I'm expecting serious fast pulse loads, so I need to parallel a bunch of ceramic caps at the regulator output. The load pulse rate could be anything, so any ringing could explode. Adding the compensation parts is good insurance.
Apparently physically bigger (like 1210) caps hold up better with voltage, and may have lower ESR, so I'll use the biggest caps I can reasonably fit.
-- John Larkin Highland Technology, Inc trk jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Yes, I'm simply saying, when doing SPICE modeling with MLCC caps, add some reasonable esr. Your LM317 exercises wold be more compelling if we knew the esr. Maybe with an array of big ceramic caps, it wouldn't matter much, but it's nice to see that parameter.
--
Thanks,
- Win
That good little article just got squirreled away by me on my private website. Dietz more-or-less suggests that designers use a tantalum capacitor. And that's exactly what the wise guys did to the 7805 and
7812 on the PIC thermostat board that recently came under my purview. Another article that got squirreled away (IIRC) argues against the old "10uF and 0.1uF decoupling" rule-of-thumb. But, its exact location eludes me at present. :( You can blame it on my private website's perpetual reorg. But the darned thing won't elude me for long.Thank you,
-- Don Kuenz, KB7RPU "To invent, you need a good imagination and a pile of junk" - Thomas Alva Edison
What, me break rules? I would never do a thing like that.
I have a load that goes from 0 to 12 amps in under 100 ns. I need a bunch of ceramic caps to furnish that current without drooping much.
The 317 rings like a bell in that circuit. The added damping parts clean that up beautifully. Actually, one 22nF cap from ADJ to ground works great.
I don't want to use electrolytics for several reasons.
-- John Larkin Highland Technology, Inc trk jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Oh yeah, it's all starting to come back to me now. :) There's an ancient _Radio Shack, Voltage Regulator Handbook, National Semiconductor_ (1977) on my shelf. Dietz is not mentioned as a contributor, so maybe he hadn't become prominent yet. The book's full of very old school big iron power supplies. Although it neglects to specifically mention tantalum capacitors in its "Capacitor Selection" sub-chapter it does note them on its typical applications circuit diagrams. This is a very interesting thread. The more my mind cogitates on it, the more interesting it becomes. ;)
Thank you,
-- Don Kuenz, KB7RPU "To invent, you need a good imagination and a pile of junk" - Thomas Alva Edison
Is that 12 Amps, as in 12,000 mA?
If so, the LM317 is only a 1.5 Amp device and won't handle 12 A. You can parallel a number of LM317's, such as in Fig. 22 of the datasheet at
An alternative is to use an external pass transistor with a PNP driver, such as in Fig. 23 of the same datasheet.
In either case, the circuit is different from the single LM317 you are using to show ringing. You need to add the external circuits to show the response.
In addition, the circuit you show at
switches from 30mA to 200mA using a pulse current source. The current source may be giving misleading results because it is a high impedance.
You may need to use a switched resistive load to provide some damping. This may have a significant effect on the ringing.
Note the load transient response in Figs 3 and 4 of the TI datasheet show a considerably different response than your model.
Other manufacturer's versions of the LM317 may have a considerably different response.
As in all modeling exercises, the results should be verified in actual hardware.
Yes.
The load is pulsed at a fairly low duty cycle, so the average current is a few hundred mA. The ceramic caps provide the peak current and the
317 recharges them.
Amps are amps. The load step demonstrates the ringing, and the fix for the ringing.
They probably use caps with a lot of ESR. And the models differ too.
Probably.
Sure, we intend to actually build it.
-- John Larkin Highland Technology, Inc trk jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Your simulation shows considerably different ringing between current rise and fall. So the currents matter.
You need to model the actual currents you are using. I suspect the idle current may be much lower, and the actual charge current may be higher.
I'd be happier to see the results with a pulsed 100ns 12 Amp load. But I'd get a more recent model. See the TI PSice model below.
You need to model the ESR.
Say the caps need to hold 0.1V drop for 100ns. The required capacitance is
i = c dv/dt i dt = c dv c = i * dt / dv = 12 * 1e-7 / 0.1 = 12 uf
You can now calculate the max ESR for 0.1V drop:
R = E / I = 0.1 / 12 = 8.33 milliohms
A bunch of parallel caps should do. But you need to specify the ESR in your model.
