1

I need to switch on and off an audio signal without clicking or thumping. I have tried the below circuit that I designed and simulated in LTspice, but it generates a thumping sound when I wire it up and test it. Also from the simulations in LTspice it can be seen that there seems to be an DC offset when I insert the probe within the swithcing section (between C8 and C9). Also on switching off there is a strange exponentioanl curve seen. I suspect that this is generating the THUMP. The signal from the generator is positive and it never goes to negative values.

I would like to generate a signal like this: S(t) = A * sin(t) The signal should be smooth turning on and off if A is an exponential function varying between 0-1 and 1-0.

A is implemented in the circuit by the two MOSFETs shunting.

Any suggestions on how to avoid the click / thumping? (I have tried to increase the RC constant R13, C2. It doesnt solve the problem. The reason for the push pull driver is to be able to drive current in and out with the same resistance so the rise and fall times will be similar)

Circuit. Green line shows audio signal path

DC offset in the start. exponentially shaped

MarkU
  • 14,413
  • 1
  • 34
  • 53
Morten
  • 11
  • 1
  • 4
    Classic issue with single supply, DC bias and an output capacitor (C9) to remove the DC! Can you measure the voltage across C9? Can you try to mute and unmute during zero crossing? – winny May 27 '20 at 12:39
  • 1
    Try activating and deactivating when the signal passes through 0 volts. I'm not saying that's easy to do but you can do this in your sim and see the result very easily. – Andy aka May 27 '20 at 12:44

2 Answers2

1

10nF capacitor C8 charges and discharges substantially from the audio when the mosfets are ON due the low time constant. The voltage in the capacitor is unpredictable. When you turn the mosfets OFF C9 is charged from C8. When you turn again mosfets ON C9 discharges and creates the thump.

Try to use mosfets so that turning them ON or OFF doesn't change capacitor charges. For ex. remove the capacitors from the signal path totally or at least remove C9 and have much bigger C8.

Another approach is to have a fet in series with the signal, not shorting it to the ground. Driving the fet becomes more complex but it's possible. I have built a multichannel noise gate with piezo triggers for better separation in drum recordings with that principle. In that environment the extra sounds were treated differently (=nice new click in the bass drum! make it louder!)

A light dependent resistor can be a better and more easily manageable choice if you try to increase the series resistance instead of shorting the signal. You can assemble a led and LDR in a common tube or purchase LDR output optocouplers.

  • Good point, but may not be the biggest source of thump. Easily fixed with 100K across M7(source-drain) to discharge C8. –  May 27 '20 at 15:08
1

Don't forget that you are connecting a 5V step onto your audio signal via a fairly hefty capacitor (the gate-drain capacitance of the switching FET).

You have removed the HF part of that with R13/C2 leaving only the LF thump as you observed; you probably need to get rid of that too.

In one product (using JFET switches) I kept the thump subsonic, with a time constant of several seconds : the normal AC coupling eliminated what was left. It sounded like a valve amplifier warming up several seconds after power-on!

However, nobody ever complained about that and the product had some pretty good reviews (except when the reviewer was working for the main competition!)

Another more complex approach (but possibly viable given you employ 2 FETs) would be to make the second FET a PMOS with a similar Cg-d, driven from the complement of the first gate drive waveform, thus cancelling out the thump with a complementary waveform. If Cgd is different, you can compensate to some extent by adjusting the gain of the second gate drive.

Or, given a differential signal or fully balanced output, you could apply this switching to both true and inverted legs, where the thumps would naturally cancel out in a subsequent differential amplifier, or balanced line input.