I'll have a bunch of questions to post soon. I've got my two boards almost fully populated but I'm waiting for the main toroids to arrive and a couple other things. Plus I'm right in the middle of final exams right now so things are a little slow with the hobby.
I've got the same concern. I'm wondering whether or not I should buy a variac off of ebay just in case. I don't want to be blowing 044's left and right everytime I flip the switch on.
I've got the same concern. I'm wondering whether or not I should buy a variac off of ebay just in case. I don't want to be blowing 044's left and right everytime I flip the switch on.
A big bad variac is nice to have around. One morning I woke up to go to work and my truck battery was dead. I left my int lights on. I used my variac, DVM, Aleph 2 toriod and rectifiers ( 45v ) and made an instant battery booster and had it started in 15 minutes. My car was behind my wifes, so I couldn't just jump it.
But seriously, My 1200 va variac was only $45.00 on Ebay and I use it a lot more than I thought I would. By switching around the taps, you can give an old amp the real acid test and juice it up to 160V an see if it is a robust design or not.
But seriously, My 1200 va variac was only $45.00 on Ebay and I use it a lot more than I thought I would. By switching around the taps, you can give an old amp the real acid test and juice it up to 160V an see if it is a robust design or not.
Brian,
Your so right, you can't beat variac to bring a repaired or new amp to life. I saw that www.jamesco.com has 2kw variac for around 100.00.
Your so right, you can't beat variac to bring a repaired or new amp to life. I saw that www.jamesco.com has 2kw variac for around 100.00.
Q3 and Q8 should not be turning on at all, unless there is sufficient current through R5/R40 to bring their Vbe above 0.65V. At any rate, any current passing through this portion of the circuit will be quite small, and will be sent to ground through the relatively low impedance of R1||R4 and R44||R45. This should not pull the gate voltage of the input pair much if at all, so I suspect you have a different issue. Perhaps a floating gate on Q6?protos said:
BTW... if your power supply is healthy, there is no way that the rail voltages should be 'pulled' with repect to ground by the loading of the amplifier circuit (especially sans power transistors!). Your power supply should deliver a rock solid +V/GND/-V, with +V and -V being equal. If not, you're going to have real trouble when you install those power transistors!
Check the resistance of R20 with a DMM (while the circuit is off, of course!), and make sure there's a solid connection to the gate of Q6. Also check your power supply voltages to make sure they're stable. If you're using a test supply setup (eg. with a virtual ground like the one I recommended for koolscooby) that can't source ground current, and you also use J1a, you may very well experience a problem with the virtual ground voltage wandering up toward the positive rail. In this case, the ccs will shut off.
My theory did check out actually.When I removed r2/r3/r42/r43 the ccs and diff pair switched on normally.
Although I do have a floating ground from a 30v bench supply by using a couple of splitting resistors and caps set-up.The slow charging up of the c1 c6 to .6v turn on q3/q8 is quite obvious.
Now the other problem is a high DC offset one well matched pair gave me 260mv and a another one 90mv which is still high at this stage.The macmillan resistors do not seem to help much in this case.The thing I don't understand is why VR2 would help dc offset values.What it does is adjust the current of the ccs and the gain of the diff pair but this does nothing to correct any mismatch between q5/q7.Of course changing the gain will impinge on the relative offset values but not that much since the mismatch will always be there.
What we need is to balance the voltage drop over r33/r25 so that it is exactly equal.This can only be done accurately by attaching an adjustment pot to them much in the same way it is done at vr2.This way you could zero out any offset at this stage or probably even correct for any (small) offset at the output stage , in fact you could probably chuck out all those matching considerations and expenses.
Probably getting carried away here but this is my simplistic theory.Anyone care to comment?
Although I do have a floating ground from a 30v bench supply by using a couple of splitting resistors and caps set-up.The slow charging up of the c1 c6 to .6v turn on q3/q8 is quite obvious.
Now the other problem is a high DC offset one well matched pair gave me 260mv and a another one 90mv which is still high at this stage.The macmillan resistors do not seem to help much in this case.The thing I don't understand is why VR2 would help dc offset values.What it does is adjust the current of the ccs and the gain of the diff pair but this does nothing to correct any mismatch between q5/q7.Of course changing the gain will impinge on the relative offset values but not that much since the mismatch will always be there.
