1712 192 OUT >> >>>> FLAG 96 -112 IN >> >>>> FLAG 320 624 0 >> >>>> FLAG 1104 416 0 >> >>>> FLAG 16 448 0 >> >>>> FLAG 560 448 0 >> >>>> FLAG 1520 416 0 >> >>>> FLAG 1776 416 0 >> >>>> FLAG 1312 416 0 >> >>>> FLAG 1040 192 MID >> >>>> FLAG 1072 560 OUT >> >>>> SYMBOL res 192 368 R90 >> >>>> WINDOW 0 59 55 VBottom 2 >> >>>> WINDOW 3 62 57 VTop 2 >> >>>> SYMATTR InstName R1 >> >>>> SYMATTR Value 150K >> >>>> SYMBOL cap 160 272 R90 >> >>>> WINDOW 0 -37 31 VBottom 2 >> >>>> WINDOW 3 -31 28 VTop 2 >> >>>> SYMATTR InstName C1 >> >>>> SYMATTR Value 200p >> >>>> SYMBOL voltage 16 -64 R0 >> >>>> WINDOW 0 30 95 Left 2 >> >>>> WINDOW 3 29 122 Left 2 >> >>>> SYMATTR InstName V1 >> >>>> SYMATTR Value 48 >> >>>> SYMBOL res 832 544 R90 >> >>>> WINDOW 0 -40 62 VBottom 2 >> >>>> WINDOW 3 -32 57 VTop 2 >> >>>> SYMATTR InstName R2 >> >>>> SYMATTR Value 50K >> >>>> SYMBOL res 656 544 R90 >> >>>> WINDOW 0 -37 59 VBottom 2 >> >>>> WINDOW 3 -31 59 VTop 2 >> >>>> SYMATTR InstName R3 >> >>>> SYMATTR Value 1K >> >>>> SYMBOL cap 800 416 R90 >> >>>> WINDOW 0 -34 36 VBottom 2 >> >>>> WINDOW 3 -30 32 VTop 2 >> >>>> SYMATTR InstName C5 >> >>>> SYMATTR Value 4n >> >>>> SYMBOL res 1632 240 R0 >> >>>> WINDOW 0 42 52 Left 2 >> >>>> WINDOW 3 53 81 Left 2 >> >>>> SYMATTR InstName Rload >> >>>> SYMATTR Value 24 >> >>>> SYMBOL LT8609S 368 192 R0 >> >>>> SYMATTR InstName U1 >> >>>> SYMBOL cap 960 208 R0 >> >>>> WINDOW 3 49 47 Left 2 >> >>>> WINDOW 0 57 23 Left 2 >> >>>> SYMATTR Value 56&#4294967295; >> >>>> SYMATTR SpiceLine Rser=2m >> >>>> SYMATTR InstName C2 >> >>>> SYMBOL ind 640 208 R270 >> >>>> WINDOW 0 77 58 VTop 2 >> >>>> WINDOW 3 68 60 VBottom 2 >> >>>> SYMATTR InstName L1 >> >>>> SYMATTR Value 47&#4294967295; >> >>>> SYMATTR SpiceLine Rser=37m >> >>>> SYMBOL voltage 480 560 R90 >> >>>> WINDOW 0 40 -2 VBottom 2 >> >>>> WINDOW 3 45 54 VTop 2 >> >>>> SYMATTR InstName V2 >> >>>> SYMATTR Value 0.553 >> >>>> SYMBOL res 960 288 R0 >> >>>> WINDOW 0 48 42 Left 2 >> >>>> WINDOW 3 56 65 Left 2 >> >>>> SYMATTR InstName R4 >> >>>> SYMATTR Value 1 >> >>>> SYMBOL cap 1088 272 R0 >> >>>> WINDOW 0 47 22 Left 2 >> >>>> WINDOW 3 44 48 Left 2 >> >>>> SYMATTR InstName C3 >> >>>> SYMATTR Value 10&#4294967295; >> >>>> SYMBOL res 1360 208 R270 >> >>>> WINDOW 0 70 54 VTop 2 >> >>>> WINDOW 3 62 56 VBottom 2 >> >>>> SYMATTR InstName R5 >> >>>> SYMATTR Value 0.1 >> >>>> SYMBOL cap 1504 272 R0 >> >>>> WINDOW 0 56 19 Left 2 >> >>>> WINDOW 3 53 47 Left 2 >> >>>> SYMATTR InstName C4 >> >>>> SYMATTR Value 56&#4294967295; >> >>>> SYMBOL cap 1760 288 R0 >> >>>> WINDOW 0 55 4 Left 2 >> >>>> WINDOW 3 58 34 Left 2 >> >>>> SYMATTR InstName Cload >> >>>> SYMATTR