Need help with hum in simple LM317 power supply

[Bowery Electric]

Hello,

I built a simple pedal power supply using LM317s.  The circuit is very typical, has six sections fed by one transformer (schematic below).  Now I read that it is a good idea to have separated supply sections.  Supposed to be like batteries?  But in my case it hums like crazy.  When I connect the negative ends together things quiet down.  When I choose one section and daisy-chain from it things are also quiet.  Using one bridge to power three regulator sections is mostly quiet, but not as quiet as ganged negatives or one section and daisy-chained hookups.  This is all on perfboard so it is easy to change things up.  I also have different filtering on sections (470mf vs. 1000 vs. 2200) and no real difference there.

So why does the isolated configuration cause so much noise?  I thought that was supposed to be the slick deal.  I assume the fault is in my circuit.

[psychedelicfish]

You have ground loops. These act as aerials and pick up all the hum. Here's a quick crappy diagram of your problem, the ground loops are in red:

You should read about pedal power supplies somewhere, I think geofex has a good article on them, but I'm too lazy to find it

[tubegeek]

EDIT: psychedelicfish got in there ahead of me, with a picture no less!

You most likely are getting noise via a ground loop: I think your pedals each are grounded twice: once via the shield of the connecting cable (grounding each pedal to each other) and once via the "-" end of your power supply (grounding each pedal to its own power supply section.)

Also: there are some things that are important that aren't visible from the schematic of a setup like this. You have to watch where the currents are flowing in each of those grounds, too. You need to make sure that the power supply return current isn't being routed through any wiring that the audio signal is also routed through. This means that each power supply "-" should be connected as directly as possible from the regulator's filter cap (25 mF) "-" end to the pedal's "-" supply connection. And that all the pedals should have "best practices" grounding internally, which they may well not.

Some very detailed info on "best practices" of grounding here: http://www.rane.com/note151.html I'll warn you, it's complicated. Essentially the idea is to use the pedal's box and the cable's shield as just that, shields, and use the power supply "-" and the signal "-" as just that: references for the power supply "+" and audio signal "+" . This would mean, among other things, making sure that exactly one of the "-" connections made contact with exactly one box in exactly one place. Then the rest of the shield connections should only connect box to box to box and not carry any signal or power supply connections, so you'd have to use isolated jacks everywhere and I guess 1/4" TRS connectors to isolate the shields from the "-" connections. Not a typical setup, I know. Among other things, that'd totally remove the opportunity to do the "inserting TS plug turns on your pedal" trick. Which is actually a good thing (that trick always puts power supply current through an audio wiring route, as mentioned above, a Bad Idea.)

I'll be honest, I don't have a good recommendation for which "-" or which box, but I do know that the connection should be immediately inside the box where the chosen cable enters it.

I realize this isn't super helpful, maybe someone else can come up with an easier way to be sure this supply won't hum that wouldn't require so much modding.

ANOTHER EDIT:

As I think about this, I guess there's at least an easier way to explain what you need to do: you need to isolate the audio "-" between devices that also have isolated power supply "-" connections.

Because, each (typical guitar pedal type) device is always going to use its power supply "-" to establish its "audio ground" reference. But with the devices connected by, typically, unbalanced interconnections, they are now connected by those cables to both their own ground reference AND the ground reference of devices that are supposed to be isolated from them.

Notice that NONE of the parts of your circuit, as drawn, connect to ground anywhere. This is actually jim dandy if it is true - those power supplies should all float with no shared connections. I have a strong feeling, though, that SOMEWHERE each power supply is connected to a "ground" that is showing up in the audio circuit, because the hum is there - they shouldn't hum if the supplies are all each actually floating, and only the audio cables are connecting the devices together in any way. Hunt for unnoticed connections between the power supply "-" and box, both inside the devices and inside the power supply box too.

The only connection to the power supply's box should be to the safety ground of the incoming AC circuit. Nothing else should connect to that box.

