Netlist said:The nicest way to reduce offset is what Peter says in post #11.
Second, I used the resistors from common input sources to outputs and I set the value at 4.7k. Bringing it further down to 2.2k increased stability of the offset.
Coupling caps at the input also helps a lot, not only for the zero offset, but also for offset drift.
I never used 100ohm resistors at the output and it works great: 1 to 2mV after 10 to 15 minutes, and very stable.
Look at http://www.diyaudio.com/forums/showthread.php?postid=173696#post173696 and you will see how I mounted the input fets; very handy to change them till they match perfectly.
There's also a post somewhere that says to first adjust V1/V3, then V2. V2 has a great infuence on offset.
These are my findings when building the amp.
What value did you use for R19/29 and for feedback and input R's?
Did you use C2/4?
/Hugo - can't wait to see pics of the finished amps.
Hi,
First of all; I am talking about DC offset between OUT- and OUT+, I am NOT talking about absolute DC offset to GND...
Ok, I use 5pF for C2 and C4. I have the 4k7 from source to output in place. R19 and R29 are 10k (like in the 'final' from hifizen's PCB).
I have found this article from Klaus; http://www.diyaudio.com/forums/showthread.php?s=&threadid=10718&highlight=offset+aleph
in which he writes the higher zener output gives him problems. When I measure between my two amps (the prototype 25W Aleph-X with 4 mosfet and my 130W Aleph-X with 16 mosfet) the first has a zener voltage of 9.2V and the latter of 9.4V , can this be a source of my problem?
I have to go to work now, but I will test it tonight by swapping the zener...
Edwin
P.S. All my 1/2W resistors are 0.1% too!
grataku said:
That is what I thought; but what is 'the resistor' of the CCS of the input differentials??? The only thing I can come up with is the R24, R26, VR2 thing...
But these have a VR2 trimpot as default. Where o where to put another trimpot? I just have put in the 100 ohm resistors to ground, it does not do a thing for the OUT- OUT+ DC offset
Edwin
Edwin,
The 100 ohms between outputs and ground do nothing to differential DC offset. They only help a little so settle absolute DC offset ( i.e. between outputs and ground ) .
With 100 ohms they help only a little . If you reduce this value down to 30 ohms you will see a more effective control of absolute DC,but as Nelson stated we begin to charge the circuit with loads which affect the circuit , if you go below 30 ohms .
The key for me is to match merely 9610s first,at 10 mA each and have several pairs to swap in and out,and play on the VR trimmpot of the 9610 current source until you achieve lowest offset at iddle.
Implement 4k7 magical , or between 2k2 and 4k7.You will see it gives a different absolute DC offset at turn on , and that you will have to readjust the trimmpot to get a 0V DC between outs and grounds.
I don't think it is a good idea to adjust current in the 2 output sections to control the offset. In my prototype 23 V rails 4,25 A bias I used fixed resistors on both sides to get proper current in each Aleph out , and stays the same R value on each mirror-side.
Hope that helps !
Regards
The 100 ohms between outputs and ground do nothing to differential DC offset. They only help a little so settle absolute DC offset ( i.e. between outputs and ground ) .
With 100 ohms they help only a little . If you reduce this value down to 30 ohms you will see a more effective control of absolute DC,but as Nelson stated we begin to charge the circuit with loads which affect the circuit , if you go below 30 ohms .
The key for me is to match merely 9610s first,at 10 mA each and have several pairs to swap in and out,and play on the VR trimmpot of the 9610 current source until you achieve lowest offset at iddle.
Implement 4k7 magical , or between 2k2 and 4k7.You will see it gives a different absolute DC offset at turn on , and that you will have to readjust the trimmpot to get a 0V DC between outs and grounds.
I don't think it is a good idea to adjust current in the 2 output sections to control the offset. In my prototype 23 V rails 4,25 A bias I used fixed resistors on both sides to get proper current in each Aleph out , and stays the same R value on each mirror-side.
Hope that helps !
