PIMETA - DC offset problem (and solution), additional questions
Mar 12, 2006 at 9:39 PM Thread Starter Post #1 of 17

civilmonkey

100+ Head-Fier
Joined
Feb 17, 2006
Posts
126
Likes
10
Hello,

First off, great forum folks. Everybody here makes it a joy to DIY audio because of this resource here. Sorry in advance for this long post!

After creating a few CMOYs, and A47s, I moved up to a PIMETA, and it sounds great! My configuration is stock as per Tangents site, here are questions people might have:

-R4 = 4.75K for a gain of around 5.6
-AD8620 (L/R channel), AD8610 (Ground) on Brown Dog adaptors
-Powered by 22 VDC from stock TREAD (supplied by 26V 400mA AD-DC wallwart)
-Biased into Class A using JFET cascodes (saturated current for Q2 around 3mA, and 2ma for Q1)
-2 buffers on all channels (including ground). R11 resistor of 100 ohms (again, all chanels).
-ALPS097, with ground strap to IG pad (to fix slight hum when touching POT)
-6.8 WIMA caps
-ELNA Cerafine 470µF x 4

1. I created all my buffer stacks before realizing that most people don't stack buffers on the ground. I'm assuming that it won't negatively affect my sound quality, but please let me know.

2. Here was my first problem and solution: measuring from the IG pad to the left, right, ground output pins (26 mV, 16 mV, 4 V). I removed the ground buffer stack, and fixed a cold solder joint connecting the buffers. New DC offset is 10mV,11mV,26mV. Measuring from IG to OG, I can less than 1 mV. DC offsets are still higher then I'd like, but I've spent the weekend on this now! (BTW, my source puts out less then 1 mV offset.)

3. I understand about DC offset on L & R channels damaging headphones. Does DC offset on the ground channel have implications?

4. Finally: My PIMETA draws 58.8 mA (steady). This seems high to me:
Amps: 3 x 4 mA = 12 mA
Class A Bias: 2 x 3 mA = 6 mA
Buffer stacks = 3 x 10 = 30 mA
LED: 2mA (measured)
TLE2426: < 1mA
Total is around 50 mA, unless I'm missing something.

As far as touble shooting goes:
-flux is off the board
-I can't see any cold solder joints, v- and v+ on pins is consistent from chip to chip and within 0.1V.
-resistance from input voltage (positve and negative) to power pins on all chips is 0 ohms
-amps stay cool, the buffer stacks are luke warm (finger need to press hard to feel the heat, but it's there).

I can live with the current draw if I can't explain it, but I like to know what's going on!

Thanks in advance.
 
Mar 12, 2006 at 10:08 PM Post #2 of 17
DC Offset is measured from each channel to the virtual ground.
 
Mar 13, 2006 at 9:15 AM Post #5 of 17
I've built a few pimetas where swapping the buffers fixes the offset (i.e. swapping left and right buffers evens things out a bit more). In terms of your current draw, there's another thread on here where Tangent outlined what you should be seeing in terms of current draw for a given pimeta setup. If you search, you should be able to find it.
 
Mar 13, 2006 at 1:23 PM Post #6 of 17
I have very similar setup. Offset seems too high. Things I'd try:

1)Bridge IG and OG, and check again. If DC goes away, 8610 or buffer stacks are the issue.

2)Bridge LR buffers. Same check.

3)If your resistros are much off the Tangent's numbres, such DC could be...

4)Follow the power. For the numbers you mention there could be a +100Kohm leak somewhere or misplaced resistor. Check resistors markings.

Oh yes, eliminate cascode for a start if it's not too hard.

Good luck!
 
Mar 13, 2006 at 2:54 PM Post #7 of 17
This amp sounds really good, I think it's worth the extra money put in from an a47. You're giving me some inspiration to pick up the slack with some projects I've put off. I think my wallet will want to have a few words though...

good project. I'm going to give it a listen in a minute. Thanks for letting me borrow it.
 
Mar 14, 2006 at 3:37 AM Post #8 of 17
Swapping buffer stacks around helped balance the DC offset, so I'm guessing the problem might be there.

I'm grappling with measuring DC offset, and have had problems finding an answer in the forums. When I meaure DC offset, I have a 3.5mm to 3.5mm cable, in the headphone (output) jack. I measure from the ground to the left and right. With my PIMETA, this test gives 0 mV all around. (This is also how I measure my source as well, mini to mini in souce output and measure)

Now, in Tangent's (essential) guide to the PIMETA, he suggests testing from input ground to output left and right. With this test I get the values I listed at the beginning of the thread (~11 mV for left and right).

