Oscillations on the output of power supply (LM317/LM337)

[Ahura Mazda]

Hi everybody, I've spent a few last days building linear supply with adjustable voltage using LM317 and LM337 combo. But I noticed some strange behavior.

Before changing Cout (2200uf):

Thinking it would be nice to add as much capacitance as possible, I firstly put 2200uf on the output. I missed that datasheet says 1uf recommended for better ripple rejection at low frequencies, but not needed.
Either way, the circuit seemed to work until I AC coupled the signal and saw some strange behavior I still don't understand. Random spikes.

Another example,  but here I understand LM317 is no longer regulating since headroom of 3V is lost.

Removed caps / 1 uF / 10 uF:

Signal is oscillating, frequency of arounf 4 Mhz, with or without load. The difference is, as higher in capacitance I go, the higher the Vout until the oscillation occurs. Also, both outputs are oscillating.

------

Steps taken:

Pretty certain all connections are good on the PCB.
Disconnected Q2 and Q6 thus also eliminating Q5 and Q5. Still oscillating.
Cut open connection on R8, eliminating Q1 (so Q3, too). Still oscillating.


Any help or suggestion for the oscillation would be greatly appreciated. Thank you guys.

[fcb]

Layout could be a contributor. How close are the caps to the regulator?

Also have you tried removing the cap on the ADJ pin and/or bypassing it with perhaps 100nF?

What's your static load? They like to be loaded a bit (10mA?) - although you indicate it's the same loaded and unloaded, your load might be quite inductive. Try a small load right up close to the regulator.

Have you tried a zobel network on the output (perhaps 2R2 in series with 220nF-pure guess).

[Ahura Mazda]

Well, it turns out it's the pots. Brand new, yet obviously not made for this purpose. I've swapped them with precise trimmers just to test whether they were the problem, and they were. Now the signal looks quite clean, would say under 1mVpp noise. Just need to figure out what potentiometer to buy so I can actually mount them on the case. Am worried they might do the same thing.

I thought the signal is common for both outputs, but apparently it's not. After changing one, I checked the outputs, and one was crisp clean, while the other was oscillation.

Also have you tried removing the cap on the ADJ pin and/or bypassing it with perhaps 100nF?

As a matter of fact I did with another 1uF. 100nF probably wouldn't change much since it would be 10uF in parallel with 100nF, right though, and the freq?

Have you tried a zobel network on the output (perhaps 2R2 in series with 220nF-pure guess).

No, but now I must research this. I don't know this stuff.

I appreciate you responding. Thank you.

Edit:
Oh my, I forgot to write the most important thing. I couldn't get the signal not to have huge ripple without Cout being more than 470uf. With pots changed, the ripple was still present.

[magic]

Unlike LM317, negative regulators tend to have some minimum requirements regarding output capacitance and possibly ESR. Check your LM337 datasheet.

The second photograph looks like the positive regulator might be dropping out due to insufficient input voltage, if the bottom channel is scaled vertically more than the top. Not sure how pots would affect that.

[Ahura Mazda]

Thanks for replying.

Yeah, I thought the same thing about the insufficient headroom.

The pots are noisy. I touch the lead, signal jumps up and down. It looks like I had problem with both pots and Cout. When switching the pots, no signal was jumping around, and when I increased Cout, there was no ripple.

Now, I might have made a mistake confusing the videos. In the last one, there might have been load connected, I'm just not sure now.

The question I have is, should there be a ripple on LM137 when resistive load is connected, but with Cout<470uF in my case? Looking by datasheet, Darlington pair is always on, supplying current as needed, so I in fact I see no reason for the filter cap. I've connected 680 Ohms as load, drawing around 17 mA at 12V, and yet there was that ripple.

I'm just not knowledgeable yet to answer precisely, but switching pots and inserting high value cap solved my problem.

