On Fri, 16 Feb 2018 06:51:47 -0500, bitrex <bitrex@de.lete.earthlink.net> wrote:>On 02/15/2018 11:54 PM, John Larkin wrote: >> On Thu, 15 Feb 2018 23:11:05 -0500, bitrex >> <bitrex@de.lete.earthlink.net> wrote: >> >>> On 02/15/2018 10:39 PM, Lasse Langwadt Christensen wrote: >>>> Den fredag den 16. februar 2018 kl. 04.31.28 UTC+1 skrev bitrex: >>>>> On 02/15/2018 08:32 PM, John Larkin wrote: >>>>>> >>>>>> >>>>>> LM317's like some ESR in their output capacitors. I don't want any >>>>>> electrolytic or tantalum caps in my new thing, just ceramics, and the >>>>>> sim sure rings: >>>>>> >>>>>> https://www.dropbox.com/s/11b3w42nsvpliki/317_nocomp.jpg?raw=1 >>>>>> >>>>>> But this fixes it: >>>>>> >>>>>> https://www.dropbox.com/s/9q80heyfbwh5frp/317_comp.jpg?raw=1 >>>>>> >>>>>> This ain't rocket science, but I haven't seen it done before. >>>>>> >>>>>> >>>>> >>>>> Straight Darlington output emitter follower voltage regulators are meh, >>>>> maybe relics of when you couldn't make a decent high beta PNP on a chip >>>> >>>> apart from the dropout, whats meh about them? >>>> >>> >>> All things being equal the quasi-complimentary/Sziklai topology is >>> better in just about all respects than the straight Darlington; the >>> complex output impedance of a straight Darlington has an inductive >>> component that increases rapidly with frequency (hence stability >>> problems), the complimentary less so >> >> The emitter follower is basically low output impedance. The PNP starts >> with high output impedance. Sziklai types, like the LM1117, have >> plenty of ESR constraints too. There are few regs around that are >> stable down to zero ESR, and most are low-voltage parts. > >A Sziklai output as I'm thinking of would have a PNP in line with the >output and an NPN driving the base; it would be the functional >equivalent of just a regular NPN emitter follower, "in disguise."The 1117 is the opposite, a small PNP driving a big NPN follower. That saves a junction drop. I call that an MDO, medium dropout regulator. Its load ESR requirements are similar to a 317 type, I usually hang a 33 uF tantalum on an 1117, after which it's happy driving a bunch of ceramic bypasses.> >Unfortunately for high power regulators that's also a non-starter >because IIRC you can't make a high power monolithic PNP to pass the >current, either. It works pretty well with discretes.There are pfet-output LDOs.> >As far as I can tell design work on high power non-LDO voltage >regulators ceased around 1980 they pretty much all seem to be variants >on the LMxxx LMxxxx topologies.Everything is going low-voltage, low loss, and switchers. And the 317 types work fine, except that there could be a version compensated for ceramic caps. -- John Larkin Highland Technology, Inc lunatic fringe electronics
LM317 compensation
Started by ●February 15, 2018
Reply by ●February 16, 20182018-02-16
Reply by ●February 16, 20182018-02-16
On Fri, 16 Feb 2018 15:05:40 -0000, "Kevin Aylward" <kevinRemovAT@kevinaylward.co.uk> wrote:>"John Larkin" wrote in message >news:vecc8dt14p89hcaeb09rd3mkplt5vergrb@4ax.com... > > > >>LM317's like some ESR in their output capacitors. I don't want any >>electrolytic or tantalum caps in my new thing, just ceramics, and the >>sim sure rings: > >>https://www.dropbox.com/s/11b3w42nsvpliki/317_nocomp.jpg?raw=1 > >Urhhhh...... the 1 uohm resister is finger nails down a blackboard for me.That's there to let me snoop the current. I find it hard to aim the LT Spice current probe at device pins, so I'll stick in a resistor here and there to make current snooping easy. I usually use 1m, but I wanted an accurate sim of regulator ringing and 1m might have affected that a little. I suppose I could use 0 ohms... tries it... no, LT Spice will not probe the current in a 0 ohm resistor. Pity.> >For a regular spice engine on regular doubles, numbers can't span more than >12 digits and have it solve correctly. So, 1 uohm means 1 Meg max.Looks like I should be OK here: https://en.wikipedia.org/wiki/Double-precision_floating-point_format In real life, PCB traces can be many milliohms. I have rarely used them as current shunts.> >Don't use smaller than 1m ohm > >er... This ain't rocket science. > >And alternatively, ... you could put a low value resister in the ic output >prior to the feedback, and use a feedback cap direct from the ic output to >its input. Effectively, if configured correctly, the output resister does >the job of an out of loop esr, but without the dc error.A current snoop resistor could be on the input side too. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●February 16, 20182018-02-16