Value 100&#4294967295; >> >>>> SYMBOL res 1152 208 R270 >> >>>> WINDOW 0 72 60 VTop 2 >> >>>> WINDOW 3 66 63 VBottom 2 >> >>>> SYMATTR InstName R6 >> >>>> SYMATTR Value 0.5 >> >>>> SYMBOL schottky 1328 336 R180 >> >>>> WINDOW 0 -47 -3 Left 2 >> >>>> WINDOW 3 -123 -33 Left 2 >> >>>> SYMATTR InstName D1 >> >>>> SYMATTR Value RB095T-90 >> >>>> SYMATTR Description Diode >> >>>> SYMATTR Type diode >> >>>> SYMBOL res 1040 544 R90 >> >>>> WINDOW 0 -37 58 VBottom 2 >> >>>> WINDOW 3 -30 59 VTop 2 >> >>>> SYMATTR InstName R7 >> >>>> SYMATTR Value 1K >> >>>> SYMBOL cap 832 320 R0 >> >>>> WINDOW 0 48 27 Left 2 >> >>>> WINDOW 3 50 57 Left 2 >> >>>> SYMATTR InstName C6 >> >>>> SYMATTR Value 2&#4294967295; >> >>>> TEXT 808 88 Left 2 !.tran 0 10m 0 20n startup >> >>>> TEXT 1384 224 Left 2 ;polyfuse >> >>>> TEXT 840 16 Left 2 ;P943 Power Supply >> >>>> TEXT 856 48 Left 2 ;JL Oct 31 2023 >> >>>> TEXT 528 56 Left 2 ;<< TI LMR38010 >> >>>> TEXT 536 272 Left 2 ;0.774 >> >>>> TEXT 1184 224 Left 2 ;shunt >> >>>> TEXT 408 504 Left 2 >> >>>> TEXT 96 472 Left 2 ;300 KHz >> >>>> >> >>> >> >>> It uses a rather compressed range of the DAC output but looks alright. >> >>> >> >>> piglet >> >> >> >> The TI chip has a 1 volt feedback setpoint, so I could use a 1.2v >> >> bandgap for the DAC reference. The resulting DAC range might make the >> >> output go 0 to 48 volts or some such, and we'd calibrate the exact >> >> limits. >> >> >> >> The TI has an enable pin too. We could set up the DACs for zero out >> >> and then enable. >> >> >> >> One issue with a buck switcher is that it works in both directions, so >> >> we can possibly pump load power uphill into our 48 volt supply. That >> >> leads to an ultraviolet catastrophe doom loop, so we would disable the >> >> switcher of we sense it doing that. That logic will be tricky. >> >> >> > >> >Current sense metrology might be one way? >> > >> >piglet >> > >> The current sense, using the shunt, can be used to detect reverse >> power and do something. >> >> One complication is startup. Suppose we want to use this supply to >> charge a battery. With our supply disabled or set low, there's a stiff >> 24 volts at our output. How can we start charging it? > >How about a diode to the battery? It's good to protect reverse connection anyway.
Yeah, a diode in the output path makes sense. The supply would be 1 amp max, so the diode drop and dissipation wouldn't be a big deal. The diode affects loop dynamics, but I already split the feedback into AC and DC paths, so that should work out. One of my guys says he can import the TI switcher model into LT Spice, so I'll cut over to that if it works.