What we need is to balance the voltage drop over r33/r25 so that it is exactly equal.This can only be done accurately by attaching an adjustment pot to them much in the same way it is done at vr2.This way you could zero out any offset at this stage or probably even correct for any (small) offset at the output stage , in fact you could probably chuck out all those matching considerations and expenses.
Probably getting carried away here but this is my simplistic theory.Anyone care to comment?
Are you using 4.7K at R11/R33? If you've got 47.5K in there like the stock schematic, then Q3/Q8 should never turn on. With 4.7K in there, you can just turn up VR1/VR3 to prevent the Q's from turning on.
There are actually two different DC offsets to adjust. The first is absolute (aka common-mode) DC offset, which is how close the outputs sit relative to GND. This doesn't really affect much, except that you won't be able to get full power out of the amp and it will clip asymmetrically. CM DC offset will also cause a DC current at the inputs if they are DC coupled. The MacMillan resistors and VR2 are meant to adjust this parameter. VR2 works to adjust the current through Q5/Q7, thus increasing or decreasing the voltage across both R23 and R25, which in turn causes Q2 and Q11 to conduct more or less current and the outputs will both float up or down in voltage depending on how much DC current flows through R1/4/44/45 (ideally zero).
Then we have our differential DC offset, which will cause a voltage to appear across the speaker. This is comparable to the DC offset you'd have on a conventional amplifier, and is what you adjust by matching transistors. Adding trimmers to R23/R25 will cause an imbalance in the gain of the left and right sides of the amp. You're welcome to try it, of course, but this amp is all about symmetry, so matching is definitely the preferred way to ensure optimal AC performance.
There are actually two different DC offsets to adjust. The first is absolute (aka common-mode) DC offset, which is how close the outputs sit relative to GND. This doesn't really affect much, except that you won't be able to get full power out of the amp and it will clip asymmetrically. CM DC offset will also cause a DC current at the inputs if they are DC coupled. The MacMillan resistors and VR2 are meant to adjust this parameter. VR2 works to adjust the current through Q5/Q7, thus increasing or decreasing the voltage across both R23 and R25, which in turn causes Q2 and Q11 to conduct more or less current and the outputs will both float up or down in voltage depending on how much DC current flows through R1/4/44/45 (ideally zero).
Then we have our differential DC offset, which will cause a voltage to appear across the speaker. This is comparable to the DC offset you'd have on a conventional amplifier, and is what you adjust by matching transistors. Adding trimmers to R23/R25 will cause an imbalance in the gain of the left and right sides of the amp. You're welcome to try it, of course, but this amp is all about symmetry, so matching is definitely the preferred way to ensure optimal AC performance.
Chad,
Your analysis is of course correct.What I suggested i.e. the use of a pot to match differential offset which is what the speaker sees is not going to affect gain that much the way I see it and in a way that will have extremely small consequences in terms of sonic differences or distortion etc.The 1or 2% adjustment that you would need at r33/r25 would affect the gain by the same amount but do you really think that if the amp outputs 20w or 20.4W you could ever hear the difference?.Ok so maybe some of the SUSY distortion cancellation could suffer so instead of .034% we will have .035% - again its a silly difference and only one that a very good distortion analyzer will be able to detect.
In any case 2% is within many components variability ratings that the majority are quite happy to live with particularly with power amps.Are you even sure that your preamp LR channels/volume control etc are better matched than 2%?
So what I am saying is that the adjustment option would solve a lot of headaches for many Aleph X builders that don't manage to get a low DC offset with a really minimal compromise in unimportant parameters.Of course since this would be an option the super matcher purists would have their cake too and just stick with a single resistor and not use the parallel adjustment circuit.
Perhaps a thought for the REV.2 boards?
Your analysis is of course correct.What I suggested i.e. the use of a pot to match differential offset which is what the speaker sees is not going to affect gain that much the way I see it and in a way that will have extremely small consequences in terms of sonic differences or distortion etc.The 1or 2% adjustment that you would need at r33/r25 would affect the gain by the same amount but do you really think that if the amp outputs 20w or 20.4W you could ever hear the difference?.Ok so maybe some of the SUSY distortion cancellation could suffer so instead of .034% we will have .035% - again its a silly difference and only one that a very good distortion analyzer will be able to detect.
In any case 2% is within many components variability ratings that the majority are quite happy to live with particularly with power amps.Are you even sure that your preamp LR channels/volume control etc are better matched than 2%?
