Thanks Nelson,
However I am using the group buy boards and the output is loaded to ground- everything according to the schematic.The diff mosfets are heatsinked together.
I find I get huge deviations with temperature.Seems like almost a volt for each 1 degree centigrade.This is the same for both monoblocks.
In fact the offset is so huge as I saw today when the amp is cold -up to 10v!! -that sometimes it doesn´t bias properly and turn on.I have to adjust the offset all the time to get it to work properly after two hours or so when the temp is stable then the offset stabilizes.But who can wait around for two hours or leave the amps on all the time?
Anybody else out there notice these variations?
As I said the main problem I think is with the CCS for the diff pair.That is the circuit that is thermally unstable.An extra simple circuit to stabilize it would be necessary.Or a kind of servo linked to the output.
I am sure Pass labs has solved this problem with the ax series.
However I am using the group buy boards and the output is loaded to ground- everything according to the schematic.The diff mosfets are heatsinked together.
I find I get huge deviations with temperature.Seems like almost a volt for each 1 degree centigrade.This is the same for both monoblocks.
In fact the offset is so huge as I saw today when the amp is cold -up to 10v!! -that sometimes it doesn´t bias properly and turn on.I have to adjust the offset all the time to get it to work properly after two hours or so when the temp is stable then the offset stabilizes.But who can wait around for two hours or leave the amps on all the time?
Anybody else out there notice these variations?
As I said the main problem I think is with the CCS for the diff pair.That is the circuit that is thermally unstable.An extra simple circuit to stabilize it would be necessary.Or a kind of servo linked to the output.
I am sure Pass labs has solved this problem with the ax series.
What I was always doing to add stability to differencial pair, was mounting devices directly to the chassis (main heatsinks). This provides more stability and when the amp reaches constant temperature, even opening the top don't change things much.
The best way for adjusting trimpots is throught the holes in a chassis, that way you don't upset termal stability (by removing the cover) and it provides convenient way for adjustement. That requires special layout though.
I believe Aleph 0 had it done this way
IIRC, in my Aleph X the absolute offset was starting hgh as well (5V or so), but after time it went down to 0V and stayed this way (no more that 100mV deviation). The differencial offset was suprisingly low after adjustement (less than 10mV).
The best way for adjusting trimpots is throught the holes in a chassis, that way you don't upset termal stability (by removing the cover) and it provides convenient way for adjustement. That requires special layout though.
I believe Aleph 0 had it done this way
IIRC, in my Aleph X the absolute offset was starting hgh as well (5V or so), but after time it went down to 0V and stayed this way (no more that 100mV deviation). The differencial offset was suprisingly low after adjustement (less than 10mV).
Hi Protos,
maybe you´ve seen my thread "one Aleph-X working, one to go"? Here I´ve described exactly the same problem you´ve got now. Mine also start at -8V, go down to -2V quite fast and then take forever to get to zero.
I also had to adjust the current source through a hole in the top cause otherwise absolute offset changes (just blowing into the amp changes it!)
Until now I did two things:
changed the output to ground resistors from 100 to 80 Ohms. This made the abs. offset go down faster.
changed R46/47 from 4k7 to 3k9 to 3k3.
This lowered the startup value(but not very much).
What I will try next is to make a thermal connection between the heatsinks and the diff pair as suggested by Peter, lower the output resistors to 60 Ohms and R46/47 to 2k7 (Netlist even uses 2k)
William
oh and I also tried changing the current through the zener (doubled it from 3 to 6mA) without any effect
maybe you´ve seen my thread "one Aleph-X working, one to go"? Here I´ve described exactly the same problem you´ve got now. Mine also start at -8V, go down to -2V quite fast and then take forever to get to zero.
I also had to adjust the current source through a hole in the top cause otherwise absolute offset changes (just blowing into the amp changes it!)
Until now I did two things:
changed the output to ground resistors from 100 to 80 Ohms. This made the abs. offset go down faster.
changed R46/47 from 4k7 to 3k9 to 3k3.
This lowered the startup value(but not very much).
