White Noise Problem FV-1 Dry Path

[alsirazi]

Hi all,
I have a problem with one of my builds, if someone could help me, I will be very happy.

I am trying to build an FV-1 DSP chip based pedal and I couldn't solve the background white noise problem of this circuit. Noise is not coming from the DSP, but from the analog dry path. I tried several configurations; with all smd components, all through hole components or hybrid. For smd components I used x7r caps in one pcb and also C0G caps in another pcb (1nF and pF caps). For 1uF coupling caps, I used film type capacitors (Wima and other brands also). (Yeah, I tried several different PCBs with no success )
The circuit is based on RG's Panning for Fun circuit (http://www.geofex.com/Article_Folders/panner.pdf) and used in many FV-1 designs. It creates a dry path and a wet path for DSP, then mix them with a panner design.
I tried different op amps (TL072, NE5532, OPA2134), same noise for all.

The problem is when I plug it in an amplifier clean channel (with room levels) it is hard to hear any noise, but when volume is up, an audible background hissing noise arise. It is even worse with gain channels of amplifiers.

Is there something that I am missing with this design? Here is the schematic:


Thanks for your help.

[ElectricDruid]

The only thing that strikes me is the large value resistors used for generating the bias supply. Is 1M really necessary? Could you use 10K instead? The large resistors will generate more noise, and then you're feeding that into the rest of the circuit. That said, 1M/1u is pretty heavily filtered, so I dunno.

Nothing else stands out. It's pretty standard stuff and I can't see any reason why it should hiss.

HTH,
Tom

[Rob Strand]

Noise is not coming from the DSP, but from the analog dry path.

Maybe not from the DSP a/d and d/a but it could be coupling through other paths eg digital crap on the ground.
If you can lift R3 and R4 and remove the power from the chip easily, try that and see if it goes away.
(Be careful that *all* power and signals are removed, you don't want to fry the chip.)

[alsirazi]

The only thing that strikes me is the large value resistors used for generating the bias supply. Is 1M really necessary? Could you use 10K instead? The large resistors will generate more noise, and then you're feeding that into the rest of the circuit. That said, 1M/1u is pretty heavily filtered, so I dunno.

Nothing else stands out. It's pretty standard stuff and I can't see any reason why it should hiss.

HTH,
Tom

Thank you for your reply Tom, I never tried this before, will give a shot for lower resistance values, thanks.

[alsirazi]

Maybe not from the DSP a/d and d/a but it could be coupling through other paths eg digital crap on the ground.
If you can lift R3 and R4 and remove the power from the chip easily, try that and see if it goes away.
(Be careful that *all* power and signals are removed, you don't want to fry the chip.)

I tried this, actually dsp and other digital components not present in the pcb now (A test pcb for analog path only). Noise is the same, actual pcb has solid ground plane and never had issue with digital noise.
I also tried this circuit with different power supplies in different places and with different guitars and amplifiers, result is the same.
Thanks.

[Ice-9]

It could be a good idea to post a picture of the build as this may shed some light on the problem. There is no reason that the blend schematic should be that to noisy. It could help a little to change the PSU Bias op amp section using two 10k voltage divider resistors and a 47uf cap.

[Rob Strand]

I tried different op amps (TL072, NE5532, OPA2134), same noise for all.

I have an idea where it could come from now. 
- The signal is divided down by the mixer (as there is no gain from IC1A).
   So the signal to noise at IC1B is low.
- Your bias supply comes from an opamp which has noise from IC2B.  This adds
   to the noise of IC1B, effectively making it 1.4 times noisier.
-  From the *+input* of IC1Bs point of view, both these noise sources are multiplied by 3.6.
   But the signal is only at the input signal level.
-  The noise after IC1B is 1.4 * 3.6 = 5.0 times  (14dB)  higher than that due the input opamp.

I would bias IC1B directly from the bias network not from the opamp output.
Adding gain to the front-end will also help provided you don't overload your DSP stuff.
(The way around it then would be to put some pre-emphasis at IC1A then
de-emphasis after IC1B or further up.)

Using a different mixer circuit will also help.

[Danich_ivanov]

I catched white noise myself in one of the builds recently, and the problem was as insignificant as flubby connection to the op-amp. Could be as simple as that.

[PRR]

> large value resistors used for generating the bias supply.

But they have 1uFd across. Amply low Z through most of the audio band. That won't hiss.

None of the other values are suggestive.

Hiss is UNIVERSAL. There's always some. Guitar "Hi Gain" pushes the hiss up (it assumes you will play LOUD and won't stop to hear the hiss).

[Rob Strand]

But they have 1uFd across. Amply low Z through most of the audio band. That won't hiss.

This table shows the minimum bias filter cap size where the bias network will degrade the A-weighted signal to noise by 1dB; assuming a JFET opamp with noise voltage 10nV/rtHz

R's                      Cfilt    (effective -3dB)
1M//1M=500k      3.3n   (96Hz)
100k//100k=50k   10n   (320Hz)
10k//10k = 5k      18n    (1.8kHz)
4k4//4k4=2k2      18n     (4kHz)
2k4//2k4 =1.2k   none                     ; below this degrades less than 1dB

So anything more than 18nF will work with any choice of bias resistors.
Pretty darn small values.   
Obviously doesn't include noise from the PSU.

[PRR]

> noise after IC1B is 1.4 * 3.6 = 5.0 times

Ah, yes.

The mix-pot wiper needs a very low impedance to avoid leakage. So right on the 1meg+1uFd may not do.

The other Vb points can be fed at 1Meg+1uFd. The mix-pot needs buffering and filtering. An added 100r will get sub-sonic sneakage down to -40dB. The 100ufd will make it much lower over the audio band. The 100uFd will be essentially hiss-free over the audio band. We still have the hiss-gain of 3+ in the output stage, so this only gains <3dB better hiss level.

This type cross-fade mixer should be run at higher levels. But that leads to gain of 2 or 3 at the input, then I assume a loss of 2 or 3 to feed the DSP? Then gain 2-3 for DSP recovery. Much more complicated. Gain of 5 or 10 would be better, but for guitar-cord work needs a lot more than 9V supply.

[Rob Strand]

Ah, yes.

Just noticed I made a minor screw-up in that I didn't allow for the connection of the mixer pot wiper in the noise estimate.  Maybe changes the results a bit.

[Digital Larry]

How about AC coupling the mixer pot wiper to ground rather than directly to the bias supply?  Seat of the pants observation, no idea what difference it would make.

[PRR]

> AC coupling the mixer pot wiper to ground

If the cap is not infinitely large, there will be crosstalk in bass.

This IS sometimes done. A little bass leakage hurts nobody.

It would maybe be better to cap-couple the whole fader network so its DC can be zero. This means more caps, and adding-back the DC bias. But they may be smaller caps. This is almost certainly the way "pro" stuff would do it. DIY pedals may be tight on space, or design insight, so may use other ploys.

[amz-fx]

Down-size the resistors for a little less noise too... output gain of IC1b is essentially unity. Flat across the guitar range.



regards, Jack

[Rob Strand]

IMHO this form of single-gang blend pot is better for noise.  It's got a gain of 1 and doesn't need a recovery stage which is where noise gets in (it gets in by amplifying the noise of the opamp IC1B, regardless of the resistors.)
I don't mean use the transistors here only the pot connection.