So what I am saying is that the adjustment option would solve a lot of headaches for many Aleph X builders that don't manage to get a low DC offset with a really minimal compromise in unimportant parameters.Of course since this would be an option the super matcher purists would have their cake too and just stick with a single resistor and not use the parallel adjustment circuit.
Perhaps a thought for the REV.2 boards?
protos,
I really don't mean to be argumentative and I hope you don't take it the wrong way... the following is all well intentioned dialogue, and I think you raise some good points of discussion which can help other members too, especially in developing a more complete understanding of this circuit's intricacies, and idiosyncrasies. The Aleph-X is quite an unusual and unique beast.
There is an additional issue in play here, which I failed to mention in my previous post: You would agree that the differential DC offset is primarily caused by the input pair's Vgs mismatch, yes? A little examination shows that this Vgs mismatch is outside of the main feedback loop, and is therefore not directly subject to any error-correction characteristics of the circuit. Hence the need for precise matching of these two transistors. The same applies to conventional amplifier topologies as well, which suffer this problem somewhat less due to the lower Vbe of BJTs typically employed.
However, any effort to compensate the differential DC offset by fiddling with something inside the feedback loop (such as R23/R25) will be blunted by the negative feedback attempting to un-do what you've just done. The mismatch in Vgs, which is the root of the problem, still exists. The best you can hope for by trimming R23/R25 is to have a large enough effect on the currents through Q5/Q7 to move their bias points until the Vgs matches. This implies a fairly significant bias current mismatch, and I'm pretty sure that unless you're correcting a very small Vgs mismatch, this will require large adjustments to R23/R25, likely exceeding 2%. Affecting the gain of each side will only be a second-order effect. In the meantime, we've lost a fair amount of the symmetry and AC performance of the circuit, which I think you'll agree isn't a reasonable trade-off. After all, it's really the AC performance we care about most, right? Therefore, I would only recommend small adjustments to R23/R25 as a final step to take out the last bits of DC offset. Personally, I wouldn't mess with them at all.
So, what can we do that will be effective in reducing the DC offsets? Well, unfortunately, with a very simple and elegant design like this, not too much. Your first and best line of defense is good ol' trasistor matching. I know, I know, it sucks to buy more transistors than you need, and I feel your pain too, but alas if you want that chocolate cake, you'll first have to choke down those last two cold brussel sprouts!
In the eternal words of Pink Floyd "How can you have any pudding if you don't eat your meat?" 
Hehe. OK seriously... there are still a few little things we can do to help with the DC offset. Mainly, what we need to do is work with parameters outside the feedback loop so that our efforts are effective. The first thing you can do is AC couple the inputs with some nice capacitors. Second, you can try AC coupling R19/R29 to GND. I'm not sure what this will do to the absolute DC offset, probably not much. AC coupling the inputs should reduce the DC offset by a factor of 2, and adding caps in series with R19/R29 by another factor of 11 or so (depending on the value of R16/R30)... all without affecting the symmetry of the amp. Of course, you'll want to use good, big caps for this to keep the rolloff point very low (most important with R19/R29).Another possibility which is sometimes seen is to employ a potentiometer for source degeneration of the input pair. You can then adjust the wiper to trim the resistance at the source of Q5 vs. Q7. This has it's own drawbacks as well, and I don't really recommend it for this amp. Another approach would be the use of DC servo circuits, though this goes somewhat against the principle of simplicity of the Pass designs and also adds a layer of difficulty for the hobbyist to deal with. I've been playing with this idea myself, but again, it's not an approach I would recommend to everyone at this point, unless you are confident that you can design a sonically neutral servo circuit which will function well.
The Aleph-X is already near the upper limit of complexity and sophistication amongst the DIY Pass designs and derivatives, and it walks a fine line in terms of performance vs. complexity. I think Grey has already done an excellent job in terms of design choices, while remaining true to the Pass tradition of simplicity. As with any design, one must be prepared to accept the Aleph-X's shortcomings, including the need for matched components.
I really don't mean to be argumentative and I hope you don't take it the wrong way... the following is all well intentioned dialogue, and I think you raise some good points of discussion which can help other members too, especially in developing a more complete understanding of this circuit's intricacies, and idiosyncrasies. The Aleph-X is quite an unusual and unique beast.
