Help with AX100 dc offset-Grey-GL-NP are you out there

I posted this on GL's AX100 thread and I am getting ready to listen to the newly constructed AX100 mono amps but I have trouble. I have to travel again so therefore my urgency.

In doing final testing for the Aleph 1.2 conversion to the AX100 using 6 output devices per quad and after carefully adjusting the abosolute dc offset to .25mv and doing sine wave testing etc I thought things were ok.

However my relative dc offset is too high as it is also .25mv. My amps are dc coupled in balanced mode. I tested in single ended mode.

Of course I first suspected the input differential pair (I am using GL's schematic of Grey's design). So I measured in circuit with the input shorted (my rails are about +/-27 volts under load).

I measured the Gate to Source voltage of each half of the diff pair and got these results (from in front of the 221 ohm gate resistor directly to the source pin of Q6&Q7):

3.805Volts
3.799Volts

diff is 6mv-seems ok

I then measured across each of the diff input pair 392 ohm drain resistors R22 and R26 on GL's schematic and got these results:

4.86 volts
4.77 volts

diff is 90mv

I then measured absolute dc offset for each half of the amp and got

+50mv--positive leg
-200mv--negative leg

My conclusion is that the matching in the 24 output devices-12 per half and 6 per quad is not good enough and I should measure each output source resistor and then swap from side to side to get better balance?

Or

Change the input diff pair, I have one set of IR9610's whose gate to source measurement at 10ma is 3.766 or 0 difference?

Or

In a past thread NP mentioned that this amp should be ac coupled to prevent dc offset problems if the inputs have any dc.
I don't think this will matter in my case but I can try it.

All the other values are the same as on GL's schematic. I have not yet tested the 50% gain since I used his value.

The PM wants her xmas lights up, the dogs walked etc and I have only time for one experiement-any advise? Thanks for your time.
I am nearing the finish line for this project and I am trying to listen to it before xmas.

dave
 
Hi Dave,

Wow. I almost never check in on Saturday morning. Too many chores! You say .25mv but again I think you have a typo. .25mv is really really good for any offset on any amp. I think you mean 250mv relative offset. That's not as good as you could do. I would suspect the input diff pair. But then again I didn't really like the spread of Source resistor voltage readings you were measuring earlier. You are in places here where I haven't traveled so I'm just making a best guess. I wish I was actually there - then maybe I could get some better traction on this.

Anyway I now have to go and get the Christmas decos down from the attic.

Graeme
 
Assuming that you meant 25mV...
One, that's not too bad.
Two, if you want to get it down lower, look at output device matching. You might consider matching the outputs at operating current, as opposed to something less. Transistors don't necessarily track linearly and no matter how closely you match them at 20mA, it doesn't necessarily mean that they're matched at 1A.
Three, you can try to fiddle it a bit with the Aleph current source pots.

Grey
 
ax100

Grey and Gl thanks, I am just trying to make this thing bullit proof.

I changed the differential transistors today and again my absolute dc offset after adjusting on and off all day was good at -100mv on the +signal speaker output and +100 at the negative signal ouptut. The gives a relative offset of 200mv. AC signal tests looked very good and my heat sinks are working well at about the same temp as my Aleph 1.2.

I believe I don't have my outputs matched well enough so I am going to measure all 24 source resistors while the amp is running and warm. I can do this by facing them upwards. I left enough connecting wire to allow me to do this. I will then swap between quads until all quads have close to the same total. I think this should do it.

I am pleased that the testing is almost over and this last thing is a pain to do but hopefully this will allow many years of good listening.

Do you know what NP is doing with his dealers? No Pass products in NJ or NYC right now. I was hoping to hear his new inventions. dave
 
Hi Dave,

Pass Labs is currently re-establishing their domestic dealership network. This has been discussed elsewhere on this forum. The quickest and best thing to do is to email Pass Labs or better yet phone them.

I think your idea of checking the current through each output transistor is a good one. Please publish the results when you have them. I suspect that you had a mistake in your matching process.

Graeme
 
ax100 dc offset

GL and NP thanks, I will measure all 24 output fets quad by quad and try to match both sides and then recheck the relative dc offset. I am very happy with the absolute dc offset as it is.

I prefer to do this than play with the bias pots as I have to reinstall them. I am using your value of 100k because my heatsinks are good enough for this output. I could raise the bais slightly on one side by using the pots but I am not sure if this will affect the absolute dc enough.

NP thanks for the dealer information.

I would love to hear an xa anything in action so I can hear to the best of my ability what to expect out of my project or who knows I may be the 1,000,000 customer to cross the dealers threshold entitling me to a 50% discount.

I am away on business this week and will get back to this final correction this weekend. dave