Volume control of a differential signal

[soongsc]

I assume it would be better to try and keep load impedance constant.

[berni8k]

Well Andrews idea was for normal volume control.To get power control i need the exact opposite.Turn a log pot around so its fast on the beginning rather the end as this makes the output power linear with the pot setting.

Oh and is matching resistances that important?

[soongsc]

Originally Posted by berni8k

Well Andrews idea was for normal volume control.To get power control i need the exact opposite.Turn a log pot around so its fast on the beginning rather the end as this makes the output power linear with the pot setting.

Oh and is matching resistances that important?

I would think it depends on how critical of sound quality one desires, but really have not explored this part yet. Some old Wireless World articles talked a lot about impedance matching.

[jcx]

Originally Posted by berni8k

...
I like to avoid digital attenuation in the DSP chip thats driving the DAC because it causes loss of bit depth.

not likely true with Delta-Sigma DACs, many have differential linearity far below the analog output noise floor - much less than the noise floor of any real world recording - and the DACs digital attenuation should track much better than any analog solution

[rdf]

Originally Posted by AndrewT

twin pots will ruin the balance of the impedances.

I didn't recommend tying the 'ground' end of the pots to ground so I don't think that's correct. To see why, feed identical signals to the two signal ends of the pots to emulate common mode interference. Since no current flows through the pots the voltage drop across any two points is zero, no matter where you tap off to feed the differential input.

[berni8k]

Okey i did some spice simulation on this differential signal impedance thing. And it is true putting different value resistors on each leg introduces interference in the signal, but its not as critical as it seams. 1K of mismatch between them caused 0,11dB of the 6dB ground noise to bleed in to the signal.

So yeah it is true that non matching impedances cause noise to get in but even at a 1K difference the noise that got in is very tiny, in my opinion too tiny to worry about.The DAC and amp are in the same case so ground shouldn't be very noisy.(Amp also has a analog ground)

[AndrewT]

Bernie, go and read Jung and Self.
0.1% error reduces the interference rejection of a balanced feed. 1% error massively reduces it.

[berni8k]

Well the simulation has shown that the noise that comes in to the signal is not dependant of the % of difference between the impedances but by the value its off.This is because 1K and 2K on the differential line has the same noise as 10K and 11K.

This again is impacted by the amplifiers input impedance.(My chip has 33K) The higher the impedance of the amps inputs are the more difference is alowed on the impedance of the differential lines.

So in the end it all comes down to what you have on the end of the differential line.Lower its impedance the less tolerable it is to impedance differences between the two differential lines.At leats that whats SwictherCAD IV has shown.

[janneman]

Originally Posted by AndrewT

Bernie, go and read Jung and Self.
0.1% error reduces the interference rejection of a balanced feed. 1% error massively reduces it.

Yes, true. At 1% difference between the two source impedances limits your common mode noise and hum reduction to max 40dB. At 10% imbalance, a puny 20dB.
It's the difference ratio that determines the cancellation, not the value, although the value as such does of course impact the noise generation.
Bottom line: use lowest value resistors that you can handle, and then make them match.
There's a lot of this material on the web when you google for instance 'Bill Whitlock'

jd

[tritosine]

theres also an AES lecture in mp3 by Bill Whitlock somewhere