On Thursday, November 2, 2023 at 3:01:55&#8239;AM UTC+11, John Larkin wrote:
> On Wed, 1 Nov 2023 08:19:32 -0700 (PDT), Anthony William Sloman > <bill....@ieee.org> wrote: > >On Wednesday, November 1, 2023 at 11:59:33?AM UTC+11, John Larkin wrote: > >> This is a DAC-programmed power supply, 48v in and 0-36 out maybe. > >> > >> I want to use the TI switcher, but I don't know how to wedge it into LT Spice. One of my guys can run the TI simulator, so I'll let him tune it with the TI part. > >> > >> We might do programmable current limiting in an FPGA, based on the current shunt measurement. > > > >I ran it. It looked sort of sensible, until I looked at L1 which hasn't got an parallel capacitance at all. I put in 1pF - which would be low (series resonance at 23MHz) and got big switching spikes. A more realistic 10pF (series resonant at about 7MHz) got me huge spikes. A 1uH ferrite bead in series (out of the Wurth range, good for about 3A) tamed the spikes back to merely big. > > > >Replacing L1 with four 12uH Wurth beads in series gave even smaller switching spikes > > > >It pays to use realistic models in LTSpice. If John had worked out what he was going to use for L1 the manufacturer's data sheet should have given him the self-resonant frequency and the parallel capacitance. > > You were doing great until you switched to insult mode. You can't help yourself.
i hit insult mode in the first line " until I looked at L1 which hasn't got an parallel capacitance at all". It's not actually an insulting observation - simply a statement of fact. You should know better. This isn't the first time you've made this mistake and I've pointed it out before. John Field's made exactly the same mistake in our "low-power 100kHz oscillator" tussle years ago, and didn't seem to feel insulted when I pointed it out. If you want to feel insulted less often, make fewer minor errors. I was rather hoping you'd worry about the extra dissipation in the LT8609S generated by the switching spikes. but you don't seem to have thought about that either. -- Bill Sloman, Sydney
On Wed, 1 Nov 2023 19:50:11 -0700 (PDT), Anthony William Sloman
<bill.sloman@ieee.org> wrote:

>On Thursday, November 2, 2023 at 3:01:55?AM UTC+11, John Larkin wrote: >> On Wed, 1 Nov 2023 08:19:32 -0700 (PDT), Anthony William Sloman >> <bill....@ieee.org> wrote: >> >On Wednesday, November 1, 2023 at 11:59:33?AM UTC+11, John Larkin wrote: >> >> This is a DAC-programmed power supply, 48v in and 0-36 out maybe. >> >> >> >> I want to use the TI switcher, but I don't know how to wedge it into LT Spice. One of my guys can run the TI simulator, so I'll let him tune it with the TI part. >> >> >> >> We might do programmable current limiting in an FPGA, based on the current shunt measurement. >> > >> >I ran it. It looked sort of sensible, until I looked at L1 which hasn't got an parallel capacitance at all. I put in 1pF - which would be low (series resonance at 23MHz) and got big switching spikes. A more realistic 10pF (series resonant at about 7MHz) got me huge spikes. A 1uH ferrite bead in series (out of the Wurth range, good for about 3A) tamed the spikes back to merely big. >> > >> >Replacing L1 with four 12uH Wurth beads in series gave even smaller switching spikes >> > >> >It pays to use realistic models in LTSpice. If John had worked out what he was going to use for L1 the manufacturer's data sheet should have given him the self-resonant frequency and the parallel capacitance. >> >> You were doing great until you switched to insult mode. You can't help yourself. > >i hit insult mode in the first line " until I looked at L1 which hasn't got an parallel capacitance at all". > >It's not actually an insulting observation - simply a statement of fact. You should know better. This isn't the first time you've made this mistake and I've pointed it out before. John Field's made exactly the same mistake in our "low-power 100kHz oscillator" tussle years ago, and didn't seem to feel insulted when I pointed it out. > >If you want to feel insulted less often, make fewer minor errors. > >I was rather hoping you'd worry about the extra dissipation in the LT8609S generated by the switching spikes. but you don't seem to have thought about that either.
Bizarre. LT Spice has macromodel test fixtures for their LTC and ADP switcher chips. The inductors sometimes include ESR but no shunt capacitance. Check for yourself. I've simulated and released to manufacturing maybe 30 various switchers in the last few years and never bothered with inductor parasitic capacitance. There's not enough capacitance in the power inductors to matter. Do some math. One power inductor that I use, 22 uH 2 amp Coilcraft, has 2.4 pF shunt capacitance. The PCB capacitance of a DPAK is more than that. How many switchers have you built lately?