What I will try next is to make a thermal connection between the heatsinks and the diff pair as suggested by Peter, lower the output resistors to 60 Ohms and R46/47 to 2k7 (Netlist even uses 2k)
William
oh and I also tried changing the current through the zener (doubled it from 3 to 6mA) without any effect
I would hate to waste more power with the output resistors.
I still think stabilising the current source is the way to go.I don´t think it is the diff pair temp.The diff pair affect more relative offset while the ccs affects absolute offset to ground.I will test this theory.Actually it is easy.
When the amp is warm with the top open touch the ccs with a frozen object and observe the offset change.Then do the same for the diff pair.If one affects the abs offset more than the other then ypu know which part needs more stability.
I think a slight variation in vgs of the current source is the cause.That´s why I suggested the current sink option which is included anyway with the boards.However nobody has commented on this aspect i.e. simple zener reg vs zener + current sink.
I still think stabilising the current source is the way to go.I don´t think it is the diff pair temp.The diff pair affect more relative offset while the ccs affects absolute offset to ground.I will test this theory.Actually it is easy.
When the amp is warm with the top open touch the ccs with a frozen object and observe the offset change.Then do the same for the diff pair.If one affects the abs offset more than the other then ypu know which part needs more stability.
I think a slight variation in vgs of the current source is the cause.That´s why I suggested the current sink option which is included anyway with the boards.However nobody has commented on this aspect i.e. simple zener reg vs zener + current sink.
Hi Protos,
you´re right about the waste of power with the output resistors but 60 Ohms is still 15 to 20 times as high as my Loudspeakers are so it won´t be that much waste......
I connected the zener to the negative power supply for a better imitation of a current source but I will have a look at the other option and see what parts I will need to test this.
Let me know when you´ve got results with the frozen object (or take ice spray as an alternative)
William
you´re right about the waste of power with the output resistors but 60 Ohms is still 15 to 20 times as high as my Loudspeakers are so it won´t be that much waste......
I connected the zener to the negative power supply for a better imitation of a current source but I will have a look at the other option and see what parts I will need to test this.
Let me know when you´ve got results with the frozen object (or take ice spray as an alternative)
William
My theory is that the zener is never going to be very stable and will change with temp.This might affect vgs by some mv and so make the ccs drift.This effect is amplified by the ccs and the diff pair subsequently.
The trick is to find which of all the destabilising factors is more responsible for the abs offset drift and try and minimize it as simply as possible.
I don´t know if I can find the recommended parts for the current sink but a small jfet can do the same job.
The trick is to find which of all the destabilising factors is more responsible for the abs offset drift and try and minimize it as simply as possible.
I don´t know if I can find the recommended parts for the current sink but a small jfet can do the same job.
I guess either /or.I don't have the schematic right here.But I remember it is a very simple idea.Both a zener and a cc sink would be a tighter regulation than either alone.But I am not an EE.I know the main purpose of the zener is to isolate the gate of the ccs from rail fluctuations noise.
I have to apply the Mr.Ice zapping technique to the fets to see if I am going in the right direction or I am spouting bs.
I have to apply the Mr.Ice zapping technique to the fets to see if I am going in the right direction or I am spouting bs.
Protos,
the zener won´t isolate from power supply noise. It´s only there to provide a biasing voltage for Q6. That´s why some people wanted a better (less noisy) reference voltage for the CCS. The other posibility would be to use a precision voltage reference like the LM329 and adjust R24/26 VR2.
William
the zener won´t isolate from power supply noise. It´s only there to provide a biasing voltage for Q6. That´s why some people wanted a better (less noisy) reference voltage for the CCS. The other posibility would be to use a precision voltage reference like the LM329 and adjust R24/26 VR2.
William
I took some time to re-measure the DC values on my AlephX and the behavior of it over a period of time.
As a reminder, I don’t use the output resistors to ground as recommended mainly because I never felt the need to do so.
When cold, I measure 3.5V absolute DC which goes down very quickly to below 0.5V and stays there with some +/- 100mV of wobbling. This is with the 2k resistors from common diff pair sources to outputs. The differential DC stays around +/-0mV-35mV, depending on how loud I play.