You can get the TI Unencrypted PSpice LM317 Model at
Other TI Spice models are at
Let us know the results.
Sure, the output transistor emitter has a very different impedance from 30 to 200 mA. So the pole from that impedance into the ceramic caps is different for the two currents.
-- John Larkin Highland Technology, Inc trk jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
I did it for you. The compensation cap is critical and very different from your result. See below.
Version 4 SHEET 1 1184 680 WIRE 64 128 -96 128 WIRE 368 128 320 128 WIRE 400 128 368 128 WIRE 512 128 480 128 WIRE 624 128 512 128 WIRE 688 128 624 128 WIRE 752 128 688 128 WIRE 368 144 368 128 WIRE -96 160 -96 128 WIRE 512 160 512 128 WIRE 624 160 624 128 WIRE 192 256 192 224 WIRE 368 256 368 224 WIRE 368 256 192 256 WIRE -96 272 -96 240 WIRE 368 272 368 256 WIRE 512 272 512 224 WIRE 624 272 624 240 WIRE 192 288 192 256 WIRE 192 368 192 352 WIRE 368 384 368 352 WIRE 192 464 192 448 FLAG -96 272 0 FLAG 368 384 0 FLAG -96 128 IN FLAG 512 272 0 FLAG 688 128 OUT FLAG 624 272 0 FLAG 192 464 0 SYMBOL voltage -96 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 48 SYMBOL res 352 128 R0 SYMATTR InstName R1 SYMATTR Value 200 SYMBOL res 352 256 R0 SYMATTR InstName R2 SYMATTR Value 4.53k SYMBOL LT317A 192 128 R0 SYMATTR InstName U1 SYMBOL cap 496 160 R0 SYMATTR InstName C1 SYMATTR Value 20uf SYMBOL current 624 160 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName I1 SYMATTR Value PULSE(30m 200m 1m 1u 1u 4m) SYMBOL cap 176 288 R0 SYMATTR InstName C2 SYMATTR Value 20n SYMBOL res 176 352 R0 SYMATTR InstName R3 SYMATTR Value 20m SYMBOL res 496 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1u TEXT 72 -16 Left 2 !.tran 10m TEXT 80 -40 Left 2 ;'LT317A Demo With Capacitive Load
That seems to be my sim; same currents, no ESR in the output cap, just not as pretty. Any engineering doc should have a title, author, and date.
I tried adding 20 mohms ESR to the output caps in my sim. Nothing changed; it rings badly without the added compensation, and doesn't ring with the RC, or just the C, from ADJ to ground.
Pulsing at 10 amps for 100 ns, the results are about the same: lots of ringing, fixed by adding the same comps.
The compensation parts help.
You sure like to tell me what I need to do. But I don't report to you.
-- John Larkin Highland Technology, Inc trk jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
John Larkin wrote:
Sorry, I picked the wrong file. You can see the original by changing the current, pulse width and cap values as listed below.
This is a newsgroup discussion. The title is shown at the top of the schematic. The date is the date of the post and is shown in the header. If I put my name on the document, you will get pissed.
Sensitive? Not at all. Try the generic "you".
I increased the output cap from 12 uF to 20uF to match your value, amd changed the pulse width from 100ns to 160ns to maintain the same dv.
The compensation cap was very difficult to optimize. You either get underdamped with overshoot or overdamped with overshoot. But I ended up with the same value as you - 20nF. So the response is sensitive to pulse width, compensation cap and output cap values. Relatively small changes have a big effect on the response.
The compensation cap ESR seems to have no effect. The output cap ESR has a very significant effect on the shape of the response.
(I checked the title - this is the correct file. I need to always do that in the future.)
