Signal Leaking Into Power Rails? EMR? - Last one, I swear.

[plumsolly]

Something has gone haywire with the attachment feature and I can't even view the other two versions of this thread - moderators, if you could delete those, that would be great.

I am finally getting around to putting a spring reverb I built a while back into a box and I am running into an issue: The signal (from both channels) is coupling into the power supply and some dry signal is showing up on the outputs.

Schematic Here: https://drive.google.com/file/d/0B88JRHsJhP2CZE1QRE9DZmVOZnM/view?usp=sharing

I am measuring about 40mV p-p of signal on the +15V rail.

I am pretty sure the coupling is happening at the constant current driver (IC4 and the two transistors after it). I can see a huge amount of radiation in that area with my scope probe.

Everything is built on perf board.

My question: Is this strictly a layout thing, or is there something I could be doing better, electrically?

Thanks,

Ben

 

[ruffrecords]

It may well be the power supply. Can you post a schematic of it?

Cheers

Ian

 

[plumsolly]

Hey Ian,

Thanks for responding. The attachment feature doesn't seem to be working - that link is the best I can do for now.

The PS is housed externally because of the sensitivity of the coils - it is a linear job and much larger than it should be for this little box.

Best,

Ben

 

[ruffrecords]

As the power supply is remote then you really should have some bulk decoupling where the power enters the unit. It does not take much change in current draw over a few feet of wire to create the 50mV you are seeing. I would suggest you slap 1000uF across each of the power rails.

Cheers

Ian

 

[abbey road d enfer]

+1 to everything Ian said.
Also check the point where R25 is grounded. It should coincide with the junction of the two suggested 1000uF decoupling caps.

 

[PRR]

8000+ pixels seems extravagant for a 7-chip FX box. It could be 20% the size. I dunno if it is the display size which keeps blowing-up the thread.

 

[plumsolly]

8000+ pixels seems extravagant for a 7-chip FX box.

Only the best, PRR. Only the best. But, yeah - I'm sure that's what broke it.

Thanks for the suggestions, Ian and Abbey. I did not mention (and it isn't shown in the schematic) but I have 220µF across both rails on each "card". I went ahead and tried another 470µF (I don't have 1000µF on hand)  at various points - where power enters the box, closer to R25, and it did not help. I also tried running it from a bench supply with much shorter cable length and that did not help either.

To get an idea of the scale of the coupling, the unwanted dry signal that is bleeding into the signal path is only 15dB quieter than the wet signal.

I tried unplugging the reverb return (which has no effect on the bleed) and removing IC5 (that's the recovery amp with a gain of 101) and the dry signal dropped by another 30dB, but it was still there, which leads me to believe that the unwanted signal is coupling before and after the input of IC5.

As I mentioned before I am measuring quite a bit of radiated signal near the constant current amp - 150mV P-P at an unconnected row of holes in the perfboard for a 3V P-P signal. Is this a normal amount of radiation from a circuit like this?

If I disconnect the signal after R31, there is nothing on the output, which leads me to believe maybe it not coupling through the power rails because, if it were, there would still be some bleed through IC1C and the output driver.

I am open to any suggestions.

Thanks again,

Ben

 

[abbey road d enfer]

I tried unplugging the reverb return (which has no effect on the bleed) and removing IC5 (that's the recovery amp with a gain of 101) and the dry signal dropped by another 30dB, but it was still there, which leads me to believe that the unwanted signal is coupling before and after the input of IC5.

Did you short the input of the recovery amp when doing that?

As I mentioned before I am measuring quite a bit of radiated signal near the constant current amp - 150mV P-P at an unconnected row of holes in the perfboard for a 3V P-P signal. Is this a normal amount of radiation from a circuit like this?

How did you measure that? It seems to me you're measuring capacitive leakage. The actual reading depends very much on the signal frequency and the impedance of what you use to measure. It's not a significant measurement. Can you post a picture/drawing of the perf board?

If I disconnect the signal after R31, there is nothing on the output, which leads me to believe maybe it not coupling through the power rails because, if it were, there would still be some bleed through IC1C and the output driver.

Looking at teh schemo I notice IC1c is not galvanically referenced, can you measure the DC voltage at the output?

 

[plumsolly]

Thanks for responding, Abbey.

Did you short the input of the recovery amp when doing that?

No, but I sure can. I'll get back to you.

How did you measure that? It seems to me you're measuring capacitive leakage. The actual reading depends very much on the signal frequency and the impedance of what you use to measure. It's not a significant measurement. Can you post a picture/drawing of the perf board?

Gotcha. I measured with a scope on X1, signal is ~1kHz.  I can definitely get you some pictures.

Looking at teh schemo I notice IC1c is not galvanically referenced, can you measure the DC voltage at the output?

