Help with Levinson ML-9 offset voltage

I have a problem with an ML-9 amplifier channel, and I need advice. One channel works fine, but the other doesn't. I have disconnected all output transistors, and although I am able to see an amplified input signal at the output of the second amplification stage (the input to the 7-diode array), the DC offset at that point should be about +2.25 volts; instead, it's 420mv. The voltage across the 7-diode array is 4.53 volts (0.42 volts at the top of the array and -4.11 at the bottom, for a total voltage drop of 4.53.) Because the DC offset is almost 2 volts more negative than it should be, the pre-amplified signal saturates when the input voltage level is increased beyond about 100 mv. So far, I have not been able to determine why the DC offset across the diode array is 2 volts more negative than it should be.

I bought this amp for a really great price because someone before me unsuccessfully tried to find the problem; several parts were replaced. As a result, there may be more than one problem layered on top of the original problem. I don't mind clipping component leads or circuit board etches, but removing transistors will be a last resort, given the difficulty of desoldering these components. I have lots of voltage measurements, and I'm happy to take more measurements or replace some of the components.

Thanks for your help.
 

Attachments

  • ML-9_schema_amplifier board_JP_V1.jpg
    ML-9_schema_amplifier board_JP_V1.jpg
    204.4 KB · Views: 60
I’m assuming the “lollipop” next to the output and the one at the rightmost base on the LTP means that they’re connected as a global feedback loop? If that’s the case, removing the output transistors breaks the feedback loop and I wouldn’t trust any voltage readings. It looks like it has some kind of local feedback too, but I’d still be suspicious about any voltage I measure without the full loop in place.
 
Thanks so much! I'll reconnect those transistors and monitor the changes. By the way, the "lollipop" is a power transistor with a second transistor glued to the surface. I believe the purpose of these pairs is to provide heat compensation by adjusting the bias as the temperature rises. I think that may help prevent thermal runaway in the output stage.