I then changed the 2k’s in 4.7k because I remember Nelson saying:
Difference with 4.7k is a greater tendency to drift and a smaller range to adjust with V2.
One particular thing I noticed now and in the past is that all adjustments affect each other.
V2 takes care of absolute DC as well as differential DC be it of a smaller magnitude.
Adjusting V1/V3 also has influence on DC in some way. Up or lower the bias with say 0.5A and a new adjustment has to be made to the offset. I tend to believe there’s a sweet spot somewhere.
The amp is now rather stable, knowing it’s a full open design. Literally every breath I take has effect on DC values.
Because the amp is now fully warmed up I’ll measure again tomorrow at first startup and report.
Schematic in link, notice also that the ccs is the one with the TL431. Feedback resistors have also been changed to 150k.
http://www.diyaudio.com/forums/showthread.php?postid=222538#post222538
/Hugo
As a reminder, I don’t use the output resistors to ground as recommended mainly because I never felt the need to do so.
When cold, I measure 3.5V absolute DC which goes down very quickly to below 0.5V and stays there with some +/- 100mV of wobbling. This is with the 2k resistors from common diff pair sources to outputs. The differential DC stays around +/-0mV-35mV, depending on how loud I play.
I then changed the 2k’s in 4.7k because I remember Nelson saying:
Whatever that may mean but I do understand not to go too low.
Difference with 4.7k is a greater tendency to drift and a smaller range to adjust with V2.
One particular thing I noticed now and in the past is that all adjustments affect each other.
V2 takes care of absolute DC as well as differential DC be it of a smaller magnitude.
Adjusting V1/V3 also has influence on DC in some way. Up or lower the bias with say 0.5A and a new adjustment has to be made to the offset. I tend to believe there’s a sweet spot somewhere.
The amp is now rather stable, knowing it’s a full open design. Literally every breath I take has effect on DC values.
Because the amp is now fully warmed up I’ll measure again tomorrow at first startup and report.
Schematic in link, notice also that the ccs is the one with the TL431. Feedback resistors have also been changed to 150k.
http://www.diyaudio.com/forums/showthread.php?postid=222538#post222538
/Hugo
One thing I found out doing simulations with the Aleph-X is at different DC biases the absolute DC offset responds to different values of the Rs from output to diff-pair source. At 3A bias on each side of the balanced output, 2.5k resistors yield about 20mV of offset at the +/- output. At 4A bias 3.3k works best at about 3mV offset at each output. These were with or without resistors to ground at each output. I suggest using 5k trimpots to adjust the DC offsets. Easier than substituting in and out different R values.
Of course, simulations does not take into account heat variations and other factors but it's worth giving it a try.
Of course, simulations does not take into account heat variations and other factors but it's worth giving it a try.
Nelson Pass said:Within a lot code, matching Vgs is perfectly adequate.
The bottom half are common-Source connection. The
top half would probably be considered an unusual
application of Common Drain.
Thanks for your help Nelson. You will recall that all the mosfets I purchase have the same date code. I completed the mosfet matching with fairly good results and was able to get many pairs to match within 0.002 volts at current values of 10mA, 100mA and 200mA. I don't know if anyone else has noticed, but each mosfet has a part number stamped on it which I discovered after measuring and matching them. Well, it turns out that the ones which have consecutive part numbers (ie; 5J and 5k) match almost perfectly in most, but not all, cases. You still have to go through the matching process to be sure.
Regarding the DC offset problem that most are experiencing, it appears this is thermal related. If one adjusts the DC offset to be minimum at operating temperature, at start-up , it is way off. Just a thought, has anyone ever tried using a thermister in parallel or in series with R46 or R47. For a negative gradient thermister, the resistance decreases with increasing temperature at a well defined relation for each type of thermister. If a thermister were selected that could match the DC offset R adjustment required, this "thought" should work. In short, it would effectivly serve as an automatic trim pot controlled by temperature. Again, just a thought.
If this doesn't work, I will resort to a microcontroller based DC correction tool which would select various resistors in parallel with R46 and R47. The key here is to sample the DC offset an make adjustments at a time when there is a lull in the music ie: no input signal. This condition occurs frequently even when playing music. This will work but it is a lot more complicated that the thermister option.