[Labrat101]

Remove the Diodes  D7 & D8 that sits across the in and out of the lm337 & 317
and put them between the out to centre ground .
The lm337 & Lm317 are not matched so there is a voltage swing .
 So causing an unbalance . what you are seeing is the higher voltage pulsing in reverse .
 To get a perfect match between the plus 0 Negative  will always be slightly higher on the  Lm317 side.
so placing the diodes across the outputs in reverse polarity will stop the feed back.  ie across the two   C 470uf
 This is showing on the funny ramp on your scope .
This is also recommended by the manufactures as well

The diodes as they are now are causing the internal overload to trigger
on reverse voltage .
once those diode are moved to the right place the circuit will function correctly .
D4 & D6 are also not necessary  and should be moved as per picture
& MUST remove  D7 & D8
Sorry about the bad paint over . did it by hand
 make sure they are in the right direction .

[Ahura Mazda]

Hey, thank you for replying.

Are you sure? I've browsed a lot of LM317-LM337 combo psus and not once have i came across that. Where can I see the recommended schematics?

The lm337 & Lm317 are not matched so there is a voltage swing .
 So causing an unbalance . what you are seeing is the higher voltage pulsing in reverse .

I guess I can add additional diodes and try, but for example, when set the output of LM337 to the lowest, the signal on Vout of LM317 will look the same. Shouldn't the signal change if what you are saying is true?
Also, I see that both signals are clear now and only when I set the output Voltage too high do I see the funny ramp, as you say.

Quote from TI:

Protectiondiode D2 is recommendedif Cadj is used.The diode provides a low-impedance discharge path to prevent the capacitor from discharging into the output of the regulator. Protection diode D1 is recommendedif Co is used.The diode provides a low-impedance discharge path to prevent the capacitor from discharging into the output of the regulator.

ONSEMI:

LM317  with  the  recommended protection diodes for output voltages in excess of 25 V or high capacitance values (CO > 25 uF, Cadj > 10 uF). Diode D1 prevents Co from discharging thru the IC during an input short  circuit.  Diode  D2  protects  against  capacitor  Cadj discharging through the IC during an output short circuit. The combination of diodes D1 and D2 prevents Cadj from discharging through the IC during an input short circuit.

Edit, I'll definitely try now. Won't hurt to try either way. Thank you again.

[Labrat101]

Hey, thank you for replying.

Are you sure? I've browsed a lot of LM317-LM337 combo psus and not once have i came across that. Where can I see the recommended schematics?

The lm337 & Lm317 are not matched so there is a voltage swing .
 So causing an unbalance . what you are seeing is the higher voltage pulsing in reverse .

I guess I can add additional diodes and try, but for example, when set the output of LM337 to the lowest, the signal on Vout of LM317 will look the same. Shouldn't the signal change if what you are saying is true?
Also, I see that both signals are clear now and only when I set the output Voltage too high do I see the funny ramp, as you say.

Quote from TI:

Protectiondiode D2 is recommendedif Cadj is used.The diode provides a low-impedance discharge path to prevent the capacitor from discharging into the output of the regulator. Protection diode D1 is recommendedif Co is used.The diode provides a low-impedance discharge path to prevent the capacitor from discharging into the output of the regulator.

ONSEMI:

LM317  with  the  recommended protection diodes for output voltages in excess of 25 V or high capacitance values (CO > 25 uF, Cadj > 10 uF). Diode D1 prevents Co from discharging thru the IC during an input short  circuit.  Diode  D2  protects  against  capacitor  Cadj discharging through the IC during an output short circuit. The combination of diodes D1 and D2 prevents Cadj from discharging through the IC during an input short circuit.

It will work I made similar circuits and they work fine . change as per my drawing and it will work . the diode across is only for single . not duel output the diode I moved will protect the circuit just make sure they are
large enough to take a surge .

[Ahura Mazda]

Nope, same thing happening.  At 15V on output, this starts happening. Vin was 18.18V, so that is that 3V headroom needed for regulation. Thank you for suggestion though.

I think the circuit is working as it is supposed to.

Edit: The noise literally doubled to around 1,3-1,4 mV. Though I can't see the spiked nicely. It's even greater.

[Labrat101]

Nope, same thing happening.  At 15V on output, this starts happening. Vin was 18.18V, so that is that 3V headroom needed for regulation. Thank you for suggestion though.

I think the circuit is working as it is supposed to.   

Edit: The noise literally doubled to around 1,3-1,4 mV. Though I can't see the spiked nicely. It's even greater.