Version 4 SHEET 1 1184 680 WIRE 64 128 -96 128 WIRE 368 128 320 128 WIRE 400 128 368 128 WIRE 512 128 480 128 WIRE 624 128 512 128 WIRE 688 128 624 128 WIRE 752 128 688 128 WIRE 368 144 368 128 WIRE -96 160 -96 128 WIRE 512 160 512 128 WIRE 624 160 624 128 WIRE 192 256 192 224 WIRE 368 256 368 224 WIRE 368 256 192 256 WIRE -96 272 -96 240 WIRE 368 272 368 256 WIRE 512 272 512 224 WIRE 624 272 624 240 WIRE 192 288 192 256 WIRE 192 368 192 352 WIRE 368 384 368 352 WIRE 192 464 192 448 FLAG -96 272 0 FLAG 368 384 0 FLAG -96 128 IN FLAG 512 272 0 FLAG 688 128 OUT FLAG 624 272 0 FLAG 192 464 0 SYMBOL voltage -96 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 48 SYMBOL res 352 128 R0 SYMATTR InstName R1 SYMATTR Value 200 SYMBOL res 352 256 R0 SYMATTR InstName R2 SYMATTR Value 4.53k SYMBOL LT317A 192 128 R0 SYMATTR InstName U1 SYMBOL cap 496 160 R0 SYMATTR InstName C1 SYMATTR Value 20u SYMATTR SpiceLine Rser=8.33m SYMBOL current 624 160 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName I1 SYMATTR Value PULSE(30m 12 5u 10n 10n 160n) SYMBOL cap 176 288 R0 SYMATTR InstName C2 SYMATTR Value 20n SYMBOL res 176 352 R0 SYMATTR InstName R3 SYMATTR Value 0 SYMBOL res 496 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1u TEXT 72 -16 Left 2 !.tran 0 200u 0 50n TEXT 80 -40 Left 2 ;'LT317A 12A Load Transient Response
At the optimal compensation cap value, the results hardly depend on pulse width at all.
I ran your sim with pulse widths from 50 ns to 1 us, and with a compensation cap of 15 nF the droop is at most 70 mV over the whole range, i.e. 0.23%, which isn't bad at all.
With no compensation cap, it's a mess for sure.
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 https://hobbs-eo.com
I tried to keep the drop under 0.1V during the high current pulse. With 1us pulse width, the drop calculates to
i = c dv/dt i dt = c dv dv = i * dt / c = 12 * 1e-6 / 20e-6 = 0.6V 0.6 / 28 = 2.142e-2 = 2.142%
I measure 0.637V. This is
0.637 / 28 = 0.02275 = 2.275% I don't know where you are measuring the drop. With 15nF compensation, the peak current from the LT317A is 1.7A, a bit over the spec of 1.5A. The recovery from the pulse is slow and draggy but probably acceptable at low duty cycle.Here's the file:
Version 4 SHEET 1 1184 680 WIRE 64 128 -96 128 WIRE 368 128 320 128 WIRE 400 128 368 128 WIRE 512 128 480 128 WIRE 624 128 512 128 WIRE 688 128 624 128 WIRE 752 128 688 128 WIRE 368 144 368 128 WIRE -96 160 -96 128 WIRE 512 160 512 128 WIRE 624 160 624 128 WIRE 192 256 192 224 WIRE 368 256 368 224 WIRE 368 256 192 256 WIRE -96 272 -96 240 WIRE 368 272 368 256 WIRE 512 272 512 224 WIRE 624 272 624 240 WIRE 192 288 192 256 WIRE 192 368 192 352 WIRE 368 384 368 352 WIRE 192 464 192 448 FLAG -96 272 0 FLAG 368 384 0 FLAG -96 128 IN FLAG 512 272 0 FLAG 688 128 OUT FLAG 624 272 0 FLAG 192 464 0 SYMBOL voltage -96 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 48 SYMBOL res 352 128 R0 SYMATTR InstName R1 SYMATTR Value 200 SYMBOL res 352 256 R0 SYMATTR InstName R2 SYMATTR Value 4.53k SYMBOL LT317A 192 128 R0 SYMATTR InstName U1 SYMBOL cap 496 160 R0 SYMATTR InstName C1 SYMATTR Value 20u SYMATTR SpiceLine Rser=8.33m SYMBOL current 624 160 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName I1 SYMATTR Value PULSE(30m 12 5u 10n 10n 1u) SYMBOL cap 176 288 R0 SYMATTR InstName C2 SYMATTR Value 15n SYMBOL res 176 352 R0 SYMATTR InstName R3 SYMATTR Value 0 SYMBOL res 496 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1u TEXT 72 -16 Left 2 !.tran 0 200u 0 50n TEXT 80 -40 Left 2 ;'LT317A 12A Load Transient Response
If I modify someone else's sim or schematic, I show both names.
