Help with noise in simple bi-polar power supply

[Matt C]

I'm looking for some advice in tracking down some noise in a power supply I'm building on stripboard.  So far it's a real mystery to me.  It's a run of the mill bi-polar  supply with both +-15Vdc and +-12Vdc.  I'm getting noise on the outputs that is leaking into the audio circuits and becoming audible.  Here are some details:

- I've been measuring on a scope with the ground wire connected to the chassis safety ground bolt, and with this arrangement I can see the noise on all the PSU outputs AND ground at the output.  So it seems like ground is somehow being contaminated in the PSU circuit
- noise is 120Hz
- gets more severe with more current, but still present with only the voltage regulator's idle current and nothing else connected
- changed slightly but still present when I used a different power transformer
- present when using both 78** regulators and LM317 adjustable regulators
- did not change after putting in fresh regulators, new (faster) rectifier diodes, and new (bigger) filter caps
- no change after connecting and disconnecting the PT's static shield lead to chassis
- no change after connecting and disconnecting PSU circuit common to chassis ground
- no change after rebuilding the circuit on a new circuit board
- no change when connecting only +-15V or only +-12V
- when looking at the rectifier output/regulator input, I see the usual small, sort of triangular wave of ripple, but there are what look like extra quick dips and peaks at the corners that seems to maybe correspond to this noise.  Tough to explain the shape properly, and I can't get a very stable image  on my scope to take a picture for you.

alright, if anyone has any idea what could be causing this weird ground noise, please let me know.  I'd love to finally get this thing working! Thanks for the help!

 

[JohnRoberts]

Since you haven't shared details I will guess... if your PS has reservoir caps that get charged through diodes, the current from charging the + supply cap resembles your waveform (but the picture is sideways too).

So my guess is your noise is coming from the ground organization- layout. You want to cleanly connect the voltage regulator grounds to each other and your circuit ground reference without the reservoir capacitor ground currents corrupting that ground quality. So cap grounds connect to each other, and transformer center tap (another guess?) but only to the regulator ground reference node though a single connection that isn't carrying AC current.

Perhaps search for past discussions here about power supply layouts.

JR

 

[RuudNL]

Maybe too low AC voltage from the transformer. (An 7812/7815 needs at least 3 V. more on its input than the output voltage.)
Maybe buffer capacitors are too small.
Maybe load current is too high. (For the given transformer/buffer C combination)
Difficult to tell, without the schematic and more details...

 

[Matt C]

I don't have a great way to post the exact schematic right now but here's one I found online that is almost exactly the topology I followed, pretty much straight out of the lm317 datasheet

power transformer is 18-0-18, center tap goes to circuit common.  there is around 25V at the regulator inputs so that's definitely not too low.  Input filter caps are 2200uF on each regulator, which should be able to handle the ~10mA of idle regulator current.

John your suggestions about layout are probably right, I'll do some research.  You're suggesting that the regulator grounds should be connected as directly as possible to the PT center tap, NOT the bottom of the reservoir caps?

 

[JohnRoberts]

I don't have a great way to post the exact schematic right now but here's one I found online that is almost exactly the topology I followed, pretty much straight out of the lm317 datasheet

power transformer is 18-0-18, center tap goes to circuit common.  there is around 25V at the regulator inputs so that's definitely not too low.  Input filter caps are 2200uF on each regulator, which should be able to handle the ~10mA of idle regulator current.

John your suggestions about layout are probably right, I'll do some research.  You're suggesting that the regulator grounds should be connected as directly as possible to the PT center tap, NOT the bottom of the reservoir caps?

NO... they all connect together but exactly how they connect matters.

It will help to understand what is happening if you visualize wires and PCB traces as small resistors, so the current flowing will cause small voltage drops everywhere. 

Looking at the schematic you grabbed, the transformer center tap connects to the ground node on the left side of the schematic. There is heavy charging current in those ground wires from the reservoir caps that will cause small voltage drops, but we don't care because we reference the regulator clean ground to the output side and connect that to the dirty capacitor common with only one path that has no charging current flowing in it to corrupt that clean ground reference.  So basically the dirty ground only connects to the clean end at one point, that becomes the ground reference. The voltage regulators reference wrt to that single point so by definition it is clean.

The ground symbols on both ends suggests they are the same, they aren't...  the left side is where the transformer center tap current return is connected, the right side is the output reference ground.

Your layout has some dirty charging current flowing in a clean reference trace.  Looking at the scope image probably the + reservoir cap.

JR

 

[Matt C]

You're suggesting that the regulator grounds should be connected as directly as possible to the PT center tap, NOT the bottom of the reservoir caps?

NO... they all connect together but exactly how they connect matters.

I think maybe we're getting at the same thing and I probably just didn't describe it well.  There's a point where the PT center tap enters the circuit board, which will be ground.  I want to run separate connections from this point to the reservoir caps (dirty) and to the regulator ground (clean), this way I minimize the amount of conductor that is shared by these two things, even though they do get tied together. 

I just made these changes to the circuit board and it seemed to do the trick.  Thanks for the advice!

 

[JohnRoberts]

Yup, that works.

We want to keep the sewage out of the clean water pipe.

JR

 

[Matt C]

I finally got this thing working properly, but I ran into a weird issue in the process, and I'm wondering if anyone has an idea why this was happening:

on the output of the lm317/337 regulators, I added a simple RC filter in an attempt to reduce noise.  I used 22r and 560uF.  When I applied power, the negative rail would sit at around -2.5V and VERY slowly move, with increasing speed, to it's final value of -15V.  This took probably 45 seconds.  It only did this when the PSU was connected to its load, and the positive rail never did this, it went straight to +15V.  I tried shorting out the 22r resistor in the RC filter and everything started working properly again.  Also for the record I did verify that the input to the regulator was correct immediately on power up, it didn't do this weird ramp up.

I don't really understand what was going on here, but it seems like maybe the modules the PSU was feeding (analog synth oscillators and related modules) didn't like seeing the RC filter on the PSU output.  After all that hassle I'd really like to know what was happening, so if anyone has some insight I'd appreciate it!