LM317 power supply diagnostics

[microbug]

Hi,
I have been working on a simple LM317 power supply with current limit for some time now (Paul Price's design). The schematic I used to make the pcb (home-etched) is attached.
I only have a crappy UT120C pocket DMM which won't do current >400ma and a Hitachi V-223 scope (picked up for £15 on eBay ) which I recently 'calibrated' with a cheap AD9850 module.
My problem at the moment is that the current limiting isn't working properly. It seems to be very intermittent and always pulls the output to .2v or so (not 0). The voltage adjust pot is a very nice Vishay spectrol wirewound, but the current one is a cheap ebay thing from china (a fake Bourns one) - I need to replace it, but I don't see why it should cause too much of a problem at the moment because there isn't much current going through it. I don't know, but I suspect it's a 10-turn trimmer in a wirewound housing.
I need to know how I can work out what the current limit is based on the non-inverting input of op-amp A - it doesn't seem to directly relate to the limit which is applied.
Is there something I'm missing here? My PCB layout was pretty sketchy, but it seems to be OK (all components tested linked to correct places). The components should be decent quality (apart from the current set pot). The trace widths on the pcb are plenty wide. Input is just over 40v pk-pk, rectified to 20v. I have noticed that if I set the voltage adj pot to its minimum, the output is -0.4v.
Would I be better off just starting again with the pcb? I can attach pics of it if needed.
Adding to the confusion is the fact that I don't really trust the readings on my multimeter. It's not shielded at all, often reads 5-50mv background noise and my oscilloscope shows a large amount of background noise as well - I am working fairly close to our switchboard. I'm getting a BM257 for Xmas.

[Simon]

what is the current limit op-amp supposed to be doing ? most limiting designs rely on a BJT and a voltage drop resistor that shuts the transitor off (or rather turns it on so that it turns something else off)

[microbug]

As far as I can tell (this isn't my design) the first opamp amplifies the drop across a 0.01ohm shunt and feeds it into the second. The second acts as a comparator, pulling the adj pin low if the set current is lower than the output. The diode between the output of the second opamp and the adj pin prevents the opamp from pulling adj high.
There is a nice dual tracking lm317 in silicon chip. I have the digital copy of that issue and the pcb image for etching - I might just start again - scrap this - and use that as it is much better documented and also dual tracking. I believe that one uses a transistor to pull the adj pin as low as it needs to before the output current reaches the set current.

[codeboy2k]

According to your schematic, the points labelled IOUT and ISET are direct reading, at 1V = 1A , so if you put a digital multimeter there you can read amps directly at either the ISET point or the IOUT point.

what is the current limit op-amp supposed to be doing ? most limiting designs rely on a BJT and a voltage drop resistor that shuts the transitor off (or rather turns it on so that it turns something else off)

It looks like the 358 current limit opamp is drawn to do that, to shunt current away from the ADJ pin of the LM317, thus lowering the voltage output until the IOUT pin matches the ISET pin.

As the ADJ pin of the 317 only requires a few micro-amps of current, the opamp directly on the ADJ pin should be able to do it.

to the OP : note that the LM358 op amp has 3mV of input offset, so this will be amplified at the current sense by 100, so, a 500mA limit will be set at 0.500V at the ISET node, but the input  offset of the current sense opamp will likely cause the current sense to trip at 0.500V - (3mVx100) = 0.200V

In other words, as soon as 200ma is drawn from the circuit through the 10mOhm sense resistor, then that opamp will multiply that by 100-> 200mV, add that to the input offset x 100 = 200mV+300mV = 500mV and that matches your setpoint.

So you will seem to set it for 500mA but it trips at way less than that because of the sense opamp's input offset.

[Mark Hennessy]

My problem at the moment is that the current limiting isn't working properly. It seems to be very intermittent and always pulls the output to .2v or so (not 0). The voltage adjust pot is a very nice Vishay spectrol wirewound, but the current one is a cheap ebay thing from china (a fake Bourns one) - I need to replace it, but I don't see why it should cause too much of a problem at the moment because there isn't much current going through it. I don't know, but I suspect it's a 10-turn trimmer in a wirewound housing.
I need to know how I can work out what the current limit is based on the non-inverting input of op-amp A - it doesn't seem to directly relate to the limit which is applied.
Is there something I'm missing here? My PCB layout was pretty sketchy, but it seems to be OK (all components tested linked to correct places). The components should be decent quality (apart from the current set pot). The trace widths on the pcb are plenty wide. Input is just over 40v pk-pk, rectified to 20v. I have noticed that if I set the voltage adj pot to its minimum, the output is -0.4v.
Would I be better off just starting again with the pcb? I can attach pics of it if needed.