Hi sorry it was very late when I looked at the schematic .
you have other errors in the drawing  R3 should be at-least  240 ohm and the cap C1 & C3 should be
ceramic Not electrolytic  try 0.1uf  .. What is the value of your Pot ??
Also use this link to do the calculations it will make life easer .
https://circuitdigest.com/calculators/lm317-resistor-voltage-calculator .
Also why did you put R11 & R12 . they are too low a value and will draw a lot of current .
I am not sure were you found this circuit from but its very power hungry and properly only 60% efficient.
What sort of load are you aming for ie Amp and voltage output required ?? .
You have also put a current limiting and voltage . the  Q4 & Q6 Will switch on hard due
power supply connection being wrong

Also the circuit shows the bridge 4 diodes .
Where is the power coming from for the LM337 ??  I guess you stuck it to the neg of the
 bridge .. 
 Please show the complete Power source connections 
 For Dual power supply to work correctly  . You must use a centre tap transformer
 ie  15v 0 15v  . having a floating ground can work fine .
     But Not as you have done it ..
   THIS TYPE OF CIRCUIT MUST HAVE ITS GROUND CONNECTED
    to the transformer centre TAP
 This is an example I found as I was to lazy to draw it out . That is correctly Powered
  BTW 

I think the circuit is working as it is supposed to.   

     Connect the power supply as per the drawing and all will become clear.
     Plus the other recommendations

[Ahura Mazda]

you have other errors in the drawing  R3 should be at-least  240 ohm (...)

What, no. Value is to supply minimal load current for maintaining regulation. 1.25 V / 120 Ohms is around 10mA. Datasheet says, 3.5 mA is typical, yet I've read that it's for more expensive version of LM317. Either way, for  240 Ohms, it would draw around 5 mA and I'm not saying it won't regulate properly, but judging from other's experiences, 120 Ohms is a better choice.

(...) and the cap C1 & C3 should be ceramic Not electrolytic  try 0.1uf

All datasheets suggest putting 1/10uF electrolytic capacitor.
Quote from ONISEMI: "A   10uF   capacitor   should   improve   ripple rejection about 15 dB at 120 Hz in a 10 V application."

What is the value of your Pot ??

5k.

Also why did you put R11 & R12 . they are too low a value and will draw a lot of current .
I am not sure were you found this circuit from but its very power hungry and properly only 60% efficient.
What sort of load are you aming for ie Amp and voltage output required ?? .
What sort of load are you aming for ie Amp and voltage output required ?? .

To discharge the caps after power is switched off, they are too big actually, takes ages since they almost draw no current. Since neither output voltages are high, and I'm the only one using this, I just put big value.

Don't know about efficiency, but it depends on many factors, one being Vin-Vout and Iload. I'll be using for many things, don't know.

You have also put a current limiting and voltage . the  Q4 & Q6 Will switch on hard due
power supply connection being wrong

You got something wrong. Q6 will be fed through only Q1, and Q4 won't turn on all until around 7 A are passing through the R9.

Also the circuit shows the bridge 4 diodes .
Where is the power coming from for the LM337 ??  I guess you stuck it to the neg of the
 bridge ..
 Please show the complete Power source connections
 For Dual power supply to work correctly  . You must use a centre tap transformer
 ie  15v 0 15v  . having a floating ground can work fine .
     But Not as you have done it ..
   THIS TYPE OF CIRCUIT MUST HAVE ITS GROUND CONNECTED
    to the transformer centre TAP
This is an example I found as I was to lazy to draw it out . That is correctly Powered
  BTW 

It is center tapped transformer. You can see all the lines on the schematics. AC+ and AC- are from the output of the transformer, and GNDTAP is from center tap.
It is the same circuit with added current boost circuit, short circuit protection for power transistors, improved ripple rejection and transient response, and plus the diode protection for the ICs. Absolutely the same starting point.

I ask you to tell me why what the circuit is now doing is not as it should be?
1. No signal bouncing due to new pots and no ripple on the output.
2. After Vin-Vout falls below the headroom, IC doesn't regulate.
I see it's all good.