There is an additional issue in play here, which I failed to mention in my previous post: You would agree that the differential DC offset is primarily caused by the input pair's Vgs mismatch, yes? A little examination shows that this Vgs mismatch is outside of the main feedback loop, and is therefore not directly subject to any error-correction characteristics of the circuit. Hence the need for precise matching of these two transistors. The same applies to conventional amplifier topologies as well, which suffer this problem somewhat less due to the lower Vbe of BJTs typically employed.
However, any effort to compensate the differential DC offset by fiddling with something inside the feedback loop (such as R23/R25) will be blunted by the negative feedback attempting to un-do what you've just done. The mismatch in Vgs, which is the root of the problem, still exists. The best you can hope for by trimming R23/R25 is to have a large enough effect on the currents through Q5/Q7 to move their bias points until the Vgs matches. This implies a fairly significant bias current mismatch, and I'm pretty sure that unless you're correcting a very small Vgs mismatch, this will require large adjustments to R23/R25, likely exceeding 2%. Affecting the gain of each side will only be a second-order effect. In the meantime, we've lost a fair amount of the symmetry and AC performance of the circuit, which I think you'll agree isn't a reasonable trade-off. After all, it's really the AC performance we care about most, right? Therefore, I would only recommend small adjustments to R23/R25 as a final step to take out the last bits of DC offset. Personally, I wouldn't mess with them at all.
So, what can we do that will be effective in reducing the DC offsets? Well, unfortunately, with a very simple and elegant design like this, not too much. Your first and best line of defense is good ol' trasistor matching. I know, I know, it sucks to buy more transistors than you need, and I feel your pain too, but alas if you want that chocolate cake, you'll first have to choke down those last two cold brussel sprouts!
In the eternal words of Pink Floyd "How can you have any pudding if you don't eat your meat?" Hehe. OK seriously... there are still a few little things we can do to help with the DC offset. Mainly, what we need to do is work with parameters outside the feedback loop so that our efforts are effective. The first thing you can do is AC couple the inputs with some nice capacitors. Second, you can try AC coupling R19/R29 to GND. I'm not sure what this will do to the absolute DC offset, probably not much. AC coupling the inputs should reduce the DC offset by a factor of 2, and adding caps in series with R19/R29 by another factor of 11 or so (depending on the value of R16/R30)... all without affecting the symmetry of the amp. Of course, you'll want to use good, big caps for this to keep the rolloff point very low (most important with R19/R29).Another possibility which is sometimes seen is to employ a potentiometer for source degeneration of the input pair. You can then adjust the wiper to trim the resistance at the source of Q5 vs. Q7. This has it's own drawbacks as well, and I don't really recommend it for this amp. Another approach would be the use of DC servo circuits, though this goes somewhat against the principle of simplicity of the Pass designs and also adds a layer of difficulty for the hobbyist to deal with. I've been playing with this idea myself, but again, it's not an approach I would recommend to everyone at this point, unless you are confident that you can design a sonically neutral servo circuit which will function well.
The Aleph-X is already near the upper limit of complexity and sophistication amongst the DIY Pass designs and derivatives, and it walks a fine line in terms of performance vs. complexity. I think Grey has already done an excellent job in terms of design choices, while remaining true to the Pass tradition of simplicity. As with any design, one must be prepared to accept the Aleph-X's shortcomings, including the need for matched components.
jwb - you mean for absolute DC offset? I think I explained it a few posts back. I guess altering the overall diff pair current might have a small effect on the differential offset. But for better differential trim, I guess you could put small adjustable CCS's in parallel with R23/R25... hadn't thought about it much, but could be something worth further consideration.
Hi Chad,
I bow to your extensive knowledge and experience.Nice ideas for the control of offset.
Perhaps you can help me out with a new problem I am having and with which I am bit stuck at this stage.
When I power up the amp and one of the two sides pos or neg -offboard power transistors ( ccs and output)- is not connected then the connected side is working properly and measuring correctly.i.e. q1/10 gd 8.2v gs 4.5v ds 12.7v , q2/11 gd 10v gs 4.5v ds 15v with about one amp i.e. .49V over .47ohms.
The moment I connect the other side and vice versa then the whole thing goes haywire with something like q1/10 gd 0v gs 4.5v ds 4v and q2/q11 gd 20v gs 4.5v ds 25v and there is no current conduction.