First, I wouldn't worry too much about not getting precisely 0V when the voltage is at min; this depends on the forward voltage of LED2. It needs to match the 1.25V reference inside the LM317. You could replace LED2 with a VBE-multiplier (one small-signal transistor and a resistor, and a pre-set resistor) if this matters to you.

Regarding the current control circuitry, I see how it is supposed to work, but I'm slightly worried about the fact the op-amp is powered from 0V, and it expects to work with the inputs at 0V. Yes, I know the LM358 is supposed to work with the inputs at 0V, but even so, I personally would have preferred to power the op-amp from the -2V rail. Of course, the 555-generated -2V rail might not be able to supply the current; it would need testing. Indeed, when I built something like this many years back, I derived the negative rail from the AC input directly. Much simpler...

As the op-amp's output can't get to 0V, that's why the current on this PSU can't be adjusted right down to 0mA.

To debug, break it down. Does the voltage at "IOUT" seem sensible? As mentioned, it should be 1V per amp. This will be there all the time, so I'd lift one end of D3 so that the current limit can't affect the output, then try varying the load (use some chunky resistors). If you can borrow a second DVM, you can measure the actual current while monitoring the voltage at IOUT.

Once you're happy that's working, check that the voltage at "ISET" varies smoothly as you adjust the control. It should vary between 0 and ~1.6V (depends on the LED, of course). If that looks good, reinstall D3, and see what happens. Use your 'scope to see if the op-amp is oscillating when it enters current limit mode...

Oh, and the current limit circuit won't be able to pull the output right down to zero - you have whatever the op-amp can do when the output is saturated against its negative rail, plus the drop of D3, plus the 1.25V of the LM317. Don't expect lab-grade performance from such a simple supply; but that said, this ought to be pretty good when working, so it'll be worth the effort. Taking the time to figure out how it should be working will give you invaluable experience for the future

And watch the temperature of the LM317. I wouldn't expect 1.5A at low voltages from a single one of these. In the future, you might consider strapping a beefy transistor around it to share the load...

Good luck,

Mark

[Kevin.D]

Just some problems with this circuit microbug some already pointed out by other responders .
  How stable do you think that output voltage derived from the 555 is .? If it drifts then so does the output of the V reg .
On your current limiting circuitry :-
 The input offset Voltage of the 358 is (2-7 mV ?) times the gain of IC3b (you have ~100) will appear as an offset on it's output .
  In order for the output of the lm317 to be ~ 0 V  requires the adj pin being pulling down to -1.25V  ,but your opamp IC3A output can only go down to ~0 V  (and then you have a diode  drop of D3 to add to that) ,so during current limiting you would still have ~ 1.8 V at the output off the lm317 .
 Lastly  that current control loop will oscillate ,look at the enourmous gain you got around that loop there ,you would need to roll off the high frequency gain of IC3A by quite alot to make that stable .

Heres an idea why dont you drop the lm317 (which was never designed to go to 0v) and go for a nice darlington transistor based design instead ,it will cost you not much more and you will gain much more knowledge doing that and end up with a nice power supply at the end .

[microbug]

Ok, there's something weird going on. With a 50 ohm load at 10v, there was about 200ma of current (using my scope as a dvm ?). Iout read 120mv. Then the current started to drop spontaneously and Iout began to oscillate at 2.5 kHz. Is this thermal shutdown on the LM317? It should have only been dropping about a watt, it was not but not burning.
Edit:
@Kevin.D, do you have any specific designs in mind?

I'm looking at Iset. With the same 50ohm load, current limiting triggers at just below 150mv. Between 150mv and 45mv, Iout oscillates before settling a bit above 0 (I can't remember exactly, my bench is downstairs and computer is upstairs).

[Mark Hennessy]

Is the LM317 on a decent heat sink?

