This statement by Mark got me to thinking...
mwhouston wrote:
Frequency response <10hz to 56Khz into 22K @ 80% volume (this has fallen sharply)
I think that sometimes we don't get a very good idea of what is happening with output impedance in some of our line stages. This stage is a good example. The 12AU7 stage has a grid circuit output impedance of 6450Ω without the 100k volume pot on the output.
Now putting the volume pot in this position has several real advantages. The first is that the noise figure of the stage is significantly increased by putting this large resistor after the first gain stage. Hence a much quieter stage. The second is that it gives us some flexibility with what we want to do with the next stage w.r.t. headroom and total voltage swing. However, the real question is this: What is that volume pot doing to the preamp output impedance as we change the volume setting? The total output impedance in conjunction with the Miller capacitance of the next stage is what is going to set the high frequency rolloff. So let's investigate.
The output impedance is a fairly simple matter to calculate from the equivalent plate circuit model. If we look at it from a fractional resistance standpoint the output impedance looks like this.
Attachment:
Out Z linear.png
And if we are driving a next stage with a 1000Ω grid stopper and a miller capacitance of 100µµf then the high frequency rolloff looks like this.
Attachment:
Upper F linear.png
Unfortunately these plots don't give a lot of insight because in audio circuits we use audio (or log) taper pots. For these pots the 10% fractional resistance point is typically at the 50% shaft setting. So what happens to the plots when we change the x-axis scale to %shaft rotation on an audio taper pot? Well, they look like this.
Attachment:
Out Z log.png
And the high frequency rolloff looks like this.
Attachment:
Upper F log.png
Now we can begin to see why Mark made the comment above about the falloff being greatest at around 80% volume. That's where the output impedance is peaking and driving a 22kΩ input with a 26kΩ output impedance is going to produce some significant loss.
My only point in all of this is that, like all the other aspects of circuit topology, the placement of controls can have a significant effect on the circuit operation and sometimes single numbers, be they peak or average, don't always tell the whole story.
Questions or comments?