Aleph-phase lag with non-inverting input

[jupiterjune]

I built a aleph mini with boards from KK some time ago. The amp sounds amazing, but I found that when I used the positive (non-inverting input) there was a phase lag in the left (aka bad) channel. When I use the negative (inverting) input, there is no phase lag at all.

Attached is my oscilloscope output of this (I hope, I am new to using this forum). The sinewave is 20kHz and the vertical is set to 5V/ division. Both channels are fed the same signal via a y-splitter connected to the positive inputs. As you can see, there is a clear lag in the phase of one of the outputs. This goes away when I feed the negative inputs. The unused inputs are shorted to ground.

I have tried the obvious:
changed the output mosfets (irf 634's)
changed the NPN transistors
it was a long time ago, but I believe I changed the input mosfets
I am pretty sure I changed the CCS mosfet, but again, this was over a year ago
I swapped the input wires back and forth, and fed the board directly (not through chasis mounted RCA inputs)

All the voltages are similar for both channels, I will post when I get info onto the computer.

Some things that might be significant:

I did just realize that I have an IRF9630 input CCS mosfet on the channel with no variation, and an IRF9510 input CCS mosfet on the channel with the phase lag -- could this be producing a different bias current in the two channels to the input mosfets?

I changed the bias setting resistors (in series with the output gain and CCS mosfets) from .33 ohm to 1 ohm -- they show a .66 V voltage drop across each

I crammed a terminal block onto the board's input holes. It is nor really sized for this. Could I have made some capacative connection on the input that is causing the phase lag?

I plan on getting out the 'iron this weekend and trying some more changes. My ultimate plan is to cascode the input mosfets with jfets, but I want to work this problem out first...

[GRollins]

Are the front end differential MOSFETs the same in both channels?

Grey

[PrestonTom]

At the risk of .....
Are you sure the phase shift is not in the scope? I have actually seen this happen before (B & K and Fluke scopes).
-Tom

[Babowana]

Originally posted by jupiterjune


Attached is my oscilloscope output of this (I hope, I am new to using this forum). The sinewave is 20kHz and the vertical is set to 5V/ division. Both channels are fed the same signal via a y-splitter connected to the positive inputs. As you can see, there is a clear lag in the phase of one of the outputs.

I think that the different phases mean the voltage gain difference between the left and right channels caused by the different locations of the upper critical frequencies to the same test signal. The most probable reason seems to be that the capacitances of the plus input side MOSFETs are different between the left and right channels.

The reason why I say this is that I can see the different V peak2peak values from your scope screen.

[cviller]

Originally posted by Babowana

The reason why I say this is that I can see the different V peak2peak values from your scope screen.

I think the voltages are just shifted vertically to illustrate the difference...

Originally posted by jupiterjune

I did just realize that I have an IRF9630 input CCS mosfet on the channel with no variation, and an IRF9510 input CCS mosfet on the channel with the phase lag -- could this be producing a different bias current in the two channels to the input mosfets?

I changed the bias setting resistors (in series with the output gain and CCS mosfets) from .33 ohm to 1 ohm -- they show a .66 V voltage drop across each

I crammed a terminal block onto the board's input holes. It is nor really sized for this. Could I have made some capacative connection on the input that is causing the phase lag?

I plan on getting out the 'iron this weekend and trying some more changes. My ultimate plan is to cascode the input mosfets with jfets, but I want to work this problem out first...

Have you tried to switch your probes and channels on scope? What is the bandwidth of your scope?
I'm not sure about the CCS transistors if they supply the same constant current, the difference shouldn't matter, but I guess the 630 might supply much more current unless you have made some adjustments. I think a higher current through you diff pair would give you a higher corner frequency. You could try to swap them or put in some identicals.
Try to measure if you have the same current through the diff pair on both channels.

[jupiterjune]

Gray-
The input mosfets are the same in both channels, IRF9510's. They were matched fairly closely for Vgs for another project, but I don't have the exact numbers--I will check if I end up removing them.

Cviller-
You are correct about the traces being offset vertically to show the phase difference, they sort of fuzz together otherwise. When I switch the feed to the negative (inverting) inputs it is clear that he lag is gone.
I am not sure that the input CCS mosfets are supplying the same current, they both have app. 10volts on their gates. __I will try puttting a little more voltage on the gate of the input CCS mosfet tonight--it should then pass more current....
Measuring the current through each individual mosfet in the differential pairs seems impossible--I will post the voltages I get once a get a circuit dwg to attach

PrestonTom, Cviller-
I am 99% certain it is not in the probes or scope, but I will check it again tonight. If I switch the feeds to the neg. input, it will disappear. It should stay gone if I then switch the probes on the output--I will recheck this. After that, I will switch channels that the RCA connections from y-splitter are feeding and repeat. I don't remember the bandwidth of the scope--I will check and post. I think it is a B&K.

Bobowana-

The outputs are offset vertically. The gain is essentially the same for the two channels at 20KhZ. At 100KhZ, I think I noticed the bad (left) channel had a little less gain, and the offset was definitely much greater.

Also, I can try using the capacitance meter on my fluke DMM to see if I can read a capacitance in the input circuit...(did not think of this before)


I have my list of things to re-check tonight!

[cviller]

Is your offset in the "bad" channel positive or negative? If negative, then you might have more current through your diff pair in that particular channel - at least that is how it works in my fishamp.
If you short the input, I think you can get a relative estimate of the current by looking at the voltage over the resistor connected to one of the legs of the diff pair.

[Nelson Pass]

Even if it's the amp, I don't think that degree of phase
shift at 20 KHz is actually significant sonically.

[GRollins]

Is the bandwidth the same in both channels?

Grey

[Babowana]

To have that phase difference
the bandwidth must be unequal
at the range of the sonical peace
as Papa smiles . . .

Let's hear what he says.