[microbug]

No - I can mount it to the fairly large metal back plate of the enclosure the supply will eventually be in. I do have heatsink compound.

[Mark Hennessy]

It must be on a heat sink. With 10 volts out, you've got approximately 10 volts across the LM317. At a current of 200mA, that's a power dissipation of 2 watts.

The thermal resistance of an LM317T with no heat sink is 80C/W, so the temperature rise will be 160C.

Be careful - that could hurt! In general, if a component is too hot to hold, it's more than 60C, and for a semiconductor, that's usually too hot for long-term reliability. In my experience, despite the claimed built-in thermal protection, they will fail when allowed to overheat.

When you attach it to the heat sink, remember you'll need to insulate the metal tab.

Remember, you can't debug the current controller until you're happy the thing works reliably in "constant voltage" mode. As I said above, break it down

[microbug]

OK- I have an insulating kit for to-220 packages so I'll drill a hole in the back panel and mount it. I'll post back afterwards!

[microbug]

OK, the LM317 is properly mounted with a TO-220 mounting kit and some thermal grease. The tab of the LM317 is not shorted to the back plate. The aluminium back plate seems to be working well - it is getting very slightly warm on a 50ohm load @ 10v.
Iout and Iset both seem to have a ~20mv offset. With the same load as before (2x100ohm 5w resistors in parallel @10v) Iout is 120mv, which is about right. When Iset is at or below the same value, Iout oscillates. The pk-pk voltage of Iout gets lower and lower until when Iset is near 20mv, Iout stays at 30-40mv.
I'm a little concerned about the oscillation - is there anything I can do to fix it?

[Mark Hennessy]

Sounds like progress

Try a 100pF capacitor between pins 1 and 2 of IC3...

All the best,

Mark

[microbug]

I put a 100pf ceramic cap between pins 1 and 2 of the LM358 and the oscillations reduced but didn't completely go, so I swapped it for a 1nf cap. However, that was with a 50ohm load. With a 33.3ohm (3x100ohm parallel) load, the oscillations re-appear. They are at ~8khz.
Also, with no load Iout is 10mv. With a 100ohm load, it reads 20mv. 50ohm reads 120mv and 33.3 reads 220mv. This doesn't seem right - it is both non-linear and has a -80ma offset :/

[Kevin.D]

Why are you so commited to using this design ?,when theres plenty of other less problematic  designs for you to choose from ,just do the usual search with the right key words .
ok lets see :-
 quote :-  "The pk-pk voltage of Iout gets lower and lower until when Iset is near 20mv, Iout stays at 30-40mv"  .
well that will be because of what I posted earlier about there being a minimum  of ~1.8V output that your current limit can reduce the output to ,so the minum current you can set would then depend on what the resistance of the load is,so 1.8/RLoad .( so with a direct short your  current limit would be infinite and so have no effect whatever).
 A possible fix for above would be to remove D3 and use somthing like a small N chan mosfet (2n7000) to drive the lm317 adj pin (source to -2V, ) ,so the opamp now drives gate of fet instead of adj pin directly ,this would also allow the use of a lower voltage  opamp ( it would also invert opamp output control (going higher now pulls down adj pin)so you would swap the inputs of ic3a.)  .

Quote :-"Also, with no load Iout is 10mv. With a 100ohm load, it reads 20mv. 50ohm reads 120mv and 33.3 reads 220mv. This doesn't seem right - it is both non-linear and has a -80ma offset :/"
 
Well the first three seem like what I would expect from that opamp but i'm not sure where the last (33R3) is coming from ,you have at least two other offsets  , there is the input offset votage and also the input bias current (which is typically 50-100nA for 358, so 50nA through 100k ~5mV offset).If you really dont need to have a 1V=1A output of ic3b then reducing the gain would solve your offset probs(or use a better opamp).

quote :-" I put a 100pf ceramic cap between pins 1 and 2 of the LM358  "
  Yes putting a  feedback cap to - input would help you with stabilty by lowering bandwidth of ic3a ,but  for that cap to be really effective the - input needs to be high impedance (but here it's directly connected to the output  of ic3b ,so it's low impedance),so put a resistor (say 10k) between the inv input and output ic3b. Without looking at a bode plot of that loop I could not guess at how big a cap it needs to be but just experiment till it's stable under various loads then add some   .(with a little bit of reading you could do a simple transient step load test on the real circuit and watch the response on your scope to determine the stabilty and approx optimum size,).

