J-fet mute DC offset movement problem

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Tompson17

Active member
Joined
Nov 28, 2017
Messages
29
Hi,

I made some analog crossover and I need to choose only one output from all 8 toutputs I have. So I designed well known circuit with one j112 in series with inverting op amp and one shunt j112 to the ground. The controll circuit is based on Soundcraft GB-8.

When I forget on the op amp resistor values (can be higher), the DC offset on the end of previous stage, separate ground for reference (the difference is constant and really small), distortion (which is not as low as I want, but still very low), I ges some magic on this. The relay (left in the schematic) is driven from separate 12V source, which is conected to analog +-15V in one point (ground to ground).

When the stage is muted, I get normal DC offset on the output of the mute stage (0,5mV) , but ... when I get UNMUTE the stage, the output offset is not something like DC one stage + DC (inverted) second stage. The result is about 2 - 6mV and it moves (!!!) when I MOVE the switch with 8 + 1 wires, which are driving the relays...

I dont understand it. I understand noise from AC line, noise from bad layout... but DC i don't understant.

Note that the control voltages are ok, for the series fet it is about 15mV when it is open, -14,8X when it is off. For the shunt fet is is similar, signal goes thru it well when the circuit is ON, nothing get thru it when OFF, the traces are realy short, all smd, j-fet op amp...

Function OK, but I want to get from all pops and clics. I know i need to block the stages and so... but I dont realy understand the fu**** DC movement on the output with relationship to moving the relay cables... Help me please  :-[

 

Attachments

  • mute stage.PNG
    mute stage.PNG
    15.2 KB · Views: 74
Tompson17 said:
When I forget on the op amp resistor values (can be higher), the DC offset on the end of previous stage, separate ground for reference (the difference is constant and really small), distortion (which is not as low as I want, but still very low), I ges some magic on this. The relay (left in the schematic) is driven from separate 12V source, which is conected to analog +-15V in one point (ground to ground).
I se that the shunt FET is connected to another ground (GND) than the output opamp (AGND). How good is the connection between those two "grounds"?
It is possible you have charge injection from the gate of Q16. You may want to decrease the value of R183.



 
abbey road d enfer said:
I se that the shunt FET is connected to another ground (GND) than the output opamp (AGND). How good is the connection between those two "grounds"?
It is possible you have charge injection from the gate of Q16. You may want to decrease the value of R183.

Between grounds is something like 70uV DC offset. Connection seems to be relatively good, the GND is ground plane and the AGND is "low current reference" ground. When i switch between channels, the DC is still the same, without any transient (RC), there is only visible pop about hundrets of uV, which is very thin and seems to be similar to the output stage, but the difference is the constant DC level between grounds and "floating" DC offset on the output.

Is the two ground design pointless? The walue was on some schematics, I haven't any experience with fets...

Many thanks for your answers :)
 
Tompson17 said:
The schematic is attachet as png.
OK, when Q16 is high Z output of IC19A will be 1x the input DC offset V.

When Q16 is low Z  output of IC19A will be 2x DC input offset V, + the DC offset voltage of IC 19B, + DC voltage at input of IC19B (not shown, going off the page).

So while perhaps small there will be a finite DC voltage change between on/off.
=====
While switching states. gate capacitance from both JFETs can couple into the - input of op amp... In an ideal world there will be some cancellation , in the real world where we live these charges will not cancel completely, if much at all.

Finally you want to sequence the gate drives such that Q16 is high Z before Q15 goes low Z or that will create a high DC gain situation at IC19A. From inspection it looks like the gate logic is reasonably sequenced. And small Cs should slow the edge rates of logic signals to reduce cap coupling.

FWIW the gate voltages should be 0V when turned on, if positive, it suggests a leaky diode, or measurement issue (like ground reference is not really 0V).

Generally U19A would have a feedback capacitor to somewhat mitigate gate capacitance.

You may need a DC blocking cap in series with R179 and - input of U19A. Note high Z resistors would also need to be used to bias the floating end of these caps to 0v, for proper operation and less clicks.

Good luck...

Note: if you mute while music is playing there may be clicks caused by muting while audio signal is not at 0V, just like a mechanical switch can click.

