Yet another Aleph3 question

[Kanawat]

Hi, I'm a newbie in this forum. In fact I just finished building a pair of Aleph 3 mono block, thanks to you all .. I really get a lot of insight in building an amplifier from this forum. The amp sounds really great ... until ... I decided to stack my DIY A3 up against my uncle original Aleph 3 from Pass Labs. Overall sound quality is almost identical except that my A3 has a little more pronouced high/ more treble. It can get to the point when I feel uncomfortable after listening to it for a while.

I used IRFP244s as the active current source and the output stage. I reduced the source resistor of the active current source to get about 2.4 amps bias. A little higher than then spec which helps reduce the problem a little bit. But I don't think this is the right way to go.

Does anybody experience the same problem as I am?

[Nelson Pass]

If the AC current gain ratio of the current is much higher than
50%, then this will result in sonic degradation. Also, as the
open loop gain is increased, you again get degradation. Lowering
the value of the Source resistances will get you both, and you
may need to check the ratio of current supplied by the current
source against the output current, and perhaps take a look at
the square wave with a scope. While you're at it, look for
oscillation.

[Kanawat]

Thank you Mr.Pass!! I was under the impression that pumping more bias will make the device more linear without knowing the downside of it.

Could you please describe more about the sonic degradation you've mentioned? Will I be able to observe it in the output waveform? either in time/frequency domain?

[Nelson Pass]

Pumping more DC bias will make the devices more linear, but
the Aleph current source has a constant DC component upon
which is overlayed an AC component which is an arbitrary
percentage of the output current, anywhere from 0% (CCS)
to close to 100%. As a practical matter, greatest efficiency is
seen at about 50% if you want to always remain in Class A
up to voltage clipping. As you increase this percentage, the
performance measures better at the lower levels and less well
somewhere below clipping, where the current source starts
shutting down. There is also a character that creeps in at high
gain settings which is not preferred, so the general
recommendation is around 50% or less.

In actual testing, I have listeners tending to prefer figures of
40% over 50%, and 40% over 30%, so it looks like the sweet
spot is maybe the 40% range. This does vary from version to
version and depends a lot on the taste of the listener, so I
don't spend a lot of time worrying about the absolutes.