SansAmp GT2 whistling problem

[Rob Strand]

All grounds are common. I used the same jack connection structure several times without any problem. Here is the schematic drawing of my build. Layout of this drawing: https://gerber-viewer.easyeda.com/showcase/?#!id=4e556e22371311e899d9026a86b9cae7&type=top&layer_list=1-2-3-4-5

I didn't try to pick holes in your layout by tracing through tracks but at first glance it looks pretty good.  There's quite a bit of ground plane and it *is* connected to ground.

have an idea, even though it may not be sensible. As I put TS808 in front of the SansAmp, whistling is gone. In order to change the input impedance, can it be helpful to change the input resistor of the SansAmp circuit?

You could connect IC1a pin 3 to IC1b pin7 using a thin wire.  That effectively buffers one part of the circuit.
Maybe that will do the same thing as preceding it with a TS808.  You need to I IC1a away from IC1b:  I would cut the track between IC1a pin 3 and IC1b pin5 at *both* ends to keep the tracks to IC1b pin5 to a minimum.

[BlackTooth61]

You could connect IC1a pin 3 to IC1b pin7 using a thin wire.  That effectively buffers one part of the circuit.
Maybe that will do the same thing as preceding it with a TS808.  You need to I IC1a away from IC1b:  I would cut the track between IC1a pin 3 and IC1b pin5 at *both* ends to keep the tracks to IC1b pin5 to a minimum.

Is this what you mean?

[Rob Strand]

Is this what you mean?

Yes, that's it.  It's worth a try.

[rankot]

If PCB is your design, then proximity of input and output traces could be the problem. I had similar experience with my RAT, fixed it with voltage divider at output. Overall volume is lower, but not whining at all.

[BlackTooth61]

If PCB is your design, then proximity of input and output traces could be the problem. I had similar experience with my RAT, fixed it with voltage divider at output. Overall volume is lower, but not whining at all.

Now I'm pretty sure that the source of the problem is PCB design. I drew the PCB inspired by user nomorebetts' layout. Jacks, switches, pots and footswitch are placed similar positions. Everything else are different. And nomorebetts' layout has exactly the same problem. They argued about the problem in this topic: http://www.diystompboxes.com/smfforum/index.php?topic=103894.msg985218#msg985218

DIMstompboxes proposed some fixes for each pot. Actually I want ONE fix to rull them all. But I will try all of them.

Thanks for the suggestion, but your suggestion requires external operations. I'm sure it will work but I'm saving it for the last.

[MrStab]

when i asked about the TS808 bypass, i assumed it was an original with a buffer. on a forum called "DIY Stompboxes". i'm not the sharpest tool in the shed. anyways, was trying to figure out whether it was low impedance or the ~700Hz LPF that was reducing the whistle.

one thing worth mentioning re. proximity: i've found the best 3PDT wiring scheme for high-gain pedals is one with the grounds on the middle poles. your connections to the jacks could be 10 miles apart, but they're forced close together by the nature of the bypass switch. Not only does putting the I/O signals on the outermost poles put them further apart, but the ground connection in-between should provide some degree of shielding.

[rankot]

I had a squealing problem with this RAT:



So I tried to redesign PCB and separate in/out traces as much as possible, but I still didn't have time to test this new PCB. It is a little bit unusuall design with MOSFET optocouplers to provide popless switching. Switching works fine, but squeals (tested with different op amps and they all work the same), contrary to my first built RAT which works excellent, but pops (switch sucks)! I will report what's happening with new design, but I can't do it in a week or two so you have to be patient

[Rob Strand]

Now I'm pretty sure that the source of the problem is PCB design. I

It's highly likely. 

There's two output tracks one under the 10k input resistor and one under the 22nF input cap.  This is probably the last thing you want to do.  Also the input tracks are quite long they go from the bottom-middle to the right-middle near the DC jack.  If you used the right-hand set of contacts to switch the input signals then the input tracks would be shorter and naturally keep away from the output tracks.

There may be some close tracks on the underside as well.

The 47pF to 100pF cap across the input normally protects against this but maybe the PCB's too far over the edge.

Because both sides of the 10k resistor are exposed there's a chance the opamp side is also picking up feedback from the output.  However the fact the TS808 save it might point more to the input side.

As a first step, I'd do the following to get the input parts away from the output tracks:
- pull the input 22nF cap
- pull the 10k input resistor
- pull the input 1M to ground.

- Cut the track running from the switch to the 1M + 22n.  Cut at the switch end.
- Solder the 22n in series with the 10k resistor as a single part.  Use leads short as possible.
- On the back of the PCB, solder the 22n+10k directly from the switch to the old hole where the 10k went.
- On the back of the board solder the 1M from the switch to ground.
- Solder a 47pF to 100pF cap across the 1M.