JR
 
Problem solved. It was a 16MHz rezonace around the GND and AGND. Maybe parasitic capacitance of output THAT1646 and some on... The parasitic resonance changed its waveform with mowing the switch and this caused a shift in average value - DC. Now all devices are grounded to GND

In this situacion, I have a -100dBu noise level with 50Hz mess, this is fine, because of open PCB, without any shielding, which I want to do. Maybee it will be -110dB in future

But still the switch function is not as I need. I placed a coupling cap between IC19B and R179, another cap is between source Q15 and GND. Between the IC19A output and source of Q15 is 39kOhm resistor that should balance the DC offset voltage of the output and input. Value of R182 was changed from 4M7 to 3M3 - Q16 get close faster, before the Q15 gets open.

Unmuting works wery fine. With signal I cannot discover any problem, mute worsk smoothly, muting works smoothly too, but...

If I see the muting process without any signal, I find out there are some AC mess. I did not some photo, but in principle, 8 - 9 times burst  of some oscilacion cca 100kHz, with repeating cycle about 1 to 2 ms,  with maximum amplitude about 5mV. The shape is fade in the rezonance, then 2 - 3 periods peak (5mV) and similarly fading down. The signal envelope is not completely regular. The 39kOhm resistor is THT and is above the mute circuit, because all another components are SMD and its relatively small.

Many thaks for Your answers that have me led to solve the main problem  :) :). Any ideas where the problem may be now?
 
Tompson17 said:
Problem solved. It was a 16MHz rezonace around the GND and AGND. Maybe parasitic capacitance of output THAT1646 and some on... The parasitic resonance changed its waveform with mowing the switch and this caused a shift in average value - DC. Now all devices are grounded to GND

In this situacion, I have a -100dBu noise level with 50Hz mess, this is fine, because of open PCB, without any shielding, which I want to do. Maybee it will be -110dB in future

But still the switch function is not as I need. I placed a coupling cap between IC19B and R179, another cap is between source Q15 and GND.
Floating the source from ground means Vgs could be a little indeterminate.
Between the IC19A output and source of Q15 is 39kOhm resistor that should balance the DC offset voltage of the output and input.
This should work in principle and should peg the source voltage close enough to 0V to switch properly.  I might have been tempted to put the cap in series with the drain, but if it works it's all good.
Value of R182 was changed from 4M7 to 3M3 - Q16 get close faster, before the Q15 gets open.

Unmuting works wery fine. With signal I cannot discover any problem, mute worsk smoothly, muting works smoothly too, but...

If I see the muting process without any signal, I find out there are some AC mess. I did not some photo, but in principle, 8 - 9 times burst  of some oscilacion cca 100kHz, with repeating cycle about 1 to 2 ms,  with maximum amplitude about 5mV. The shape is some fade in of rezonance, then 2 - 3 periods peak (5mV) and similaryly fading down. The signal envelope is not completely regular. The 39kOhm resistor is THT and is above the mute circuit, because all another components are SMD and its relatively small.

Many thaks for Your answers that have me led to solve the main problem  :) :). Any ideas where the problem may be now?
A small lead capacitor in parallel with R178 could swamp out stray capacitance from - input node to ground (like I already suggested)..

The cap in series with R179 is now getting it's Dc bias through one or the other JFETs, perhaps a high z resistor between output of U19A and R179 to provide DC bias when both JFETs are open.

JR
 
@JohnRoberts

The last problem was hard to find, but it is funny... and as a many problems, very simple.

The soundcraft circuit is driven by switch, but my crossover by relay. Its function is get the R188 to GND and open the T16. When i mute the circuit, relay gets off by "inner spring". However, the switching contact leave the "close pin" nad touch the "open" pin... but 8 times!!! It seems like a dribble on the open contact because of the spring. And due to the fet control system this transients gets on Q16 which modulates the audio signal.

So i put the caps between R187 R188 and -15V, between base T16 and -15V and 100k res from D51 to gate of Q16. As recomended higher, I lowered the value of R183 to 100k too and the system works wery well yet.

Thanks for all your recommendations guys!
 

Latest posts

Back
Top