On Thursday, November 2, 2023 at 2:36:05&#8239;PM UTC+11, John Larkin wrote:
> On Wed, 1 Nov 2023 19:50:11 -0700 (PDT), Anthony William Sloman > <bill....@ieee.org> wrote: > >On Thursday, November 2, 2023 at 3:01:55?AM UTC+11, John Larkin wrote: > >> On Wed, 1 Nov 2023 08:19:32 -0700 (PDT), Anthony William Sloman > >> <bill....@ieee.org> wrote: > >> >On Wednesday, November 1, 2023 at 11:59:33?AM UTC+11, John Larkin wrote: > >> > > >> >> This is a DAC-programmed power supply, 48v in and 0-36 out maybe. > >> >> > >> >> I want to use the TI switcher, but I don't know how to wedge it into LT Spice. One of my guys can run the TI simulator, so I'll let him tune it with the TI part. > >> >> > >> >> We might do programmable current limiting in an FPGA, based on the current shunt measurement. > >> > > >> >I ran it. It looked sort of sensible, until I looked at L1 which hasn't got an parallel capacitance at all. I put in 1pF - which would be low (series resonance at 23MHz) and got big switching spikes. A more realistic 10pF (series resonant at about 7MHz) got me huge spikes. A 1uH ferrite bead in series (out of the Wurth range, good for about 3A) tamed the spikes back to merely big. > >> > > >> >Replacing L1 with four 12uH Wurth beads in series gave even smaller switching spikes > >> > > >> >It pays to use realistic models in LTSpice. If John had worked out what he was going to use for L1 the manufacturer's data sheet should have given him the self-resonant frequency and the parallel capacitance. > >> > >> You were doing great until you switched to insult mode. You can't help yourself. > > > >i hit insult mode in the first line " until I looked at L1 which hasn't got an parallel capacitance at all". > > > >It's not actually an insulting observation - simply a statement of fact. You should know better. This isn't the first time you've made this mistake and I've pointed it out before. John Field's made exactly the same mistake in our "low-power 100kHz oscillator" tussle years ago, and didn't seem to feel insulted when I pointed it out. > > > >If you want to feel insulted less often, make fewer minor errors. > > > >I was rather hoping you'd worry about the extra dissipation in the LT8609S generated by the switching spikes. but you don't seem to have thought about that either. > > Bizarre. LT Spice has macromodel test fixtures for their LTC and ADP switcher chips. The inductors sometimes include ESR but no shunt capacitance. Check for yourself.
I'm happy to take your word for it. It just means that the people who set up the test fixtures made the same mistake that you did. Since the switching spikes that appear when you put a realistic parallel capactiance into your inductor model hit the switches inside the LT8609S just when they are turning on and off, they signicantly increase the dissipation in the switches reducing the maximum output currents that the device can deliver, you can see why marketing wouldn't be keen to see that mistake corrected,
> I've simulated and released to manufacturing maybe 30 various switchers in the last few years and never bothered with inductor parasitic capacitance. There's not enough capacitance in the power inductors to matter.
There hasn't been enough so far. Getting stuff wrong and getting away with it means that you have been over-doing the safety margins. It may be prudent design but it is also extravagant design.
> Do some math. One power inductor that I use, 22 uH 2 amp Coilcraft, has 2.4 pF shunt capacitance. The PCB capacitance of a DPAK is more than that.
So what? The parallel capacitance of the inductor has to be charged or discharged whenever you switch it's input from rail to rail, and you do that a lot.
> How many switchers have you built lately?