If I put that extra 0805 cap on my PC board, I can elect to leave it out, or to change its value if that improves transient response. But just changing the ring Q from 20 to 2 is a big improvement. What I don't want is the trigger rate (controlled by my cutomer) to push the ringing into bad places.
This is, to me, a new way of using the LM317. I thought I'd share it.
-- John Larkin Highland Technology, Inc trk jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
John Larkin wrote:
To me, it is amazing that a device can take such a horrible whacking and recover so gracefully. This is in a device that has zero phase shift at DC and a constant voltage difference between the output and the feedback pin.
Before we terrorize users with the response to transient pulses, the behaviour is much more docile with softer loads. Here's the result with a sine wave load. The behavior is very much the same with huge variations in load and adj cap values and load current. It is very well behaved.
Version 4 SHEET 1 1184 680 WIRE 64 128 -96 128 WIRE 368 128 320 128 WIRE 400 128 368 128 WIRE 512 128 480 128 WIRE 624 128 512 128 WIRE 688 128 624 128 WIRE 752 128 688 128 WIRE 368 144 368 128 WIRE -96 160 -96 128 WIRE 512 160 512 128 WIRE 624 160 624 128 WIRE -96 272 -96 240 WIRE 512 272 512 224 WIRE 624 304 624 240 WIRE 192 336 192 224 WIRE 336 336 192 336 WIRE 368 336 368 224 WIRE 368 336 336 336 WIRE 368 352 368 336 WIRE 192 368 192 336 WIRE 624 400 624 384 WIRE 192 448 192 432 WIRE 368 464 368 432 WIRE 192 544 192 528 FLAG -96 272 0 FLAG 368 464 0 FLAG -96 128 IN FLAG 512 272 0 FLAG 688 128 OUT FLAG 192 544 0 FLAG 336 336 R1R2 FLAG 368 128 R1R4 FLAG 624 400 0 SYMBOL voltage -96 144 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 15 SYMBOL res 352 128 R0 SYMATTR InstName R1 SYMATTR Value 470 SYMBOL res 352 336 R0 SYMATTR InstName R2 SYMATTR Value 3242 SYMBOL LT317A 192 128 R0 SYMATTR InstName U1 SYMBOL cap 496 160 R0 SYMATTR InstName C1 SYMATTR Value 1u SYMBOL cap 176 368 R0 SYMATTR InstName C2 SYMATTR Value 10n SYMBOL res 176 432 R0 SYMATTR InstName R3 SYMATTR Value 20m SYMBOL res 496 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1u SYMBOL res 608 144 R0 SYMATTR InstName R5 SYMATTR Value 100 SYMBOL voltage 624 288 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value SINE(0 1 1e3) TEXT 72 -16 Left 2 !.tran 4m TEXT 80 -40 Left 2 ;'LT317A Sine Wave Load
Another aspect of op amps and linear regulators is the ouput impedance. It often rises with frequency, giving an inductive characteristic. There is a paper by Erroil H. Dietz, Senior Technician, National Semiconductor titled "Understanding and Reducing Noise Voltage on 3-Terminal Voltage Regulators" that describes this.
The article is very hard to find. Some pdf links have it inverted which is a pain to have to rotate twice to read it.
The article is also on p204 in Appendix C of Bob Pease's book titled "Troubleshooting Analog Circuits". There is a a copy online. Scroll down to page 204 to read it:
There is a series of Bob's articles at
There is a possibility that the Dietz paper may be included in them, but it will take a while to look for it.
Ha! There is a better source at
It is on page 201. You need WinDjView to read it, which you can get at
I posted it upthread. It's on my website.
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 https://hobbs-eo.com
He used three values of Cadj, 0, 10u, and 1000u. He didn't try anything like 22nF. I'm sort of surprised that nobody seems to have tried that, or at least publicized it.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.