Netlist said:
I had a pretty hard time to re-adjust both DC settings and finally ended with adding the output R's to ground. 43ohm is what I had at hand. I can tell for sure that it is possible to get both values well within specs being <1V differential DC and <100mV absolute DC.
In fact both values are now close to zero after some hours playing. When cold, things are worse but only take half an hour to meet the specs.
Hope this helps.
/Hugo
Wuffwaff,
I have to apologize for not being up to speed yet on the Aleph-X. My son bought the PCBs last year but decided not to build it because he heard of the "DC offset" problem inherent to the design. I said what the heck and am about half way to finishing the amp.
I have found nothing in the wiki or the elsewhere describing "relative DC offset". What is it, what does it affect and what is done to get rid of it, if indeed it is a problem. The Wiki only discusses the Absolute DC offset adjustment and I have not come accross it in the posts. I have not read all the posts as yet and doubt that I will ever have the time to.
I have to apologize for not being up to speed yet on the Aleph-X. My son bought the PCBs last year but decided not to build it because he heard of the "DC offset" problem inherent to the design. I said what the heck and am about half way to finishing the amp.
I have found nothing in the wiki or the elsewhere describing "relative DC offset". What is it, what does it affect and what is done to get rid of it, if indeed it is a problem. The Wiki only discusses the Absolute DC offset adjustment and I have not come accross it in the posts. I have not read all the posts as yet and doubt that I will ever have the time to.
I played around with the zener and ccs.The part that is most temperature sensitive and affects abs dc offset is - believe it or not- the zener.Then the ccs mosfet and last the diff pair.Even just putting a coldish screwdriver on the zener would affect offset values quite quickly by hundreds of mv.The ccs is also quite sensitive.
In my highly biased ccs the amp begins at 10v offset drops after 5-10 min to around 4v and then very slowly until it reaches final temp stability after two hours to quite a few hundred mv - I gave up trying to get it lower.
What I see happening is that there are two variables one is the temp stability /working temp of the ccs which is reached quite quickly in the first 5-10 min and then the temp inside the case which affects all other parts but which takes very long to stabilise because of the large external heatsinks.
Because of my higher bias I believe my problems are multiplied by a factor of two or three.There is not much you can do with the ccs mosfet however I was thinking of changing the zener for another biasing arrangement that is less temp sensitive.I still don´t understand why the zener affects the mosfet so much.Normally the resistors biasing the mosfet at the source should be the most sensitive part.At 9v bias the mosfet should be fully switched on so a few mv here and there at the gate shouldn´t affect it .
In any case I don´t know if it´s worth the trouble in the end.The amps sound really fine and a a few hundred mv abs offset is not going to affect anything but a very small percentage of rail swing in terms of operation.
You just have to switch it on for a few minutes before you put it in gear.However if I am around the house I just leave them on more or less permanently.
In my highly biased ccs the amp begins at 10v offset drops after 5-10 min to around 4v and then very slowly until it reaches final temp stability after two hours to quite a few hundred mv - I gave up trying to get it lower.
What I see happening is that there are two variables one is the temp stability /working temp of the ccs which is reached quite quickly in the first 5-10 min and then the temp inside the case which affects all other parts but which takes very long to stabilise because of the large external heatsinks.
Because of my higher bias I believe my problems are multiplied by a factor of two or three.There is not much you can do with the ccs mosfet however I was thinking of changing the zener for another biasing arrangement that is less temp sensitive.I still don´t understand why the zener affects the mosfet so much.Normally the resistors biasing the mosfet at the source should be the most sensitive part.At 9v bias the mosfet should be fully switched on so a few mv here and there at the gate shouldn´t affect it .
In any case I don´t know if it´s worth the trouble in the end.The amps sound really fine and a a few hundred mv abs offset is not going to affect anything but a very small percentage of rail swing in terms of operation.
You just have to switch it on for a few minutes before you put it in gear.However if I am around the house I just leave them on more or less permanently.