[earthtonesaudio]

First know that isolation and regulation are techniques to mitigate hum, not eliminate it.  That said your circuit does not have isolated outputs.  For that you need separate secondaries or separate transformers. 

Since you said it's not too bad with the daisy chain I would go that route as the cheap solution.  Or invest in separate secondaries for your transformer as the expensive solution.

[Bowery Electric]

Thanks to all for your comments and for all of the great threads that I read while searching for this little project.

All of the outputs of my power supply sections are isolated, as has been described in several of the online PS build tutorials including RG's.  They "float".  But here this results in hum.

As for ground loops, RG's Spider PS shows that having isolated regulation and distribution eliminates this, but it does not in my case.  

Altering the pedals was suggested.  How do the manufacturers of pedal power supplies now available on the market do that?  Do they not have to contend with the customer's pedals as they are?  

I keep reading about how to make a quiet pedal power supply and have followed the instructions but my reward is noise UNLESS I do "the wrong thing" and tie the negatives together.  

As for the daisy chain, the negatives of all pedals are tied together through the chain in parallel with the shields of the interconnecting cables at the PS inlet's connection to the input jacks but do not cause loops?  

I am cornfused!

Also, tubegeek, there are no unnoticed connections to the PS box because there is no box.  It is all on perfboard at this point and every connection has been placed there by me.  The only common point between any regulated section is the single transformer secondary winding.  After that I use six bridges that lead to six floating regulator circuits.  Only when I solder on a link between negatives is there a common connection between sections.  I did a lot of searching and reading before I prototyped this and with the exception of using a common secondary winding I have built six completely separate floating regulated DC supplies.  This ain't my first electronic project but it is my first attempt at the pedal power supply.  Tube amp power supplies are much simpler!  Also, obviously I am not an EE, just a hobbyist who can't go out and work on the car because it's too hot!  

earthtonesaudio, I have another power transformer.  This evening when I get home I will add that  to the mix so that I have two sections that are completely isolated all the way back to the mains.  I will report the results.  Thanks!

[R.G.]

So why does the isolated configuration cause so much noise?  I thought that was supposed to be the slick deal.  I assume the fault is in my circuit.

I think it's because the AC inputs to the two bridges are swapped. If you reverse the pins the AC is connected to on ONE of your two bridges, the hum may go away.

The input to a rectifier bridge is not really an AC connection. It's a switched DC connection, the diodes doing the switching. If you make two bridges work from the same AC supply, and connect their outputs, there are two different ways they connect. One way works fine. The other way, the diodes are trying to connect different AC phases to the same place.

[amptramp]

You really need separate secondaries to achieve isolation.  The grounds of the power supplies are always connected by the diode bridges and any differences in diode drop will modulate the ground lines with hum (or buzz since harmonics are generated by rectification).  I would consider using a bunch of cheap doorbell transformers to replace the single transformer.  They are usually 10 VAC or 12 VAC and have enough series inductance that they can be shorted without drawing excessive current.

[Bowery Electric]

So why does the isolated configuration cause so much noise?  I thought that was supposed to be the slick deal.  I assume the fault is in my circuit.

I think it's because the AC inputs to the two bridges are swapped. If you reverse the pins the AC is connected to on ONE of your two bridges, the hum may go away.

The input to a rectifier bridge is not really an AC connection. It's a switched DC connection, the diodes doing the switching. If you make two bridges work from the same AC supply, and connect their outputs, there are two different ways they connect. One way works fine. The other way, the diodes are trying to connect different AC phases to the same place.

OK, but a bit confused in the application of your comments.  In my build I have six bridges for six regulator circuits and none of the outputs are cross-connected.  My little schematic does not show any cross-connections of the circuit on the DC side of the bridges.  As for the AC side of the bridges, they are all connected the same to the same legs of the secondary of the transformer.  No funny business at all.

It is when I DO deliberately connect the DC negative rails together *as a test* that hum is greatly reduced.

