1kHz noise, how to get rid of it?

[WickedBlade]

Hello,

my project involves both audio and digital path, and I'm getting some noise from the digital path into the audio path. I'd like some advice as to how to minimize it.

Now, for further details. I'm using one CORE, 2 DOUTs, 1 DIN, and a custom relay board that is triggered with some of the DOUT outputs (I'm using darlington arrays and everything). The MIDIBOX modules all come from Mike's shop and are unmodified except for the power supply, which was taken off the CORE module (I had the same problems when it was still on the CORE module though).

On the audio path I'm using a dual opamp to make 2 impedance buffers, and the relays I mentioned before are used to route outboard effects in and out of the signal path (I'm basically building a MIDI effects switcher). I've done all I could think of to avoid ground issues (ground loops and all), I'll come to it later.

My problem is that there is some noise coming from the digital space that goes into the audio path. The noise is basically 1kHz, and I've determined that it is directly related to the frequency to which I've told the PIC to update the DIN/DOUT (and it's there even if I don't actually call updates for the DOUTs). Taking the DOUTs or DIN off the system makes the noise disappear, I think that's proof enough. I've made sure that each module is acting for some part in the noise, and the sum of them makes the overall noise worse, of course. For the record, I did put bypass capacitors on the shift register power supply pins.

As far as power supplies go, I'm feeding 12V DC to my system, which gets first through a diode bridge (just to be safe), then to a 2200uF cap, and then splits to feed 3 power regulating circuits.

An image is worth a thousand words, so you'll find attached the schematic for the power supplies, I hope they're clear enough (ignore the voltages I give though, they're not correct).

So basically I made sure that the audio grounds are wired almost directly to the ground of capacitor C11, the darlington arrays are wired to the ground of C8, and so on and so forth. I was hoping to have different ground spaces this way (I'm voluntarily not using the word 'plane' here), but it doesn't seem to suffice. I get some low noise when the buffers are inactive, and a lot more noise when they are active. BTW I have 2 buffers, but the noise level doesn't change much between when only one buffer is active and when both are.

I'd like to know what I can do to resolve this issue. Add new capacitors with different values? Rework my grounding? I believe I could make 2 separate grounds (one for the digital path + relays and one for the audio/buffers) but I'm not sure how I could achieve that exactly.

Any help appreciated, I'm nearing completion but my project is worthless if I can't resolve this.

[lylehaze]

OK, so there's some opportunities here. The schematic really helps, but the construction method also plays a part.

I have no doubt that the power supply is carrying the noise across for you. And because you're using relays, you can completely eliminate that problem. Here's how:

A relay, unlike a semiconductor, allows COMPLETE isolation between the control (coil) and the switches. There is no need to have a common ground between them.

Get a separate power supply for your op-amps. DO NOT connect the digital power supply to the analog supply. not even the grounds.

Let the digital supply drive the core, modules, and relay coils.

Let the separate op-amp supply drive the signal stuff ONLY.. and shouldn't you be using a bi-polar supply for the op-amps?

Once the supplies are separated, the only path left would be sloppy construction methods.

Have Fun,

LyleHaze

[WickedBlade]

Thanks for your input.

Unfortunately, for practical reasons I can't really get 2 (external) power supplies, which was the obvious solution (I've thought about it but had to reject it because it is impractical for me).

The sloppy construction method is definitely not out of question, unfortunately, but I've checked my work a lot and can't find any fault that would explain the noise.

As for the buffer and opamps, I don't need bipolar supply because I'm using a reference voltage to 'lift up' the signal and get it back to 'normal' at the output thanks to a capacitor. The schematics is basically the one found there: AMZ Super Buffer. What's different is that I'm using only one opamp, let's say IC2b, and resistors R1-R3 are taken out too. I've actually built 2 reference voltage, one for each half of the opamp, which is not actually necessarily, I think, but shouldn't hurt.

About the schematics, I should point out that I've tried to represent the actual way that the wiring is done, and most importantly, the power really splits at the C1 cap, and its pins are the only place were the 3 power supply subcircuits meet, so I'm thinking that noise from the digital path should actually go through C4 then C3, IC1, and C2 to then go back the other way through C9, IC3, C12, and C11. It looks like very bad luck to me that it manages to do just that...

I'll have another look at my ground wires though...

Edited March 7, 2010 by WickedBlade

[sparx]

Yes the power supply looks like where it might be coming from. The bottom 7805, you may want to use diodes instead of resistors to jack up the voltage, or even a different regulator - 7809?

I would definately try using seperate power supplies at least until you can isolate exactly where the problem is.

S

[lylehaze]

Ouch!

Nice catch Sparx.

Using a resistor to offset a 7805 to another voltage is a bad idea..

as the load on the regulator changes, the current through the reference pin will too.

So now your "reference" is shifting with the load, this will create a unbalanced load any way you look at it.

This would get even worse if you're trying to reference the amp stages at half-voltage or something (which you are also doing)

Change that to a "correct" regulator of the desired voltage.

Wow, I missed that completely. I must be getting old or something. :)

[WickedBlade]

Ouch!

Nice catch Sparx.

Using a resistor to offset a 7805 to another voltage is a bad idea..

as the load on the regulator changes, the current through the reference pin will too.

So now your "reference" is shifting with the load, this will create a unbalanced load any way you look at it.