[rdl]

R3 and R5 are fine at 100-120 ohms, higher will usually work but is only in spec with LM117. I have never seen a datasheet with anything but electrolytic for the adjust pin bypass cap.

Do you have the wipers of the adjust pots connect to one end or the other? I saw a web page not too long ago which showed a better way to connect the adjust pot, but I can't find the link now. I will look more.

As for making the LM337 fixed, I fail to see the point but maybe I'm misunderstanding. AoE 3 has a schematic for a dual tracking lab supply which uses just one pot to adjust both regulators, maybe something like that is what is meant.

Link to free Art of Electronics 3 chapter 9. Schematic is on page 608. Anybody thinking of building a power supply should check this out.

http://artofelectronics.net/wp-content/uploads/2016/02/AoE3_chapter9.pdf

[Labrat101]

R3 and R5 are fine at 100-120 ohms, higher will usually work but is only in spec with LM117. I have never seen a datasheet with anything but electrolytic for the adjust pin bypass cap.

Do you have the wipers of the adjust pots connect to one end or the other? I saw a web page not too long ago which showed a better way to connect the adjust pot, but I can't find the link now. I will look more.

As for making the LM337 fixed, I fail to see the point but maybe I'm misunderstanding. AoE 3 has a schematic for a dual tracking lab supply which uses just one pot to adjust both regulators, maybe something like that is what is meant.

Link to free Art of Electronics 3 chapter 9. Schematic is on page 608. Anybody thinking of building a power supply should check this out.

http://artofelectronics.net/wp-content/uploads/2016/02/AoE3_chapter9.pdf

What you say is correct . But one has to take into account that the details you are referring to
are for the Texas OEM's  which are slightly different . In this manual which is now 20 ago.
In the real world most of the LM's are not OEM and are of another origin .
From my past experience of power supplies using a ceramic works far better . on the sense pin
 upto 10uf will work even on a NON OEM .
And Using op amp's is a far better way to go.
 One of his scope reading looked like the clipping of the internal LDO  which tells me that these
 are not the t series . and maybe rebranded which will work OK but they need bigger heat sinks.
 That why I suggested the 240 ohm . And the calc link .
I have made so many power supplies I have lost count over the last 50yrs.
 I used a LM317t  a few weeks ago to make a precision 3.3000v that had to be drift free.
 electrolytic caps 1uf to 10uf have a high ESR .  Ceramic have a better characteristics. 
  This type of duel power supply has been around for yonks .
 But today many of the LM317 & 337 are not equal and when paired there can be a noticeable
difference  . ie Example >>The Pos. +14.90v & Neg. -14.45v  with identical parts and layout.
 The error may not bother you but its going in wasted heat & power loss efficiency .
 
 
    RNS

[fcb]

Example >>The Pos. +14.90v & Neg. -14.45v  with identical parts and layout.
 The error may not bother you but its going in wasted heat & power loss efficiency .

LM317 and LM337 work by servoing the output to keep 1.25V between the ADJ and OUT pins.

What you describe (14.90-14.45=0.45V) is likely not the fault of the LM3x7 but whatever you have hanging on the ADJ pin.

[Labrat101]

   I did write 'Example' ...   
 

but whatever you have hanging on the ADJ pin. 

    That was the funniest thing I have hear all day
   ((Personally my ADJ pin works just fine Thanks for asking )) 

  Some how no one has read correctly anything that has been written.

I / we don't have any such problems with power supplies .  Building or repair. etc.
 Its the other people that have trouble .
 First He has Never supplied adequate info.
1 .. was it build on a push board ?
2 .. Are the components new or from junk  S/H parts ?
3.. Was that scope ever Calibrated & reset before the test ?     
4 .. Have the probes been checked .. were they set to 1 or X10
5.. what were the exact setting of the scope  during the test .
6.. Only a drawing of the schematic was shown .. Not the Actual Built assembly . !!
   How do we Know, That it was wired up correctly ?   
 
 SO Far it's only been hear Say .. 
 
