This way but the problem is still here
Can you short the 10R resistor with a wire and then trace the output?
In this particular case, the board is not too long, so connecting them to the "back" of the board is fine. On the longer board it would make sense to connect them together and then make a connector for a separate ground wire.
Why would one want to take the interference passing through these decoupling capacitors to the Signal wiring?
It's not just hum that passes through the decoupling capacitors. Any interference on the supply rail will take the decoupling capacitor route, IF the impedance is lower than any alternative.
All the HF glitches that get onto the supply rail will pass through the decoupling and enter the current route you have provided. The low voltage side of that route MUST be low impedance. Any impedance in that route will generate a voltage. Minimise the impedance as much as possible. This means very short wires/traces to the decoupling ground.
Yes, connecting the low voltage side of the decoupling caps together is better.
Do that for every decoupling pair.
Then connect all these decoupling low volts wires together to create a low impedance decoupling ground. This is where you connect the PSU zero volts to.
Here is my "opinion".
I think it is better to cross connect each pair of decoupling caps. A short thick route is better than a long wire.
Then connect all the cross connections together with another wide trace. The shorter the better. But preferable to have the highest current connects at one end of the trace and have the lower current connections spread along this trace to the smallest current at the other end.
This should be done at the PCB layout stage.
Then the current routes for each stage should as closely as possible follow the traces that return the current. This is the PCB equivalent to using a close coupled pairs of wires.
Cherry drew an example of this in a paper way back in the 80s, or 90s.
But almost no layout designer follows his advice.
Almost all spread out the Flow and Return traces so that they have BIG LOOP AREA. This is where your interference gets in.
And a big ground loop around the perimeter of a PCB is one of the biggest interference gathering loops one can build in.
I think it is better to cross connect each pair of decoupling caps. A short thick route is better than a long wire.
Then connect all the cross connections together with another wide trace. The shorter the better. But preferable to have the highest current connects at one end of the trace and have the lower current connections spread along this trace to the smallest current at the other end.
This should be done at the PCB layout stage.
Then the current routes for each stage should as closely as possible follow the traces that return the current. This is the PCB equivalent to using a close coupled pairs of wires.
Cherry drew an example of this in a paper way back in the 80s, or 90s.
But almost no layout designer follows his advice.
Almost all spread out the Flow and Return traces so that they have BIG LOOP AREA. This is where your interference gets in.
And a big ground loop around the perimeter of a PCB is one of the biggest interference gathering loops one can build in.
Last edited:
the ground lift ( a bad label in my view) reduces interference current in the signal wiring. This is not usually a problem in a mono block. (a real monoblocks, not a dual mono where two amps are inside the same chassis).
The ground lift will probably have no effect in a mono test. Compare it on your work bench using 0r0 to 10r to 100r.
When you go to a multi-channel amplifier, then you find that interference current in the signal wiring increases enough to become easily measurable and often audible.
That's when 0r0 does not work and you need to add in the extra resistance, to reduce the current. Somewhere between 2r and 22r is usually suitable.
A better label for this resistor is "signal wire current attenuator".
One disadvantage to using this extra resistance in the signal return to decoupling ground connection is that it generates a DC voltage when the +ve current is not exactly matched to the -ve current. That DC voltage error is read by the amplifier stages and becomes an output offset.
Do not unbalance the input stage to cancel this "added resistor" offset. Instead balance the +ve and -ve currents so that the DC voltage across the added resistor becomes <<1mVdc.
The ground lift will probably have no effect in a mono test. Compare it on your work bench using 0r0 to 10r to 100r.
When you go to a multi-channel amplifier, then you find that interference current in the signal wiring increases enough to become easily measurable and often audible.
That's when 0r0 does not work and you need to add in the extra resistance, to reduce the current. Somewhere between 2r and 22r is usually suitable.
A better label for this resistor is "signal wire current attenuator".
One disadvantage to using this extra resistance in the signal return to decoupling ground connection is that it generates a DC voltage when the +ve current is not exactly matched to the -ve current. That DC voltage error is read by the amplifier stages and becomes an output offset.
Do not unbalance the input stage to cancel this "added resistor" offset. Instead balance the +ve and -ve currents so that the DC voltage across the added resistor becomes <<1mVdc.
Last edited:
Hi Andrew.
In an amplifier with 2K2 input resistor and 47R gain-set resistor is it possible to balance the input and feedback current?
Also, how can the dis-balance be a DC potential when both the 2K2 and 47R resistors are in series with capacitors, passing only AC?
CFH9
Unfortunately the signal gnd network shorting can't solve the problem.
It doesn't make sense if it's shorted or open,TESTED
Tested on a new +/-50v power supply (+/-5mV ripple) with a little better results but the problem is here.
I see better results when resistors(18R/10w) placed in series with the rails.
Unfortunately the signal gnd network shorting can't solve the problem.
It doesn't make sense if it's shorted or open,TESTED
Tested on a new +/-50v power supply (+/-5mV ripple) with a little better results but the problem is here.
I see better results when resistors(18R/10w) placed in series with the rails.
Last edited:
I assume you are referring back to post633
. Could you reword your questions. I don't understand what you are asking.
Unfortunately the signal gnd network shorting can't solve the problem.
It doesn't make sense if it's shorted or open,TESTED
Tested on a new +/-50v power supply (+/-5mV ripple) with a little better results but the problem is here.
I see better results when resistors(18R/10w) placed in series with the rails.
Interesting.
Can you upload some pictures including the PCB setup, heatsink and signal wires? I am beginning to think we are reaching the PSRR limit of the amplifier, though it looks like it should have better performance.