This would get even worse if you're trying to reference the amp stages at half-voltage or something (which you are also doing)

Change that to a "correct" regulator of the desired voltage.

Wow, I missed that completely. I must be getting old or something. :)

Really? I had no idea that the resistors might be a problem! I got the circuit from a 'reknown' source, someone who seems to be well regarded on the net. Heh, go figure...

Anyway, I'm using resistors because I'm trying to get as much voltage (hence headroom) for the opamp as possible. I'm currently getting about 10.15V (with different resistor values), and I'd like to stay in this area (considering that I'm feeding a 12V DC and the voltage drop from vregs, I can't get much more). The only other spare vreg that I have is a 7809 which would reduce the voltage a bit... You mentioned that I could use diodes instead of resistors. Would that fix the issue? I'm interested in any pointers you'd have as to how to do that (I'm not an electronics expert, as you have guessed).

Now, I know that I could feed the opamp directly from the diode bridge and I would get about 12V (minus the diode voltage drop), but I was trying to be future proof and deal with higher power supplies... It's still a possibility, though...

[sparx]

The usual way of increasing the output of a voltage regulator by a couple of volts is to put diodes (e.g. 1N4001) between the Ground pin and the Ground connection. A couple of diodes will increase it by 1.4V (ish).

Am not saying that what you have done is incorrect, but its more common to use diodes.

If you need an unusual voltage or any more than a couple of diodes worth, use a variable regulator such as LM317.

As I'm old and wise, (apparently!), I would suggest you revise your power supply.

Please post back if you need any more help.

S

[WickedBlade]

All right, so it looks like I'm going to use a 7809 with one or two diodes to get ~10V, and I'll rework my grounding and we'll see what comes out of that.

Thanks guys.

[WickedBlade]

I'm trying to find references about the diodes trick for the power supply. Mainly I want to make sure I'm not putting the diodes backwards.

I have only been able to find this: Three different ways to change the output voltage of a 78 series regulator

Can you confirm that this is the correct way of inserting the diodes, with the anode towards the 'real' ground? Also, is the cap necessary and if so, is the value suggested correct?

Of course, if you have more precise websites on the subject, I'm interested...

Thanks!

[sparx]

Yes that is correct, the three diodes in series pointing down is the correct way to do it.

(Am not saying the others are wrong, just this way is more usual).

I wouldn't bother with the cap, but if I did, the value seems fine.

Good luck and post back if it still persists.

A little question though, why don't you use a dual supply circuit (+/- 12V) and tap +5V off this such as is used in the MBFM?

S

[WickedBlade]

Hmm, I haven't had a real look at the MBFM, to be frank :rolleyes:

Part of why is because I'm not really "creating" the audio, just sort of distributing it, so it's not the same use.

As for power supplies, well I'm building a pedalboard so I was trying to avoid clustering the case or the floor with more transformers. As it stands now, I don't have any space left in my case to add another power supply input or a built-in transformer... This would need some serious reworking and I'm hoping to avoid that or the project will be canceled due to too much trouble... Or I'll just dump the buffers and only use the relays, this way the audio and digital will be totally separate, but I liked this feature and will try to keep it viable for as long as possible.

Thank you for your help.

[lylehaze]

OK, I'll take a short shot at explaining this.

The center leg of the regulator is a ground reference. The output voltage is measured from this point, so it should be a stable ground reference. These linear regulators draw as much current as they deliver, and "burn off" the excess voltage as heat.. this is also done through the center pin.

So if you have a (example) 7805 making + 5 volts at a half-amp, from a 9 volt input supply, you'll be taking in 9 volts at close to 1/2 amp at pin 1, and you'll be supplying 5 volts at about 1/2 amp at pin 3. The difference, 4 volts at a half amp, will be burned off as 2 watts of heat, and this current will pass through the center leg.. WHICH IS ALSO YOUR GROUND VOLTAGE REFERENCE..

Now, back to the big picture.. the current draw through this regulator will change with the load. As the current through the regulator changes, so will the current through your resistor. As that happens, the "voltage reference" center pin of your regulator will no longer be a steady voltage. In fact, it will move in voltage giving an accurate amp gauge of your circuit load. As it moves, the output voltage of your regulator will move too, since it's trying to stay 5 volts above that reference.

So you have created a variable voltage that changes reliably with current draw. This is great if you need an ammeter, but really bad if you want a regulated voltage.

Diodes have a (fairly steady) forward voltage drop, so this would be better. Best choice by far would be to use the right regulator for the job. They DO make 7809 regulators, although you would need to supply about 12 volts input for that to work reliably.

The next choice would be to use a pair of 7805's to get ten volts. Use the output of the first one as the center pin of the second.. it's sloppy, but better than the resistors you have going now.

I hope some of this made sense.. it's early, I just woke up.

Have Fun,

LyleHaze

[WickedBlade]

It makes sense, and it's more or less what I had understood. I'm just surprised that the input current is varying so much (since the 1kHz noise comes from another circuit entirely, with only the ground in common).

As I said, I do have a 7809, and I'm using an external 12V DC supply, so I'll manage to get at least 9V. I'd like to do better though, so I'll try the diodes trick to get 9.7 or 10.4V if possible. If worse comes to worse, I'll just use the 12V supply directly, since I took the pains to buy a linear power supply, no switching involved.

Thanks again.