I have performed another test where I took three of the regulator circuits and hooked them to ONE bridge.  This results in less hum three floaters but it is not as quiet as the test immediately above.

See?  This circuit is not responding in the way I have read it is supposed to.

Do you need separate and complete schematics for all of my tested conditions?  I have AutoCAD and I am not afraid.

[Bowery Electric]

You really need separate secondaries to achieve isolation.  The grounds of the power supplies are always connected by the diode bridges and any differences in diode drop will modulate the ground lines with hum (or buzz since harmonics are generated by rectification).  I would consider using a bunch of cheap doorbell transformers to replace the single transformer.  They are usually 10 VAC or 12 VAC and have enough series inductance that they can be shorted without drawing excessive current.

Yes, I considered this upthread.  I will add at least a second tx and test two completely isolated circuits this evening.  That will be an interesting test!  That the circuits are all common through the single transformer secondary makes sense to me.

This is turning out to be an interesting project!   

[tubegeek]

I have AutoCAD and I am not afraid.

OK, looks like my comments were not close to the mark, I was guessing, and guessing wrong. Sorry.

Here's what I think is the actual problem that is causing the noise, rather than the theoretical problems that MIGHT be causing it: it's what amptramp said: the bridges are not all giving you the exact same diode drops and this results in a residual 120Hz (possibly 60, but I think 120 is more likely) component of difference between each power supply's raw rectified voltages. I think some of this is present even AFTER regulation, otherwise, the supplies would have only DC differences and you wouldn't have heard anything wrong. Then when the "-" sides are left floating,  the pedals are all operating with discrepancies between their common signal grounds and their varying-with-respect-to-each-other power supply "-" connections, driven by whatever each of their circuits' impedances presented to their power supplies are. These discrepancies are showing up in the audio. There are indeed ground loops in your system as soon as you common the regulated sections' "-" ends, but as you have correctly pointed out, ground loops are not the cause of the problem you are hearing and whatever problems they can cause are - apparently - minor compared to the other issue.

When you ground all the power supplies together, here is what I THINK is happening: there is - apparently - still a variation between the regulated positive supplies (tying the grounds together doesn't change that.) Some portion of that is the ripple - ripple that must be there because you heard it - still, because we haven't changed anything within any of the regulated sections, so their performance won't have changed.

What we HAVE done is moved all the variation from the "-" side, where there was therefore some injection into the ground system via the competing ground references, to the "+" side where we now have each circuit's own independent power supply ripple rejection (PSRR) working for us. So less hum makes it to the output when connected this way. Each pedal has no shared reference to the "+" supply of any other pedal inherent in the audio circuit and - if designed to - should also be less sensitive to ripple on their "+" supply than their ground reference. As a result, less hum comes out the output jack.

Unless I'm missing something. Again. In which case just ignore me from here out, I'm usually quite harmless as a rule.

Tube amp power supplies are much simpler!

Not when you connect a bunch of them together! But, yeah, they sure are.

[Bowery Electric]

Thank you, tubegeek.  I'm pretty sure your assessment is correct.

OK, I want to establish that although I built six regulator sections I can only test three since I only have three pedals.  So I have been making mods based on threes.  I want to now post up schematics of the original circuit and my two mods as described above.  A forth pix will show the mod I plan on doing tonight with three separate transformers. I'm doing this because if something is learned here some other future guy or gal will find it when (s)he has a PS problem and so I like to document these things as completely as possible.  So...

Original circuit:


Mod 1 (negatives ganged):


Mod 2 (one bridge):


And tonight's exciting mod to be:


I'll post up the results.  This is a great forum, yall.

[Bowery Electric]

Update.

Quick check before tearing into my little proto shows .02 volts offset between negative rails of two sections of the 3 floating sections (that's the "original circuit" schematic above).  I'm no expert but that can not be good.