Regards
Edwin,
the zener should be no problem, you could always go to the other extreme and install a 6.9 voltage reference instead of the 9.1 zener. I would stick with the 9 V reference anyways. The trimpot in parallel with the resistor was put in to reduce the adjustment range, and it only affects the absolute dc offset (btw each output and the gnd) not the actual offset at the speaker, which is the one you are interested in. If you can bring the absolute dc down to within 100mv using the ccs pot V2 then it means you are in the correct range and you shouldn't mess with it further. Of course things may change when you play with V1 and 3 and you may need to readjust V2 and that may bring you out of range.
There is no question in my mind that your problem right now is mismatched mosfets and the bigger problem for you right now would be to take them out and re-match them. It is not straightfoward to find the perfect balance.
The only tolerance I am worried about is the 3W 0.56 source resistors (r 1-6, r40-45).
Ultimately, the same current should flow on both sides ie the voltage drop should be equal on all the source resistors and both absolute and actual DC offset should be as close as possible to 0. Whether you want V1 and V3 in place or not it's up to you.
the zener should be no problem, you could always go to the other extreme and install a 6.9 voltage reference instead of the 9.1 zener. I would stick with the 9 V reference anyways. The trimpot in parallel with the resistor was put in to reduce the adjustment range, and it only affects the absolute dc offset (btw each output and the gnd) not the actual offset at the speaker, which is the one you are interested in. If you can bring the absolute dc down to within 100mv using the ccs pot V2 then it means you are in the correct range and you shouldn't mess with it further. Of course things may change when you play with V1 and 3 and you may need to readjust V2 and that may bring you out of range.
There is no question in my mind that your problem right now is mismatched mosfets and the bigger problem for you right now would be to take them out and re-match them. It is not straightfoward to find the perfect balance.
The only tolerance I am worried about is the 3W 0.56 source resistors (r 1-6, r40-45).
Ultimately, the same current should flow on both sides ie the voltage drop should be equal on all the source resistors and both absolute and actual DC offset should be as close as possible to 0. Whether you want V1 and V3 in place or not it's up to you.
grataku said:
Thanks Grataku for the time you take to help me out, much appreciated;
1. I can change the absolute witin 100mV without any trouble so the V2 range seems ok. So I will not touch it!
2. I have set VR1 and VR3 to about .47V over the .056 ohm resistors. Which gives me a total bias per side of 3.35A . This is about what I am aiming for (between 3.3A and 3.5A) ... With this bias I can change the absolute within 100mV with V2 .
3. My 'old Aleph-X' had about 0.07V difference in differentials and it work ok... still a mystery...
4. I have adjusted the ACS gain to a perfect 50%
5. My 0.56 are 2% intertechnik metalfilm 5W resistors 300ppm.
Some questions; (again!)
1. I can take out the 9610's from my test Aleph-X and put these in, just to see if these are better 'matched'. Is this wise to do?
2. When I rematch; should I do it on 5mA 10mA and 20mA ??? What is the best way?
3. Should I buy new resistors and match them? But how?
Thanks,
Edwin
The 9610 are certainly a big part of the problem, but the biggest problem maybe the output mosfets, both the ccs and gain, since I seem to recall that substitution to a better pair caused you worst offset.
I would suggest to try to re-match them at two different current levels by changing the resistor in the matching circuit. To match the 0.56 you should be able to insert them in the mosfet matching circuit by keeping the mosfet the same and changing resistor. In any case, the best thing is to use a matching current that is similar to the current at which the mosfet will operate in the actual amp circuit.
I would suggest to try to re-match them at two different current levels by changing the resistor in the matching circuit. To match the 0.56 you should be able to insert them in the mosfet matching circuit by keeping the mosfet the same and changing resistor. In any case, the best thing is to use a matching current that is similar to the current at which the mosfet will operate in the actual amp circuit.
Peter Daniel said:
Thanks,
I know this; the 9610 mosfet's give over 390 ohm (european E-serie) after 15 minutes of running 4.52V and 4.50V which means 11.58mA and 11.53mA .. So they look matched close enough so it seems... still I keep an DC offset of 118mV ...
