CAE Cascode Diffamp Front End

[gingertube]

I've got a couple of CAE PCB-1A front ends built up.
They are identical to the front end of this amp
http://www.diyparadiso.com/proj/sfm75.htm
Tubes are 6DJ8 (ECC88).

I'm a little short of gain and am wondering what I can do to squeeze out a bit more. For info: In my amp I have direct coupled 6SN7 cathode followers hanging off the diff amp and driving 2 pairs of KT88 in 30% Ultralinear + 10% Cathode Feedback.

Morgan Jones 3rd Edition tells me that the gain of a cascode is approximated to gm1xRL where gm1 is the gm of the lower tube. A quick calc suggested to me that this approximation is within 1% of the value obtained using the full gain equation.

The tubes are currently running at about 4.5mA each (with about 100 Volts Vak) so I could get a bit more gain by increasing the current to give higher gm.

Ideally I'd like to change to Svetlana 6N1P BUT these have lower gm and so problem would be worse.

The other avenue to explore is increasing RL.
QUESTION 1:
Could I run Current Mirror load on the diffamp? If I do, does the current source biasing need to be retained or could it be replaced by a simple bias resistor? Would I loose the self balancing "feature" of the common mode driven current source? Would a current mirror load provide self balancing anyway?

QUESTION 2:
Is there any benefit in changing the top triodes from self biased to a fixed grid supply (i.e. the more usual arrangement)? Any opinions on the sonic differences if any?

Any advice greatfully received.
Cheers,
Ian

[EC8010]

Hmmm. My experience when I need "a bit more gain" is that I actually need a lot more gain, like 6dB. You might be able to wring another dB or two of gain out of your existing circuit but it will be a struggle. Incidentally, I couldn't see any grid-leak resistors on the output valves. I assume they exist!

One possible solution might be to replace the upper valves with E180CC, the higher mu might enable another 3dB or so of gain (need a spreadsheet for those horrible equations).

The self-bias arrangement is probably quieter than fixed bias.