I thought I had a 12.6 tx around here but I could not find it so I moved ahead with adding a second transformer to make two completely separate power supplies.  Hooked it all up, connected my old Electra compressor pedal (sensitive to hum) and my DOD Grunge pedal (VERY sensitive to hum, the coalmine canary) and hooked that up to a Princeton and powered it with the batteries for baseline (quiet) and then plugged them into the two discrete power supplies.  Quiet!  Just like the when on batteries.

I guess I'll be getting one of those multi-secondary transformers.  I guess I never thought the sections would look back through the bridges at that single transformer winding and then deliver variations along the negative rails due to differences in the bridges.  That does seem to be what is happening, right?

[armdnrdy]

I think that there have been posts here.....R.G. maybe?....and posts at other sites that basically explain what you now know.

I did a bit of research on power isolation to connect a voltmeter display to an adjustable bipolar power supply. The voltmeter wanted a separate supply than what was feeding the power supply. A separate diode bridge was not enough!

The ultimate in power isolation is a transformer!

I believe that there are DC/DC converters that could provide isolation but.....that comes at a price.....They're expensive!

If you have the room, Transformers are the way to go.

Small PC mount transformers can be pretty reasonably priced if you shop around.

[tubegeek]

.02 volts offset

Was that AC or DC you measured? I'm guessing DC because of the word "offset." (Look at me, guessing again. Hoo boy.) Did you also look at the AC measurement?

The .02V travels in a loop from power supply to pedal back to power supply, up along the "-" connection and then back along the shield. If there is AC in it, well, we know there is because your canary heard it, it shows up clearly in the audio.

Nice work on the documentation, thanks!

[Tininho]

Hi, great thread  (and I noticed it is old)

So.. I built something very similar and I have the corresponding ground loop hum. My power supply actually looks like Bowery's mod 2, ie. a single transformer, to single bridge rectifier, to multiple regulated outputs.
Although there are other schematics looking like this on the net, no one seems to notice the ground loop problem.. Finally I see I'm not alone (hence bringing this old thread back up - I hope I can skip "you have a ground loop, disconnect your mains earth, modify your pedals, cut your signal cables" suggestions)

I totally understand what's wrong about my circuit, and btw and thank you for trying and sharing all the other mods. That has saved me a lot of time. As I understand, the best solution seems to be multiple transformers, and I agree.
But.. all due respect to that solution, it does not work for me. I built my power supply 100% out of recycled components, I am proud of it, so I will avoid buying transformers at all costs (ironic idiom) - Lets say this is a "if you were in a deserted island..." exercise.

How would you break that ground loop then? (recap: I'm suggesting to "fix" mod 2 - one bridge, because i think that's the cheapest circuit)

I am thinking of putting a decoupling inductor between the "-" connections of the regulator sections. My hope is that it will still bring all the outputs to the same DC ground reference (always good, right?), while blocking only the parasitic AC, which will be forced to take the cable-shielding route, as it was originally supposed to.

Nonetheless, I am confident that I am not that smart and the solution cant be this easy. I also have never used an inductor before and I have no idea how big or small it should be for this purpose.
So what do you guys think of that? any suggestions?

I'll probably test it anyway one of these days, if someone is interested in that please say so

Thanks for your interest and btw this forum is awesome, I've been reading it for years even though this is my first post.

[PRR]

Welcome.

> a single transformer, to single bridge rectifier, to multiple regulated outputs.

That works perfectly well 98% of the time.

So why are you special?

Have you tried your supply with _one_ pedal and made sure it is _clean_? Two pedals?

Maybe you have just one beloved pedal which "does not play well with others" on the same supply?

It is also good to lash-up six D-cells (9V) and try your pedal board on that known-clean common source.

And "all due respect", I will not admit how many times I have torn the boxes apart looking for a buzz, only to find it was a bad patch cable all along.

> a decoupling inductor

Would work BUT-- a small inductor won't do much, essentially no-difference. An inductor large enough to block 50/60Hz will be essentially the same iron+copper as a Power Transformer for the same load, and a non-standard part, so that's more costly than just salvaging more transformers.