I just found out the values of my ACS source 0.56 resistors after 20 minutes of running (hot heatsink), they all start higher but they decrease each about the same amount; otherwise said; the voltage difference between the mosfet's stay the same but overall they decrease when they heatup:
leftside:
4.63
4.59
4.62
4.48
= avg 4.58V
and
rightside:
4.55
4.59
4.64
4.51
= avg 4.57V
I will swap the 4.48V and 4.41V with new resistors to see what happens!
Edwin
Edwin Dorre said:
Sounds to me like you are getting closer to the 'source' of the problem so to speak
. Not that you couldn't listen to it now with 100 mv DC offset but it sounds you'll be able to connect the speakers in very soon. I am looking forward to reading about your listening impressions...
Edwin,
Just wanted to congratulate you on a great looking amp! Sorry I haven't had the time to read and post on the forum lately (heck, I haven't even built anything more than the test channel I put together with the beta PCB). But, I hope to change that soon... I'll be keeping an eye on this thread to see how you get on.
Cheers,
Chad.
Just wanted to congratulate you on a great looking amp! Sorry I haven't had the time to read and post on the forum lately (heck, I haven't even built anything more than the test channel I put together with the beta PCB). But, I hope to change that soon... I'll be keeping an eye on this thread to see how you get on.
Cheers,
Chad.
grataku said:
YEAH!!!!
I matched a new pair... Running on 10mA and 15mA and 20mA ... They were measured with a 0.001V scale running on 10mA - 3.960 and... 3.960... they were running on a 0.01V on 15ma 4.02 and 4.02 and running on 20ma 4.07 and 4.07 !!!
And you know what? I only changed the diff pair and now I have about 0.034V DC offset!!! I did not even had to change the source resistor or the CSS trimpots! So with the trimpots I can probably get closer to 0V!
I am a happy man! Now I can start finishing the second channel!
A very big thanks for Grataku for his persistency on the diff pair problem!
Edwin
Edwin Dorre said:
Not to minimize Grataku's efforts; I learned a lot of them, but I knew from the start it was the diff-pair, simply because I had the same problem. That’s why I made my board with those connectors so I could easily change the pair. Trimming will certainly give you even less offset.
Glad you made it!!!
/Hugo
Netlist said:
Not to minimize Grataku's efforts; I learned a lot of them, but I knew from the start it was the diff-pair, simply because I had the same problem. That’s why I made my board with those connectors so I could easily change the pair. Trimming will certainly give you even less offset.
Glad you made it!!!
/Hugo
What did you know, chief? That you felt like taking a cheap shot at me?
I never said that wasn't the problem. Actually I said it was a big part of the problem but as I recall Edwin replaced the input pair earlier with questionable success so I thought further digging might have been required. Months ago I was actually very intersted on the subject of paralleling mosfets and the way their unequal vgs would be averaged out and I run a little simulation of what various mosfet and source resistor value spread may average to and that was a very small number for very large initial vgs spread so I was surprised that with 4 parallel mosfets there would be such a large problem, I wasn't ready to discount the possibility.
In my prototype with single device per side I found the problem in a single mosfet with apparently very different transconductance... well, different enough to produce 76mv offset. This was happening with the differential matched within 1mV.
grataku said:
What did you know, chief? That you felt like taking a cheap shot at me?
Don't feel offended; glad I saw the smile....
I'm not like that, you should know that from reading my posts.
The only thing that matters is Edwin's amp in this thread.You helped me a lot with my AlephX, so thank you again.
/Hugo
Netlist said:
Don't feel offended; glad I saw the smile....
I'm not like that, you should know that from reading my posts.
You helped me a lot with my AlephX, so thank you again.
/Hugo
Too make up for it; I have another question;
My absolute DC is very fond of wandering around. I tried the 100hm but they do not do much. I had the same problems with the test-aleph-X but I thought that is a problem for later. Mine moves down, seems to stablilize and then wanders on...
How did your absolute DC behave with Ian's 4k7's in place and with or without the 100ohm?
I will test 2.35 (two 4k7 parallel) tonight, as I recal you moved to 2k2 and no resistors to ground?
Edwin
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.