On Thursday, February 15, 2018 at 8:32:35 PM UTC-5, John Larkin wrote:> LM317's like some ESR in their output capacitors. I don't want any > electrolytic or tantalum caps in my new thing, just ceramics, and the > sim sure rings: > > https://www.dropbox.com/s/11b3w42nsvpliki/317_nocomp.jpg?raw=1 > > But this fixes it: > > https://www.dropbox.com/s/9q80heyfbwh5frp/317_comp.jpg?raw=1 > > This ain't rocket science, but I haven't seen it done before.317 needs no such ESR compensation. The ringing looks suspiciously like excitation of the SRF of an output capacitor. Did your model give it any ESL? And your solution merely reduces the shunt resistance by a factor of 20x which probably has more to do with damping than anything else.> > > -- > > John Larkin Highland Technology, Inc > > lunatic fringe electronics
Reply by ●February 16, 20182018-02-16
<bloggs.fredbloggs.fred@gmail.com> wrote in message news:d4f2ca9f-ff53-44fd-8877-51c55e3b1f1c@googlegroups.com...>317 needs no such ESR compensation. The ringing looks suspiciously like >excitation of the SRF of an output capacitor. Did your model give it any >ESL? And your solution merely reduces the shunt resistance by a factor of >20x which probably has more to do with damping than anything else. >The output has an inductive characteristic at higher frequencies. It's stable as you can see, but that doesn't mean the impedance is great. Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: https://www.seventransistorlabs.com/
Reply by ●February 16, 20182018-02-16
On Fri, 16 Feb 2018 08:57:05 -0800 (PST), bloggs.fredbloggs.fred@gmail.com wrote:>On Thursday, February 15, 2018 at 8:32:35 PM UTC-5, John Larkin wrote: >> LM317's like some ESR in their output capacitors. I don't want any >> electrolytic or tantalum caps in my new thing, just ceramics, and the >> sim sure rings: >> >> https://www.dropbox.com/s/11b3w42nsvpliki/317_nocomp.jpg?raw=1 >> >> But this fixes it: >> >> https://www.dropbox.com/s/9q80heyfbwh5frp/317_comp.jpg?raw=1 >> >> This ain't rocket science, but I haven't seen it done before. > >317 needs no such ESR compensation.The data sheet says it does.> The ringing looks suspiciously like excitation of the SRF of an output capacitor.The frequency is low, and is different on the rising and falling edges of the load current pulse. It's the chip pseudo-inductance resonating, not the cap's ESL. If the ringing were local to the caps, my damping on ADJ wouldn't fix that. Cap series L makes a different waveform than paralleled inductance.> Did your model give it any ESL? And your solution merely reduces the shunt resistance by a factor of 20x which probably has more to do with damping than anything else.With a big cap from ADJ to ground, it rings badly, too. It has to be the *right* capacitor to damp the ringing. I tried this with two different LM317 models; the ringing is somewhat different (the LT317 is better), but the damping idea is the same. It's amazing that LT ever made a 317. I think they did that early on, when they needed some revenue. They want $4 for it! I'm paying less than a tenth of that for TI. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●February 16, 20182018-02-16
John Larkin wrote:> I suppose I could use 0 ohms... tries it... no, LT Spice will not > probe the current in a 0 ohm resistor. Pity.Don't jump so far. Try an order of magnitude at a time, and also adjust read values according, of course. Real current shunts are nice too. Especially for breadboard work.
