LM317 Votage Regulator Instability?

You may have two problems..

One, the timer is pulsing and thus causing a load variation.

two, the 317 is a monolithic device and is sensitive to RF, meaning it can actually act as a detector in a sense..

of course, oscillatiion is common with these things, hence the need for caps.

You should have a good storage cap on the output to handle riples from the load and also use some low inductive small caps across the electrolytic, input side and ref pin. ~50 uf for the load cap and maybe .1uf ceramic caps or the like for the RF snub. make sure you put a .1uf in parallel with the 50 or more uf cap.

Jamie

Average or peak currents? Better yet, put a resistor in series and look at the current waveform on a differential scope (using 2 probes). I would not be surprised to see the LM317 oscillating at some fairly high frequency. If so, you need a BFC (big fat capacitor).

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I can't argue with that, except that several people and the data sheet have mentioned that the LM317T doesn't like light loads and has a minimum load current specification. Since this is your problem, not mine, I can only offer suggestions and things to try. Since your regulator doesn't like a light load, kindly put a load resistor across the 12V line and see if the situation improves. (Hint: When troubleshooting, I like to do easy things first. A resistor is easy).

Incidentally, it is possible to damage an IC to the point where it still functions, but no in the manner commonly expected. I've done this numerous times with RF power transistors, where there are a bunch of transistor in parallel inside the packages. The LM317T probably doesn't have this exact problem, but might not appreciate having some holes blown in the IC's insulating layers. It's not easy to test for this, so just replace the LM317T with one that has a good chance of working.

I'm also wondering if there's something odd about your construction methods. I've seen 3 terminal regulators work well on a PCB breadboard with good grounding, and then go nuts when laid out on a PCB, where the ground follows a long and twisted path to the regulator. A big electrolytic on the output helps, but only if the leads and traces are fairly short. If you built your regulator on one of those solderless breadboard abominations, then I suggest you find a piece of PCB material and build it better. Then give the solderless breadboard to your worst enemy.

As for the 555, it allegedly draws very little current, but it doesn't hurt to check. Insert a 1 ohm resistor in the PS line and measure the drain with a differential input scope (i.e. using 2 scope probes). Look for giant current spikes and high freq oscillations, that don't belong.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

Here's another way to blow up an LM317 using nearly maximum voltage ratings: Just short the input or output to ground and the filter caps will discharge through the LM317.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I did try that resistor idea before, but it didn't make any difference. But it was 1500 ohms, so maybe too light to produce the desired effect.

I've now installed a 100uF electroylic in parallel with the existing

100nF ceramic on the output, but it's made no difference I'm sorry to say.

OK, your 'partially-blown theory' is noted. I'll try replacing it if all else fails as it's a bit tricky to get at.

Yeah, I'm aware of the importance of good layout so we can't attribute the issue to that in this instance.

I do actually have a current probe somewhere. Anyway, before I try any of that, I'm going to give these earlier suggestions from you and others a chance:

a) Load it down with a lower-value resistor in parallel with the electrolytic.

b)Insert bypass cap to the 'adjustment' pin

There is also the more remote poss that the regs I have a fakes. There are fake regs out there and I've had some myself off Ebay, but I'd be surprised if that were the case here since all the fakes I've heard about can barely provide a few tens of miliamps. I shall order some fresh ones from Farnell in a moment or two, though, just so I have *known* good reference regs for comparison purposes.

If the current requirements are that low, why don't you use something like the LM7812 instead of the 317?

They are far simpler to use, aren't as flakey as the 317 and about as cheap.

Unless the supply the 317 is in is needed to be voltage adjustable or high current (or higher than a LM78xx can produce), you'll save a lot of hair pulling using the LM7812.

-bruce snipped-for-privacy@ripco.com

It really doesn't need to be adjustable, but if I'm going to swap to a different device I'd really rather get something with a bit more 'headroom' than these 40V max regs, since I'm inputting ~37V. I wasn't aware that the 317 was regarded as flaky but bow to your superior knowledge of the subject!

This is now the front-running suggestion from tests I've just carried out. Confirmation one way or the other in a few hours.