anyway good luck and happy expermenting

[microbug]

OK. I'm beginning to think that trying to make a current-limited power supply that works is more trouble than it's worth. Especially as this is my first power supply, I want to have something that is known to work. As I said before, there is a nice dual tracking supply in silicon chip which I could use.
In the meantime I'd like to try and carry on the design I have currently; I will create a second version of the schematic (and in turn the pcb). I'll use a 2n7000 on the adj pin of the LM317 for the current limit and a better op-amp instead of LM358 (any suggestions?). I'll also experiment with the cap size. Again, any suggestions for reading material on transient step load tests would be welcome. I found this app note by the late Jim Williams.
I am using a 2x0-15v transformer which in reality reads 40v pk-pk - I am only using one winding, so I could use the other as a negative rail. If I did this, would I need to regulate it down to 2v or whatever (using 79xx reg)? I'm guessing so. This would also have the added benefit of higher stability, as I do not completely trust the 555 inverter - it reads -1.6v not -2.

[Mark Hennessy]

OK. I'm beginning to think that trying to make a current-limited power supply that works is more trouble than it's worth. Especially as this is my first power supply, I want to have something that is known to work.

In the first post, you cite the designer. Was this offered as a complete, debugged design, or as a starting point?

In the meantime I'd like to try and carry on the design I have currently; I will create a second version of the schematic (and in turn the pcb). I'll use a 2n7000 on the adj pin of the LM317 for the current limit and a better op-amp instead of LM358 (any suggestions?).

I'm not sure that an op-amp upgrade is needed at this stage. Nor would I add anything else between the op-amp and the LM317. Instead, try and arrange a negative supply for it - even temporarily as part of the debug process. Even a PP3 might help you prove the point

I'll also experiment with the cap size.

Kevin made an excellent point about the lack of resistor between the output of IC3B and the input to IC3A. I should have spotted that myself. Try to arrange for something like 1k to 10k there, as that will help a lot with the oscillation.

Again, any suggestions for reading material on transient step load tests would be welcome.

Personally, I'd suggest that you don't need to worry about this just yet. Get the thing basically stable first. I test using a MOSFET and power resistor, with a pulse generator driving the MOSFET,  while observing the waveform on an AC-coupled 'scope.

I am using a 2x0-15v transformer which in reality reads 40v pk-pk - I am only using one winding, so I could use the other as a negative rail. If I did this, would I need to regulate it down to 2v or whatever (using 79xx reg)? I'm guessing so.

It would be a waste to use the whole winding just to create a low-current negative rail. Personally, I'd simply build another PCB (once debugged) and have a dual PSU - far more useful

As I mentioned, you could derive a negative rail from the incoming AC to the bridge - it would need two diodes and two capacitors. I'll try to attach a suggested schematic...

This would also have the added benefit of higher stability, as I do not completely trust the 555 inverter - it reads -1.6v not -2.

And I explained this in the first paragraph of my first post; the voltage is entirely dependent on the forward voltage of the chosen LED. They vary a lot. Typically 1.6V for red, 2V for green, yellow, orange, and ~3.5V for blue, white, "pure green" - you get the idea.

I also suggested you could try a VBE multiplier, which would be adjustable to match the 1.25V of the LM317 - see the attached schematic.

All the best,

Mark

[microbug]

OK. Thanks for your help, I am creating a new schematic (based on yours).
How would you calculate (formula / rule of thumb) what size decoupling caps to use in certain places? I know that for most logic-level ics and op-amps a .1uf ceramic as close to the device as possible is good, but what about other places (e.g., across the 5v6 zener on your schematic). The datasheet for the LM317 recommends a 10uf electrolytic for the output and this is what I used before for that purpose - would 10uf also be a good option for the cap after the zener?