Yep.  Be back shortly.

 

[plumsolly]

Did you short the input of the recovery amp when doing that?

Shorting the input of IC5 has the same effect as removing it altogether.

Can you post a picture/drawing of the perf board?

See attached.

Looking at teh schemo I notice IC1c is not galvanically referenced, can you measure the DC voltage at the output?

-0.5V

Thanks again for your help,

Ben

 

[abbey road d enfer]

Shorting the input of IC5 has the same effect as removing it altogether.

That seems to indicate  capacitive leakage between the high level stages and the recovery amp. Shielding is necessary. I think the drive and recovery circuits should be separated on two different breadboards for proper evaluation.

See attached.
-0.5V

That suggests there is actually galvanic continuity, so either there is a resistor that is not mentioned on the schemo or capacitor C35 has enough leakage to maintain DC reference. Lift C35 and check output voltage of IC1c.

 

[plumsolly]

That seems to indicate  capacitive leakage between the high level stages and the recovery amp. Shielding is necessary. I think the drive and recovery circuits should be separated on two different breadboards for proper evaluation.

Ok. I will explore this tomorrow.

That suggests there is actually galvanic continuity, so either there is a resistor that is not mentioned on the schemo or capacitor C35 has enough leakage to maintain DC reference. Lift C35 and check output voltage of IC1c.

The circuit is as drawn. When you say "lift", like short over it? Is C35 necessary? I added it to protect the pots and the next stage from any DC at the output of IC5, but maybe I don't need it.

Thanks again,

Ben

 

[abbey road d enfer]

When you say "lift", like short over it?

Lift means disconnect one side.

Is C35 necessary?

Not strictly.

I added it to protect the pots and the next stage from any DC at the output of IC5, but maybe I don't need it.

I uderstand that, but you should have a resistor to ground at the rightmost side, of a few dozen kiloohms, to discharge any DC residual to ground.

 

[Newmarket]

The circuit is as drawn. When you say "lift", like short over it? Is C35 necessary? I added it to protect the pots and the next stage from any DC at the output of IC5, but maybe I don't need it.

If it is 'as drawn' then you definitely have an issue around IC1C as there is no source for dc bias for the non-inverting input.
I suggest you need to resolve this as a priority then re-evaluate, Just a resistor (several 10s of K eg 47K) from the non-inverting input will do it. although you might want to place it or an additional resistor after C35 as advised for similar capacitor discharge characteristic regardless of control settings.

 

[plumsolly]

Thanks for the replies and the advice, Abbey and Newmarket.

The filters are from a Harrison 32C channel as implemented in the "Harrison Ford Filters" from a while back (see attachment). I just tapped into the middle because I wanted the highpass before the tank and the lowpass after it (though in retrospect, I'm not sure I would have done it that way...). The highpass and lowpass stages are DC coupled, which is why there was no resistor there, originally. In the original circuit, the DC path from the non-inverting input looks like goes through those 390Ω resistors and the two pots and then the two 10kΩ resistors from the previous stage.

Do you think 47K to ground right after my C35 should get me sorted?

Thanks again,

Ben

 

[abbey road d enfer]

Do you think 47K to ground right after my C35 should get me sorted?

Sure, no prob. I think ATM your circuit works because the very low input bias current of the TL074 takes ages to discharge the 470uF cap, but eventually, it will hit one of the rails.

 

[plumsolly]

Sure, no prob. I think ATM your circuit works because the very low input bias current of the TL074 takes ages to discharge the 470uF cap, but eventually, it will hit one of the rails.

Thank you for helping me understand this - I added it.

That seems to indicate  capacitive leakage between the high level stages and the recovery amp. Shielding is necessary. I think the drive and recovery circuits should be separated on two different breadboards for proper evaluation.

I went ahead and breadboarded the recovery stage and moved it away and it solved the coupling issue. I'd love to leave it on the one board and do something clever to isolate the stages, but I don't have much room left to maneuver. If I need to, I can rebuild the recovery stage on a separate piece of perf board. Any suggestions for things to try before I go that route?

Thank you again,

Ben

 

[abbey road d enfer]

I went ahead and breadboarded the recovery stage and moved it away and it solved the coupling issue. I'd love to leave it on the one board and do something clever to isolate the stages, but I don't have much room left to maneuver. If I need to, I can rebuild the recovery stage on a separate piece of perf board. Any suggestions for things to try before I go that route?

You have to learn about electrostatic shielding. Make sure sensitive lines are not too close to high-level ones and insert grounded shields.

 

[derekbenjaminmusic]

@plumsolly Did you ever get this unit completed? I've been trying to find basically the exact same thing but I'm not experienced enough to do the R&D that's needed for this project. I would love to find a working model to build!