New Posts  All Forums:Forum Nav:

DC Offset

post #1 of 13
12/12/10 at 2:23pm
Thread Starter 

I had measured DC offset on Matrix M Stage headphone output and found:

 

Left channel - 0.8 mV

Right channel - 8.4 mV

 

I tuned blue trimpots after LM 337 and LM 317 to 15V each and I changed opamp (OPA 627 instead OPA 2134) and result is the same (same DC offset).

 

I noticed one more thing... When amplifier is cold, offset on channels is almost same (around 4mV) but when amplifier goes hot (after 30 min) DC offset on left channel falls below 1 mV and on right channel goes up to 8 mV. I made measurements without DAC signal on input.

 

 

I am not concern about damaging headphones (Grado SR225).

 

I would like to know why DC offset isn't same on both channels?

And why there is a change until amplifier becomes hot? 

 

AMP-1.jpg

 

 

Thanks!

 

 

Reply
post #2 of 13
12/13/10 at 12:07am

The output buffer is not wrapped in the global negative feedback loop, which means that the DC offset won't be in-check and could drift significantly due to thermal gradients.

Reply
post #3 of 13
12/13/10 at 12:29am
Thread Starter 

What would happen if I reconnect opamp feedback from its output, to headphones output? Would that solve DC offset?


Edited by BlaBlaBla - 12/13/10 at 12:31am
Reply
post #4 of 13
12/13/10 at 1:11am

Most likely yes, but this is a commercial product and you'd void the warranty by doing so.  At any rate the DC offset voltage you got are not dangerous, if it was me I would leave it alone.

Reply
post #5 of 13
12/15/10 at 3:04pm
Thread Starter 

Amb, thanks for advice!

Reply
post #6 of 13
12/20/10 at 9:30am

doesn't DC offset always color the sound?

Reply
post #7 of 13
12/20/10 at 11:24am

No but it might give impression of slight sound confinement. Especially on low impedance headphones where the offset constitutes more percentage of the signal swing. Regarding the rest - 8.4mV is no disaster especially that it's further divided on the output resistor in this particular amp and leaves 6.4mV on 32 ohms.

Reply
post #8 of 13
12/20/10 at 3:02pm

To answer the OP, as to why it changes.  Transistors (and opamps) are active devices.  As such, temperature changes the operating points.  Look at any datasheet and you'll see expected average operating points relative to temperature.  Temperature will affect all devices, negative or positive temperature coefficient devices.  That's basically saying that things change as temperature changes.

 

For the "target" temperature the transistors were probably not matched, and the resistors may not be matched.  It's also possible that the resistors for gain section are not balanced as well.  It's possible that closer matching would reduce the DC some more, but that said, 8.4mV isn't all too bad.  You can probably get it down to about 3mV or so by matching components to a tighter spec.

 

This is a general reason why DC servo loops are a good addition.

Reply
post #9 of 13
12/20/10 at 4:14pm

Don't DC servo opamps color the sound either? especially when dual opamps are used.

 

My old Firestone Spitfire DAC outputs 3 and 4mV of DC offset, but the newer MK2 version output 8mV on the left channel and 19mV on the right one.

 

I was under the impression than DC decouping caps and DC servo opamps would color the sound too much. Even though I'm not sure how they managed to reach 3mV w/o either of those in the old Spitfire(even when using LT1363, whatever opamp used as PCM1793 LPF outputs 3/4mVDC).


Edited by leeperry - 12/21/10 at 3:31am
Reply
post #10 of 13
12/20/10 at 6:09pm

Whether or not a servo colors the sound depends on where you inject the feedback.

Servos operate at very low frequency and shouldn't color the sound...

If you inject it into the current source/sink rather than the feedback loop it should

not be audible.
 

Quote:
Originally Posted by leeperry View Post

Don't DC servo opamps color the sound either? especially when dual opamps are used.

 

My old Firestone Spitfire DAC outputs 3 and 4mV of DC offset, but the newer MK2 version output 8mV on the left channel and 19mV on the right one.

 

I was under the impression than DC decouping caps and DC servo opamps would color the sound too much. Even though I'm not sure how they managed to reach 3mV w/o either of those in the old Spitfire(even when using LT1363, whatever opamp used as PCM1793 LPF output 3/4mVDC).

Reply
post #11 of 13
12/21/10 at 3:00am

I have never heard a competent servo colour the sound at all, its below the audio band. they would have simply injected some DC into the inverting input to zero the DC without a servo, this method wont adjust if the amount changes, but if constant its a simple solution

Reply
post #12 of 13
12/21/10 at 8:40am
Quote:
Originally Posted by qusp View Post

I have never heard a competent servo colour the sound at all, its below the audio band. they would have simply injected some DC into the inverting input to zero the DC without a servo, this method wont adjust if the amount changes, but if constant its a simple solution



if it changes at a slow rate, as DC offset should, the servo adjusts with it.  Most servos operate at 1Hz or less.

Reply
post #13 of 13
12/22/10 at 3:41am
Thread Starter 

I could change BC and BD transistors and find matched pairs. But I don't want to mess (too much) with this amp. I was curious because of its simplicity. It doesn't have any mechanism for DC offset suppression. It seems that all lies on quality (good matching) components. Better components = lower DC offset. I like this approach. I like simplicity. :) Next few months, I will buy parts for CKIII. And see how it will behave under same circumstances (same DAC). It has servo and different topology (more complex). I expect better results. Technically, of course. For sonic results I am not sure, because I have never heard CKIII. I must admit that I can't speak about technical sound, because I like slightly "imperfect" or "colored and warm sound". Best example is that I had AKG k701 and they (fortunately) went to hell. "Fortunately", because they were the most boring headphones ever. And I couldn't find source, that would give them warm and lively signature without "100% technically flat" approach.

Reply
New Posts  All Forums:Forum Nav:
  Return Home