None in recent years. The most recent is probably the one in the 1993 milli-degree thermostat. Figure 8 from that paper shows a bunch of 1.3uH ferrite beads put in to cope with exactly that problem. Meas. Sci. Technol. 7 (1996) 1653&ndash;1664. -- Bill Sloman, Sydney
On Wed, 1 Nov 2023 21:54:25 -0700 (PDT), Anthony William Sloman
<bill.sloman@ieee.org> wrote:

>On Thursday, November 2, 2023 at 2:36:05?PM UTC+11, John Larkin wrote: >> On Wed, 1 Nov 2023 19:50:11 -0700 (PDT), Anthony William Sloman >> <bill....@ieee.org> wrote: >> >On Thursday, November 2, 2023 at 3:01:55?AM UTC+11, John Larkin wrote: >> >> On Wed, 1 Nov 2023 08:19:32 -0700 (PDT), Anthony William Sloman >> >> <bill....@ieee.org> wrote: >> >> >On Wednesday, November 1, 2023 at 11:59:33?AM UTC+11, John Larkin wrote: >> >> > >> >> >> This is a DAC-programmed power supply, 48v in and 0-36 out maybe. >> >> >> >> >> >> I want to use the TI switcher, but I don't know how to wedge it into LT Spice. One of my guys can run the TI simulator, so I'll let him tune it with the TI part. >> >> >> >> >> >> We might do programmable current limiting in an FPGA, based on the current shunt measurement. >> >> > >> >> >I ran it. It looked sort of sensible, until I looked at L1 which hasn't got an parallel capacitance at all. I put in 1pF - which would be low (series resonance at 23MHz) and got big switching spikes. A more realistic 10pF (series resonant at about 7MHz) got me huge spikes. A 1uH ferrite bead in series (out of the Wurth range, good for about 3A) tamed the spikes back to merely big. >> >> > >> >> >Replacing L1 with four 12uH Wurth beads in series gave even smaller switching spikes >> >> > >> >> >It pays to use realistic models in LTSpice. If John had worked out what he was going to use for L1 the manufacturer's data sheet should have given him the self-resonant frequency and the parallel capacitance. >> >> >> >> You were doing great until you switched to insult mode. You can't help yourself. >> > >> >i hit insult mode in the first line " until I looked at L1 which hasn't got an parallel capacitance at all". >> > >> >It's not actually an insulting observation - simply a statement of fact. You should know better. This isn't the first time you've made this mistake and I've pointed it out before. John Field's made exactly the same mistake in our "low-power 100kHz oscillator" tussle years ago, and didn't seem to feel insulted when I pointed it out. >> > >> >If you want to feel insulted less often, make fewer minor errors. >> > >> >I was rather hoping you'd worry about the extra dissipation in the LT8609S generated by the switching spikes. but you don't seem to have thought about that either. >> >> Bizarre. LT Spice has macromodel test fixtures for their LTC and ADP switcher chips. The inductors sometimes include ESR but no shunt capacitance. Check for yourself. > >I'm happy to take your word for it. It just means that the people who set up the test fixtures made the same mistake that you did. > >Since the switching spikes that appear when you put a realistic parallel capactiance into your inductor model hit the switches inside the LT8609S just when they are turning on and off, they signicantly increase the dissipation in the switches reducing the maximum output currents that the device can deliver, you can see why marketing wouldn't be keen to see that mistake corrected, > >> I've simulated and released to manufacturing maybe 30 various switchers in the last few years and never bothered with inductor parasitic capacitance. There's not enough capacitance in the power inductors to matter. > >There hasn't been enough so far. Getting stuff wrong and getting away with it means that you have been over-doing the safety margins. It may be prudent design but it is also extravagant design. > >> Do some math. One power inductor that I use, 22 uH 2 amp Coilcraft, has 2.4 pF shunt capacitance. The PCB capacitance of a DPAK is more than that. > >So what? The parallel capacitance of the inductor has to be charged or discharged whenever you switch it's input from rail to rail, and you do that a lot. > >> How many switchers have you built lately? > >None in recent years. The most recent is probably the one in the 1993 milli-degree thermostat. Figure 8 from that paper shows a bunch of 1.3uH ferrite beads put in to cope with exactly that problem. Meas. Sci. Technol. 7 (1996) 1653&#4294967295;1664.
That stirred-water bath seems to be the highlight of your life. Your main motivation is to be nasty. That overpowers your common sense and basic numeracy and ability to learn. That's a too-common human behavior, letting emotion overpower reason.