It might be a bit tight and a bit messy.

At all times lay the parts so they are as far away from any track or switch contacts that connect to the output.  This might mean standing the parts further off the board to keep them away.

It might not be enough but this is the type of thing you have to do to save the day and push the layout back into the good zone.

[BlackTooth61]

I've tried to lower the R40(330k) on IC4a in order to limit the output level. Most of the problem is solved. There is still a bit of whistling when I crank up the HIGH pot. I'm planning to change the R9(3.3k) into a higher value in order to limit HIGH pot without affecting the tone. If necessary, I will put a cap(smallest value possible) across the pins 1 and 2 of IC1 and maybe pins 6 and 7 of IC4 as Rob said.

[idy]

yes, changing the input impedence, by adding buffering, may make the problem go away. Will make?

[idy]

You can go to "build reports" at tonepad. Several builders experinced same or similar. One says he replaced opamps with NE5532Ps. And got rid of problem. That or a buffer. Or read the rest of the build reports...

[Rob Strand]

I've tried to lower the R40(330k) on IC4a in order to limit the output level. Most of the problem is solved. There is still a bit of whistling when I crank up the HIGH pot. I'm planning to change the R9(3.3k) into a higher value in order to limit HIGH pot without affecting the tone. If necessary, I will put a cap(smallest value possible) across the pins 1 and 2 of IC1 and maybe pins 6 and 7 of IC4 as Rob said.

Before you mod the circuit it's probably worth trying any of these things:
- Put a 47pF too 100pF cap from IC1a pin 3  + IC1b pin 5 to ground.
- Move IC1a pin 3 to pin7 IC1b.
There might be something weird going with the two opamp inputs connected together. So moving the
  connection or  the 47pF mod might stop whatever is going on there.  Another possibility worth
  trying is to reduce the 10k to say 2.2k.
- Mount the 10k input resistor and the 22n input resistor off the board (ie. long legs) about 6 to 10mm
  to reduce coupling to the output tracks.

BTW, the cap values I gave previously are quite minimal.   100pF on the input is probably the only one pushing the boundary.

[BlackTooth61]

Off-topic but I've drawn a 3PDT layout for future pedals. I drew the input-output tracks opposite sides of 3PDT and I tried not to cross in-out tracks on two layers. Middle pins belong to LED.

Is this a well drawn layout for input-output issue in your opinion? Is there anything I can do for improving the layout?

[BlackTooth61]

Before you mod the circuit it's probably worth trying any of these things:
- Put a 47pF too 100pF cap from IC1a pin 3  + IC1b pin 5 to ground.
- Move IC1a pin 3 to pin7 IC1b.
There might be something weird going with the two opamp inputs connected together. So moving the
  connection or  the 47pF mod might stop whatever is going on there.  Another possibility worth
  trying is to reduce the 10k to say 2.2k.
- Mount the 10k input resistor and the 22n input resistor off the board (ie. long legs) about 6 to 10mm
  to reduce coupling to the output tracks.

BTW, the cap values I gave previously are quite minimal.   100pF on the input is probably the only one pushing the boundary.

I've moved IC1a pin 3 to IC1b pin7. I also mounted the 10k and the 22n and 1M off the board as you said before, cut the switch end of the track. But nothing has changed.

[Rob Strand]

I've moved IC1a pin 3 to IC1b pin7. I also mounted the 10k and the 22n and 1M off the board as you said before, cut the switch end of the track. But nothing has changed.

The fact those mods did *nothing* is a concern.   There's clearly a problem but not much evidence of a cause.

At this point I'm thinking the switch is the problem.   Probable the two contacts side by side.   If that's the case, separating the switch contacts as far as possible with the LED contacts in the centre is probably the only solution left.  That's of course helps separate the tracks as well.

As a strong sanity check to confirm the output wiring is feedback to the input, either by tracks or by the switch,  it would be very worth while to manually wire an output jack.   With this mod the switch won't work but it has to be set to the effect position to connect the input.   The test would be:

- Remove output components:  C30 (2u2) and probably R6 (1k)
- Wire some coax to the emitter of Q1 and to ground
- hang that wire out of the box
- Air-mount new parts C30 (2u2), R37 (100k) and R6 (1k) to a  new socket
- connect the other end of the the coax to C30 and the socket ground.

This should remove all output wiring from around the switch.   

If it were mine I'd probably do the same to the input (ie pull-up R41 (1M), C3(22n) and R11 (10K) and wire it to a new set of components and input socket).   However you can try the output first.

If the oscillation goes away then you can be fairly certain the oscillation is caused by some sort of coupling around the jacks and the foot switch.    If the oscillation is still present then you know there's another problem on the board.

Is this a well drawn layout for input-output issue in your opinion? Is there anything I can do for improving the layout?