Edit:
A couple more questions (directed at Mark Hennessy, but I'll take answers from anyone ):
- Why have you put a resistor marked as ~100r on the positive supply of the current limit op-amp (on the schematic you posted previously)?
- Why do you need the extra voltage drop on the output of the current limit op-amp (an led plus a normal diode)?
- What is the purpose of the diode in parallel with the 240r resistor next to the lm317 (protection?) ? OK, looking at the lm317 datasheet I found that it is for protection so if the input is shorted to ground current is not forced into the output of the regulator.

[microbug]

Bump.
Also, should it make any difference, I am using a 330ohm resistor in place of 240 next to the lm317 because +unreg (with my transformer) is about 19v, so a 330ohm will give my 5k pot more resolution (my calculator tells me the max voltage with 330ohm is 20.18 and with 240 is 27.29).
Edit:
Another question! I'm doing the current control stuff now. How do you work out the gain of ICB (differential amplifier)? Wikipedia only confused me further. A=Rf/R1 works in this case, but how do I know it will work in other applications?

[mariush]

You shouldn't use more than 120 ohm ... in some circumstances the regulator won't work properly if you exceed that value. 100-120 ohm is the recommended range of values.

The datasheet shows 240 ohm but that's for the LM117 regulator (the datasheet is for lm117/lm317).

If you want to increase the maximum voltage, just put a resistor in series with the pot but then you'll have higher minimum voltage.

[microbug]

OK. Is that because of the minimum load (10ma)? If that's the case, would the resistor marked 'should pass around 10ma' (going from -1.25v to ground) on Mark's schematic help to solve this problem?
Edit:
I'm debating whether to add a 7805 regulator for logic level stuff (breadboards etc). This would also be useful for a voltage divider for the current control pot. Opinions?
Thanks!

[Mark Hennessy]

How would you calculate (formula / rule of thumb) what size decoupling caps to use in certain places?

Often, it's just experience. Sometimes, manufacturers give recommendations. Sometimes, you determine what's needed during debugging. I'd encourage that, as you'll learn most from doing that...

I know that for most logic-level ics and op-amps a .1uf ceramic as close to the device as possible is good, but what about other places (e.g., across the 5v6 zener on your schematic). The datasheet for the LM317 recommends a 10uf electrolytic for the output and this is what I used before for that purpose - would 10uf also be a good option for the cap after the zener?

A cap after a zener is actually less useful than you might think, as the zener has a low slope resistance, which forms a filter with the capacitor. A "generic" 100uF might be OK there - you have to experiment.

The output cap is decided when evaluating the dynamic behaviour, but 10uF is probably a good start.

Edit:
A couple more questions (directed at Mark Hennessy, but I'll take answers from anyone ):
- Why have you put a resistor marked as ~100r on the positive supply of the current limit op-amp (on the schematic you posted previously)?

It's good practice. It forms a filter, which cuts down the noise on the +UNREG supply. It offers some protection against surges.  Should the IC (or bypass cap) go short, the resistor will smoke, but the fallout from that would be relatively small.

- Why do you need the extra voltage drop on the output of the current limit op-amp (an led plus a normal diode)?

I don't. However, you do need a diode so that the current limit works (as in CV mode, the op-amp output is hard up against the supply rail). Adding the LED gives you a convenient front-panel indication that you are in CC mode. Question for you: why not omit the diode, given that the LED is a diode?

All the best,

Mark

[Mark Hennessy]

You shouldn't use more than 120 ohm ... in some circumstances the regulator won't work properly if you exceed that value. 100-120 ohm is the recommended range of values.

The datasheet shows 240 ohm but that's for the LM117 regulator (the datasheet is for lm117/lm317).

Are you sure about that? I've never come across that before. I've always used 240 ohms as it gives a convenient ~5mA.

I've just checked a few datasheets from different manufacturers, and no sign of that... What manufacturer was you using?

[microbug]

Actually, I have the Ti datasheet open and it lists 'minimum load current to maintain regulation' as typically 3.5ma and max 10ma.

[Mark Hennessy]

Another question! I'm doing the current control stuff now. How do you work out the gain of ICB (differential amplifier)? Wikipedia only confused me further. A=Rf/R1 works in this case, but how do I know it will work in other applications?

Seriously, I would leave that until the voltage regulation is working. The golden rule of engineering is to break it down, thoroughly testing each part of the circuit before moving on.

 As part of this journey, you might be interested in prototyping this (from the National Semiconductor datasheet):