On Friday, November 3, 2023 at 1:26:47&#8239;AM UTC+11, John Larkin wrote:
> On Wed, 1 Nov 2023 21:54:25 -0700 (PDT), Anthony William Sloman > <bill....@ieee.org> wrote: > >On Thursday, November 2, 2023 at 2:36:05?PM UTC+11, John Larkin wrote: > >> On Wed, 1 Nov 2023 19:50:11 -0700 (PDT), Anthony William Sloman > >> <bill....@ieee.org> wrote: > >> >On Thursday, November 2, 2023 at 3:01:55?AM UTC+11, John Larkin wrote: > >> >> On Wed, 1 Nov 2023 08:19:32 -0700 (PDT), Anthony William Sloman > >> >> <bill....@ieee.org> wrote: > >> >> >On Wednesday, November 1, 2023 at 11:59:33?AM UTC+11, John Larkin wrote: > >> >> > > >> >> >> This is a DAC-programmed power supply, 48v in and 0-36 out maybe. > >> >> >> > >> >> >> I want to use the TI switcher, but I don't know how to wedge it into LT Spice. One of my guys can run the TI simulator, so I'll let him tune it with the TI part. > >> >> >> > >> >> >> We might do programmable current limiting in an FPGA, based on the current shunt measurement. > >> >> > > >> >> >I ran it. It looked sort of sensible, until I looked at L1 which hasn't got an parallel capacitance at all. I put in 1pF - which would be low (series resonance at 23MHz) and got big switching spikes. A more realistic 10pF (series resonant at about 7MHz) got me huge spikes. A 1uH ferrite bead in series (out of the Wurth range, good for about 3A) tamed the spikes back to merely big. > >> >> > > >> >> >Replacing L1 with four 12uH Wurth beads in series gave even smaller switching spikes > >> >> > > >> >> >It pays to use realistic models in LTSpice. If John had worked out what he was going to use for L1 the manufacturer's data sheet should have given him the self-resonant frequency and the parallel capacitance. > >> >> > >> >> You were doing great until you switched to insult mode. You can't help yourself. > >> > > >> >i hit insult mode in the first line " until I looked at L1 which hasn't got an parallel capacitance at all". > >> > > >> >It's not actually an insulting observation - simply a statement of fact. You should know better. This isn't the first time you've made this mistake and I've pointed it out before. John Field's made exactly the same mistake in our "low-power 100kHz oscillator" tussle years ago, and didn't seem to feel insulted when I pointed it out. > >> > > >> >If you want to feel insulted less often, make fewer minor errors. > >> > > >> >I was rather hoping you'd worry about the extra dissipation in the LT8609S generated by the switching spikes. but you don't seem to have thought about that either. > >> > >> Bizarre. LT Spice has macromodel test fixtures for their LTC and ADP switcher chips. The inductors sometimes include ESR but no shunt capacitance. Check for yourself. > > > >I'm happy to take your word for it. It just means that the people who set up the test fixtures made the same mistake that you did. > > > >Since the switching spikes that appear when you put a realistic parallel capactiance into your inductor model hit the switches inside the LT8609S just when they are turning on and off, they significantly increase the dissipation in the switches reducing the maximum output currents that the device can deliver, you can see why marketing wouldn't be keen to see that mistake corrected, > > > >> I've simulated and released to manufacturing maybe 30 various switchers in the last few years and never bothered with inductor parasitic capacitance. There's not enough capacitance in the power inductors to matter. > > > >There hasn't been enough capacitance so far. Getting stuff wrong and getting away with it means that you have been over-doing the safety margins. It may be prudent design but it is also extravagant design. > > > >> Do some math. One power inductor that I use, 22 uH 2 amp Coilcraft, has 2.4 pF shunt capacitance. The PCB capacitance of a DPAK is more than that. > > > >So what? The parallel capacitance of the inductor has to be charged or discharged whenever you switch it's input from rail to rail, and you do that a lot. > > > >> How many switchers have you built lately? > > > >None in recent years. The most recent is probably the one in the 1993 milli-degree thermostat. Figure 8 from that paper shows a bunch of 1.3uH ferrite beads put in to cope with exactly that problem. Meas. Sci. Technol. 7 (1996) 1653&ndash;1664. > > That stirred-water bath seems to be the highlight of your life.
That's funny. I cited some stirred water-bath work, but made the point that we couldn't use that approach (and consequently couldn't do as well as you can that way) but could get away with a much more compact and less well-insulated set up. It was fun, but scarcely a "high-light of my life".
> Your main motivation is to be nasty.
No. It's discourage people from posting nonsense, which does seem to be an enthusiasms of yours. Some of the nonsense you get from climate change denial web-sites, but some of it is less obviously derivative.
> That overpowers your common sense and basic numeracy and ability to learn.
Not areas where your own performance is famously good. You think that Donald Trump has common sense
> That's a too-common human behavior, letting emotion overpower reason.
As in your desired to be flattered, and your chagrin when your boasts fail to impress. -- Bill Sloman, Sydney