Measuring from output ground to output left/right makes sense to me, as this is what the headphone will be seeing. Can someone take a stab at clearing this up?

Thanks very much.
 
Mar 14, 2006 at 5:16 AM Post #9 of 17
Quote:

Originally Posted by civilmonkey
Measuring from output ground to output left/right makes sense to me, as this is what the headphone will be seeing.


That's true, but the two measurements should be essentially equivalent if IG to OG is very near 0V. I suggest measuring against IG only because that's often more convenient.

If you get 0 mV of offset when measuring OG to OL and OR, but 11 mV when measuring against IG, then I question your measurement that gets 0 mV from IG to OG.

On your current draw question, remove the buffer stacks and re-measure the amp's current draw. If it goes down to around 20 mA, then you know the buffers are drawing more than they should.
 
Mar 14, 2006 at 5:47 AM Post #10 of 17
Quote:

Originally Posted by Xakepa
...
1)Bridge IG and OG, and check again. If DC goes away, 8610 or buffer stacks are the issue.

2)Bridge LR buffers. Same check.
...



If by "bridge" you mean to short circuit between these points, this is a no-no. Just what do you think you would accomplish with this anyway?
 
Mar 14, 2006 at 1:49 PM Post #11 of 17
Ouch...I ment "bypass", eliminate...short circuit input-output pads for the buffers of EACH channel, if you did not soldered the buffers to the board yet.
 
Mar 14, 2006 at 1:53 PM Post #12 of 17
Quote:

Originally Posted by tangent
That's true, but the two measurements should be essentially equivalent if IG to OG is very near 0V. I suggest measuring against IG only because that's often more convenient.


That clears that up, thanks. I dislike asking questions on DC offset, as it is such a common theme on these forums, and I should have been able to find the answer myself.
Quote:

Originally Posted by tangent
If you get 0 mV of offset when measuring OG to OL and OR, but 11 mV when measuring against IG, then I question your measurement that gets 0 mV from IG to OG.


So do I. I thought it was wierd that on every amp I mesaured like this (I got 0 to 1mV). My multimeter reads in volts, but to three decimal places (i.e 0.020 V) so I don't believe that's the problem, however, I have an old multimeter which I will try to rule that out. I will investigate.

Quote:

Originally Posted by tangent
On your current draw question, remove the buffer stacks and re-measure the amp's current draw. If it goes down to around 20 mA, then you know the buffers are drawing more than they should.


Again, I'll investigate and report back. I did do this at one point, but I can't remember the values now.
 
Mar 14, 2006 at 4:27 PM Post #13 of 17
Well, it could be a systematic DMM error

DMM are rated in % error over the range, so on V scale 2-3mV could actually be below the precision limit...but 10+mV most probably should not.

If you can, try a different meter with dedicated mV range.
 
Mar 14, 2006 at 11:39 PM Post #14 of 17
Woa,

First off, Tangent you are right, my IG to OG reading was suspect.

I measured DC offset on the PCB, from IG to out L and out R pins on the AD8620, and OG to out L and out R (again pins on the AD8620) - same reading around 12 mV now. IG to OG is around 0 mV.

EDIT: If I measure ground out (or ground in) to left out and right out at the output jack solder lugs, i get 0 to 1 mV....?

The multimeter I'm using has a precision of 100 uV +/- 0.5% of the reading, so my meter is alright I think.

As suggested, I tried to see how much current various components were drawing. I let the amp warm up - with everything in it measures 58.0 mV (starts at 58.9). Removing the buffer stacks (one at a time) or the opamps I get this

-Amp with BUF stack (left) out, draw = 43.3, so the buffer stack is drawing 14.7 mA
-BUF (right) draws 14.4
-BUF (ground) draws 13 mA
-the AD8620 draws 21.4 mA
-the AD8610 draws 4.5 mA

This is seeming like a bigger problem then I thought!. I double checks the bandwidth resistors (and mesaured) and they are 113.5 ohms. So, the buffer stack should be drawing a total of around 10 mA.

any ideas?

I'm off to trouble shoot some more!
 
Mar 15, 2006 at 9:19 PM Post #15 of 17
Quote:

Originally Posted by civilmonkey
checks the bandwidth resistors (and mesaured) and they are 113.5 ohms. So, the buffer stack should be drawing a total of around 10 mA.


It may be that we don't completely understand how stacked buffers and the bandwidth resistor interact. I don't think anyone has done exhaustive tests on this.
 

Users who are viewing this thread

Back
Top