Well the main issue I see is you have the input and output on pins next to each other at the 4-pin socket.  Given you suspect there's an input/output coupling issue I'd be putting the input and output on opposite sides of the 4-pin connector.   I'd also try to keep all the output wiring to the left of pin 4 and pin 3 of the output jack, and all input wiring to the right of pin 4 and pin 3 of the input jack.   Dump as much ground plane on both sides of the board between the left (output) and right (input) sides of the board.

[BlackTooth61]

I think I have no energy to solve the problem. So I've changed the R40(330k) into 56k and R9(3.3k) into 33k. Whistling is completely gone with these values. But in California setting, HIGH pot range got a bit narrow. Whistling is not a big issue in California and Tweed settings. So I will put a resistor (controlled by empty pins of the AMP switch) parallel to the R9 so that change the R9 back to its original value.

Well the main issue I see is you have the input and output on pins next to each other at the 4-pin socket.  Given you suspect there's an input/output coupling issue I'd be putting the input and output on opposite sides of the 4-pin connector.

Yes, I realised that could be a problem after sending the photo and changed as you said.

Dump as much ground plane on both sides of the board between the left (output) and right (input) sides of the board.

I'm not sure I understand this sentence. Can you clarify?

[Rob Strand]

I think I have no energy to solve the problem. So I've changed the R40(330k) into 56k and R9(3.3k) into 33k. Whistling is completely gone with these values. But in California setting, HIGH pot range got a bit narrow. 

These tricky problems can grind you down a bit.  Just be aware sometimes when you "fix" these issues you can just be pushing the oscillation above the audible range.  This is where a CRO is extremely useful.

One thing I've learnt about these types of problems is finding the cause is more important than finding the fix.   The correct fix is easy once you know the cause and you know it will work.  However there's many fixes which only hide or reduce the problem (sometimes coming back on other builds).

I'm not sure I understand this sentence. Can you clarify?

All I'm saying is to put as much ground plane down on both sides of the board as possible.
However it is important to put as much ground tracks or fills in between any input and output signals.
The ground promotes the signals to couple to ground and not between themselves, which is the suspected cause of the problem.

The grounding on your original PCB was pretty good.

[BlackTooth61]

Here is the final version of the layout with both sides of the PCB included your suggestions.

[Govmnt_Lacky]

That GT2 PCB design looks oddly familiar... 

[Rob Strand]

Here is the final version of the layout with both sides of the PCB included your suggestions.

It looks good.   Unfortunately it's late here and I'm a bit sleepy so I might not be too sharp picking up bugs.

Now for the possible bad news.  When I looked at that layout I can see all the tracks and it is very clear what is going where - fair enough .

The other thing I do see clearly now is how the true-bypass is done.   Notice the input jack is always connected to the output set of contacts (contact 7).   The output jack goes to the centre pole of the switch.    Now, when you are in effects mode the only separation between the input and output signals is across contacts 4 and 7 of the switch.   This *might* be the real problem.    On one hand putting the 47pF to 100pF cap across the 1M input resistor *didn't* fix the problem.   On the other hand we know putting a TS808 on the input fixes the problem.   So the coupling between the input and output is either very strong or we are on the wrong track entirely.

So what to do?

Assuming the coupling across the switch is causing the problem.    We have one option.   Go here and scroll down just before half way down.  There are three true bypass circuits.   

http://www.geofex.com/Article_Folders/bypass/bypass.htm

The first one is your connection.   The second one is kind of the reverse of the first.  Here the output is permanently connected to the input set of contacts and can couple across the switch similar to the first case.   However, look at the third circuit.  In this case, when the circuit is in effects mode, the coupling from the input line to the output line has to jump two sets of contacts.   This will halve the coupling capacitance between input and output..

Is this going to fix the problem?  I don't know! If it doesn't then your only options are
- to give-up on true bypass and use a buffered bypass. 
- make the switching more complicated and use the third circuit but also short the top wire to ground in bypass mode.  This will need more contacts.
- use the third case but connect a small capacitance from the top wire to ground.   This makes it not true bypass but if the capacitance is small enough it will be less than the variations in cable capacitance.

What I've said here might not be necessary. Your new layout might just work.   But if it doesn't then the problem isn't solved.  I don't know if you can hack your current PCB to implement the bypass in the third circuit.  That would be a really good test.  It's still not 100% because your current PCB has other coupling issues due to the tracks.  (That's why I wanted to bypass the whole switching circuit a couple of posts back).  However if it did work that would be a big step forward.   If it doesn't fix the problem then maybe go ahead with the new layout.

At the end of the day I still don't know why this circuit is so hard to stabilize.   Obviously others have had similar problems.

I'll leave you to think about how you want to go about it.