Well, 'er indoors came home and set me a load of tasks around the house so I've had to put off further experimentation until tomorrow. However, the partially-blown theory looks good, because I've hooked one of the 555 boards up to a fresh 317 which is just inserted in breadboard with NO 'supporting' caps at all and it was like *totally* fine. In fact I was so sure it wouldn't work that I didn't bother to set the 12V up first, so ended up with 22V for about 10 seconds across the 555. AIUI, the max for this device is 15V, so I was convinced I'd destroyed it, but amazingly it survived! Reset the Vout to 12 and reconnected and everything was completely normal. UNLESS of course I've done some *partial* damage to the 555 in the same way as I appear to have done with the first 317, of course. Definitive results tomorrow when I'll swap out the known good 317 for the duff one.

What's flaky about the 317? You can blow them up by overvoltage, or by pushing a large current backwards into the ADJ or output terminals, but otherwise they're pretty well-behaved. Well, they will oscillate if you're running at low gain and you put 10 nF on the output. However, it's a 1-A class regulator, and just about any circuit needing anything like that much juice will have a lot more capacitance than that.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

The LM317 requires an output current of 5 mA or more, to bias it. If you really ARE only drawing 3 mA, that could be the problem.

Nope. It was still malfunctioning even when loaded down with resistors in parallel to the load to increase the current draw. I expect to have a definitive answer to what was responsible for the issue at some point today.

Well, for all the reasons you mentioned and the fact this thread is like

2 or 3 days old and the OP still hasn't made it work.

I know the 317 is probably the most common regulator used but just isn't the best choice in some cases, like this one. Cost I'd guess is the best reason to use them.

The LM78xx (and LM79xx for negative outputs) are far simplier to use at a slightly higher cost. They have excellent ripple rejection, thermal protected and short circuit protected. Usually except for a .001 cap they don't even need a heat sink (current dependent of course) for the TO-220 package.

Even if they need a heat sink, the tab is ground so you can just nut/bolt it to the chassis (provided it's at ground which is usually the case).

in-gnd-out is it, no fuss, no muss.

-bruce snipped-for-privacy@ripco.com

Yeah, I've replaced this one with the 60V variant just to be on the safe side as my first choice was too near the supply rail for comfort. I'd obviously partly damaged the old one with my prodding around under power and the cap probably partially discharged through it. So the 10 shilling postal order and box of After Eight mints goes to.... Jeff Liebermann! :-)

The logic was easy. You had already done most of the obvious troubleshooting (increase load, add capacitors, check currents). There was only one active device capable of causing a problem. All that was necessary was to contrive a reasonable explanation and failure mechanism. The near maximum voltage was an obvious problem and blowing it up by shorting the input or output to ground was quite likely. As Sherlock Holmes may have said: Once you eliminate the impossible, ridiculous, absurd, and disgusting, whatever remains, no matter how improbable, must be what's wrong.

Thanks, but no gratuities needed. Besides, the body mechanic says that I shouldn't eat chocolate. Sigh.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

False assumptions have always plagued me. In this instance I falsely assumpted that because the reg was functioning fine under a heavy passive load, there was nothing whatsoever wrong with it. I didn't admit to the possibility of partial damage which could manifest under other load conditions and that made me completely blind to fixing the problem. Lesson learned - I hope!

Assumption is the mother of all screwups. If it helps, I think everyone has the same problems with bad assumptions. It's our way of simplifying our lives. Our brains are pre-programmed to look for patterns and repetitions, so we generalize based on the available but usually insufficient evidence. Don't fight it. Assumption and generalization work well if properly used.

In my life, I've noticed that first impressions are usually correct. That applies mostly to people, but also to things. I just look at the circuit, the components, the PCB, or the schematic, and ask myself "what's the most likely culprit". Never mind logic, deduction, measurement, and flow charts. Just intuition based on experience. For example, I always look for bulging capacitors, previous repairs, crappy soldering, corrosion, bad wire connections, loose parts, signs of overheating, loose screws, and an assortment of common problems that can be found by inspection. Do it often enough, and it becomes automatic[1].