 He has ask for help and not given enough info .. The schematic should work Maybe not perfectly .
 So Far the Info shows some images of a scope screen which could rightly  displaying
 a Faulty Scope  Problem ...        or Human error more likely .   
There is no clear pictures of this layout complete with showing all the connection .
 And he is assuming that we Here on this Forum all HAVE a magic crystal Balls showing
what's on his desk /Bench .   
 

[Labrat101]

  All is Ok . 

BTW  I never said 

Q4 and Q6 would burn up because improperly wired.

I said they would switch hard .. As opposed to smooth.  ( linear )
 
Enjoy Have a 

 RNS ₁₀₁

[rdl]

How is the circuit currently assembled? Is it on a breadboard or point to point wired so that changes can easily be made? Do you believe that the regulator chips are genuine?

[Labrat101]

Hi ,
I am not sure why I am doing this But here it is.
I ran LM317 under a stress test .
 15v @ 1.5 amp as MAX rating
Ch 1 set 200mv  @ 2ms

 to show the shutting down .  or over heating
 
  with your values .
   is this same as you photo .??

Which emulator did you use to draw your circuit ??

[Ahura Mazda]

Thank you for replying.

How is the circuit currently assembled? Is it on a breadboard or point to point wired so that changes can easily be made? Do you believe that the regulator chips are genuine?

It's PCB, was since I post one. Chips are genuine. Bought from credible shop here. I've attached my assembly, and all connections are 100% made even though it looks quite ugly due to my inexperience.  PCB layout is debatable and rookie in itself; I later saw I could have made regulators more distant. I've added ferrite beads on the inputs and outputs, as for Jim Williams application note I saw, on how to reduce noise on voltage regulators.

Hi ,
I am not sure why I am doing this But here it is.
I ran LM317 under a stress test .
 15v @ 1.5 amp as MAX rating
Ch 1 set 200mv  @ 2ms

Thank you, I appreciate it. As you can see, I've added heatsinks. When drawing 1,5 amps, or any higher than a few hundred mA, transistors take the hit. Even when drawing 2 A from both sides both, LMs are cold, and transistors get nicely warm. So , I don't think that any of the LMs are getting shut down. I think it's from not sufficient headroom to regulate. Also note that NPN power can take 7A, that's why R9 is 0.1 Ohms, so it will automatically turn on Q4 an make LM317 supply more current, thus not burning the transistor. Same goes for the other side, with PNP's maximum draw being 3 A.

My reply #8 shows when I just make the output 15V, with input being 18.18V, and losing headroom. Then the second one is me increasing Vout.

I did say that even when drawing high currents, with enough headroom, output voltage is perfectly stable. I'll post some photos later today when I get back from university, giving more on current draw from the chips.

Edit: I've forgot to mirror the layout, so I got into some trouble later with LMs. Had to mount them upside down on the heatsink.

[Labrat101]

  OK
First The top board layout looks nice . I can see you have put a lot of work into this First attempt .
 The soldering side . Well . First I will make a Friendly suggestion do not power it up again until
 You have removed some of the Solder with a solder pump or Solder wick.
 You have used enough solder for 10 such boards.
 I zoomed in on your soldering photo . there are at least 3 near shorted tracks and several bad joints .
 yes I know you said that you were new at this.
 I don't know what type of soldering iron you are using . But by looking the solder wire that you are working
with is properly too thick for your iron .  maybe your working with 1.5mm
 if so find some 0.8 mm 60/40 Multicore CF10 flux 2%  not the lead free stuff .
 I recommend that you use some flux and remove all the solder and re do it ..
    Sorry . yes its a bummer  .  (look at it as a learning curve and not a punishment )
 when complete clean the entire board with a toothbrush (old) and Alcohol to remove any stray blobs etc.
 90% of all electronic problems are due to bad connections & dry joints .
 I think after you have done that.  Things will look different electronically.
  Reducing Risks of any project   ''The Dreaded  Magic Smoke '' 
 & the First Law of Murphy    .

  1 Question what are you going to use this power supply For ??  or to power ? .
    Is it by any chance A bench power supply ??

UPDATE .. on one of your earlier post you ask were did I find the 240 ohm value .
              A lot of the data sheets posted on the web had a typo and was updated a long time back but most of Google searches
             did not include the newer revisions . Photo .