Reply by ●February 16, 20182018-02-16
On 02/16/2018 10:24 AM, John Larkin wrote:> On Fri, 16 Feb 2018 04:09:14 -0800 (PST), pcdhobbs@gmail.com wrote: > >> >>> Are you referring to the 1u resistor? That's just there to >>> conveniently snoop the current in the sim. It's the RC from ADJ to >>> ground that kills the ringing. >> >> Normally a 317 running at high CL gain is pretty stable IME, but I can imagine that a bit of lead/lag could help the load transient response. >> >> I use the lead network a bit like that in buck switchers with cap multipliers--take the DC feedback from the cap mult and AC from the switcher output. >> >> Cheers >> >> Phil Hobbs > > One way to get a stiff programmable power rail is to drive the ADJ pin > of a 3T regulator with an opamp, open-loop. Maybe that could benefit > from some sort of damping too.Yes, it's cheaper than an LM395 and works pretty well for that. The compensation can be a bit squirrelly if the op amp is too fast or the range of load currents is too wide.> > For a stiff bipolar rail, the TCA0372 opamp (dual, 1 amp per, 50 > cents) is great, but it sure doesn't like low-ESR caps on its output. > A tenth of an ohm in series with the output fixes that.Probably also a bit load sensitive, no? Voltage references also vary widely in stability--the ON Semi version of the TLV431 is unstable over most of the range you'd want to use it in--check out Figure 18 at <http://www.onsemi.com/pub/Collateral/TLV431A-D.PDF> compared with the graph on P7 of <https://www.diodes.com/assets/Datasheets/TLV431.pdf>. Not exactly plug-in replacements! 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
Reply by ●February 16, 20182018-02-16
On 02/16/2018 12:19 PM, Tim Williams wrote:> <bloggs.fredbloggs.fred@gmail.com> wrote in message > news:d4f2ca9f-ff53-44fd-8877-51c55e3b1f1c@googlegroups.com... >> 317 needs no such ESR compensation. The ringing looks suspiciously >> like excitation of the SRF of an output capacitor. Did your model give >> it any ESL? And your solution merely reduces the shunt resistance by a >> factor of 20x which probably has more to do with damping than anything >> else. >> > > The output has an inductive characteristic at higher frequencies.� It's > stable as you can see, but that doesn't mean the impedance is great. > > Tim >See e.g. the Errol Dietz (*) article at https://electrooptical.net/www/sed/ErrolDietzRegulatorNoisePeaks.pdf Cheers Phil Hobbs (*) Errol Dietz rose from being Bob Pease's technician to National's CTO. Not bad. -- 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
Reply by ●February 16, 20182018-02-16
On Friday, February 16, 2018 at 12:19:07 PM UTC-5, Tim Williams wrote:> <bloggs.fredbloggs.fred@gmail.com> wrote in message > news:d4f2ca9f-ff53-44fd-8877-51c55e3b1f1c@googlegroups.com... > >317 needs no such ESR compensation. The ringing looks suspiciously like > >excitation of the SRF of an output capacitor. Did your model give it any > >ESL? And your solution merely reduces the shunt resistance by a factor of > >20x which probably has more to do with damping than anything else. > > > > The output has an inductive characteristic at higher frequencies. It's > stable as you can see, but that doesn't mean the impedance is great.Right, any single pole rolloff has an effective inductance as output impedance, but the Q is generally too low (<1) to support ringing. His waveform shows a Q of approximately 10.> > Tim > > -- > Seven Transistor Labs, LLC > Electrical Engineering Consultation and Contract Design > Website: https://www.seventransistorlabs.com/
Reply by ●February 16, 20182018-02-16
On Friday, February 16, 2018 at 12:51:24 PM UTC-5, John Larkin wrote:> On Fri, 16 Feb 2018 08:57:05 -0800 (PST), > bloggs.fredbloggs.fred@gmail.com wrote: > > >On Thursday, February 15, 2018 at 8:32:35 PM UTC-5, John Larkin wrote: > >> LM317's like some ESR in their output capacitors. I don't want any > >> electrolytic or tantalum caps in my new thing, just ceramics, and the > >> sim sure rings: > >> > >> https://www.dropbox.com/s/11b3w42nsvpliki/317_nocomp.jpg?raw=1 > >> > >> But this fixes it: > >> > >> https://www.dropbox.com/s/9q80heyfbwh5frp/317_comp.jpg?raw=1 > >> > >> This ain't rocket science, but I haven't seen it done before. > > > >317 needs no such ESR compensation. > > The data sheet says it does. > > > The ringing looks suspiciously like excitation of the SRF of an output capacitor. > > The frequency is low, and is different on the rising and falling edges > of the load current pulse. It's the chip pseudo-inductance resonating, > not the cap's ESL. If the ringing were local to the caps, my damping > on ADJ wouldn't fix that. > > Cap series L makes a different waveform than paralleled inductance. > > > > Did your model give it any ESL? And your solution merely reduces the shunt resistance by a factor of 20x which probably has more to do with damping than anything else. > > With a big cap from ADJ to ground, it rings badly, too. It has to be > the *right* capacitor to damp the ringing. > > I tried this with two different LM317 models; the ringing is somewhat > different (the LT317 is better), but the damping idea is the same. > > It's amazing that LT ever made a 317. I think they did that early on, > when they needed some revenue. They want $4 for it! I'm paying less > than a tenth of that for TI.I doubt you're going to see this energetic resonance on anything other than the LT part.> > > > > -- > > John Larkin Highland Technology, Inc > > lunatic fringe electronics