We have one big advantage with repair. We can assume that the circuit was designed correctly and that it did work at one time. All we need to do is find the one part that died. However, this was not really a repair task, but rather a construction or engineering task, which cannot make the assumption that the circuit was previously working. So, I ask myself, what is the most likely screwup someone could do and what were the likely causes of the circuit behavior? The behavior made no sense, so went through my history of all the things I've destroyed on the bench. I started with improper construction and ended with the effects of overvoltage and overload while looked for a pattern. Overvoltage and temporary wiring errors are my big problems. Often, the combination blows things up.

If intuition doesn't pay off, I make measurements. I suggested that you use a scope to look for oscillation and current spikes. It would have helped identify the exact symptoms, but would not have pointed directly to the LM317T.

I could go on forever on how I do troubleshooting. However much of it is more intuitive than logical. That's guaranteed to produce more confusion than enlightenment, so go with your first guess. It's usually right.

Good luck.

[1] I've been looking at resistor color codes for much of my life. I don't look at the colors individually and decode them into numbers and multipliers. I look at the entire resistor, automatically recognize the pattern, and (for example) 10K pops into my head. The same can be done with defective PCB's and products. Something that looks "odd" is often the culprit.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I wouldn't have minded so much, but a while ago we had a thread on the golden rules of troubleshooting and my contribution was, of all things, "assume nothing" (!)

So it's more of an art than a science? I'm not a professional technician so I don't do these sort of things on a daily basis. Consequently my reasoning skills are nowhere near as strong and well-developed as someone who's fully immersed in the business.

Nice. The problem is that the common adage "assume nothing" should really be "conclude nothing". My methods do quite a bit of assuming. For example, I assume that recurring problems will tend to reoccur. I assume that a manufacturer that makes crap, will continue to make crap. I assume that companies that make marginal products will continue to make marginal products. I assume that bad work habits will persist for a lifetime. In repair, I assume that the gizmo in question actually worked at one time (assuming it was not shipped dead on arrival). Lots of assumption, most of which are not totally true

100% of the time, but are a good start.

Where we all get into trouble is when we can't distinguish between working assumptions and a logical conclusion. Working assumption are generalizations, which may help narrow down the cause of a problem, but not necessarily point directly at the cause of a problem. Logical conclusions are when we take observations, measurements, and maybe a troubleshooting flow chart, and produce a possible culprit. The first (assumption) helps us find the problem. The 2nd (premature conclusion) usually distracts us in other directions.

So, you have my permission to make as many assumptions as you feel necessary to simplify a problem, but not to draw any conclusions from those assumptions.

It's neither an art or a science. It's a mixture. One cannot troubleshoot any problem solely on the basis of science or logic. What I think works best (for me) is an understanding of how the circuit works or should work. From that, I can guess what I should see in voltages and waveforms. Knowing what the circuit should NOT be doing is as important as what it is expect to do. That will usually reduce the problem down to the sub-section. To isolate it down to the component is where art is best. Use of a heat gun, cold spray, external signal source, self-heating, and just the look of a component, are all things that are not particularly well defined, and therefore more of an art. When both science and art fail, I try psychology. What was the designer or manufacturer trying to accomplish? I'm often amazed at how clear things become when I think about the designers motivation, instead of what is facing me on the bench.

My apologies for not being able to provide a clear and defensible picture of how troubleshooting works. It's still much of a puzzle to me. Sufficient to say that one starts with the science, which yields the fundamental functions of the circuit. One then goes to the art, where experience is more important. One then goes to psychology, where second guessing the designer provides some possible answers.

If such philosophical issues are of interest, I suggest you get a copy of Outliers by Malcolm Gladwell. Cheap enough used: "Throughout the publication, Gladwell repeatedly mentions the "10,000-Hour Rule", claiming that the key to achieving world class expertise in any skill, is, to a large extent, a matter of practicing the correct way, for a total of around 10,000 hours."

10K hrs is about 5 years of total immersion. I've found this to be true for most everything I've tried, except my crappy piano playing. Well over 10K hrs of practice does not compensate for a fundamental lack of talent.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558