Normally I would know which part of the circuit to check but in this case I am going about in circles because my reasoning is that if each side is working properly in isolation then the input diff pair is working and all the other components and devices are connected correctly and doing their job.But something happens when both + and- amps are working at the same time that makes the circuit fail.There is nothing connected at the input or output.
I bow to your extensive knowledge and experience.Nice ideas for the control of offset.
Perhaps you can help me out with a new problem I am having and with which I am bit stuck at this stage.
When I power up the amp and one of the two sides pos or neg -offboard power transistors ( ccs and output)- is not connected then the connected side is working properly and measuring correctly.i.e. q1/10 gd 8.2v gs 4.5v ds 12.7v , q2/11 gd 10v gs 4.5v ds 15v with about one amp i.e. .49V over .47ohms.
The moment I connect the other side and vice versa then the whole thing goes haywire with something like q1/10 gd 0v gs 4.5v ds 4v and q2/q11 gd 20v gs 4.5v ds 25v and there is no current conduction.
Normally I would know which part of the circuit to check but in this case I am going about in circles because my reasoning is that if each side is working properly in isolation then the input diff pair is working and all the other components and devices are connected correctly and doing their job.But something happens when both + and- amps are working at the same time that makes the circuit fail.There is nothing connected at the input or output.
Just to give the reader an idea of the matching necessary for this amp, I matched the input diff pair to within 1mv at two different temperatures, using a programmable CCS to make certain to measure the Vgs at the desired drain current. Then I matched the 392Ω resistors (R16/17 hifiZen, R23/25 GRollins) to within 0.1%, using the same current source. The result was about 20mV differential offset at the output with the amp warmed up and the inputs connected together.
The current in the diff pair seems to have a pretty strong effect on the differential offset. The lowest differential offset is achieved at the lowest offset to ground. With any offset to ground the differential offset can be pretty high. I don't know what the standard here is but I like to keep the differential offset under 100mV at all practical operating temperatures.
I got an improved result by trimming V1 and V3 for exactly the same idle current on each side.
I think the idea of an oven for the differential pair is a rather good one. It removes one thermal transient from the system. Maybe putting the diff pair on an aluminum slab with a 1W heater is worth the trouble. I'd really like the DC situation to be sorted within 10 seconds of powerup.
The current in the diff pair seems to have a pretty strong effect on the differential offset. The lowest differential offset is achieved at the lowest offset to ground. With any offset to ground the differential offset can be pretty high. I don't know what the standard here is but I like to keep the differential offset under 100mV at all practical operating temperatures.
I got an improved result by trimming V1 and V3 for exactly the same idle current on each side.
I think the idea of an oven for the differential pair is a rather good one. It removes one thermal transient from the system. Maybe putting the diff pair on an aluminum slab with a 1W heater is worth the trouble. I'd really like the DC situation to be sorted within 10 seconds of powerup.
Further to my previous post I think the problem was mainly one of lack of adjustment.When powering up both sides it seems to affect absolute offset values so that r1/4/44/45 were dropping something like 10v and were affecting the operation.
This of course can be adjusted by vr2.I haven't had time to try and adjust everything yet.
However I am still trying to understand why the offset values are very low (mv)when only one side is operating and they go sky high when the other side is on.
This of course can be adjusted by vr2.I haven't had time to try and adjust everything yet.
However I am still trying to understand why the offset values are very low (mv)when only one side is operating and they go sky high when the other side is on.
Ok the one channel is up and running correctly.However I am really surprised how finicky absolute DC offset is.When I adjust vr2 to 0 with the amp warm(in fact it swings a lot over a 150mv range within say 20 sec) then I find after half an hour it is up to 3v! I realize that the offset changes whether I have the lid on the amp closed or not with about 3v difference between lid closed or open -wow!So what I do is adjust the dc offset with the lid open at -3v so that it will eventually drift closer to zero.
Now the closest I am getting is around 200-300mv absolute offset with less than 20mv relative offset.
Now that is what I call a temperature sensitive circuit!(perhaps my higher biased front stage also has something to do with it).
As to the sound comparison of one monblock AX to Aleph5:I usually connect both speakers on one channel for a kind of mono operation.
It seems the AX is not just better on points but is a whole level or two better.Bass much heftier ,mids are similar , treble separation and extension much better.Also even in mono there is more apparent depth and separation of instruments.
Now the closest I am getting is around 200-300mv absolute offset with less than 20mv relative offset.
Now that is what I call a temperature sensitive circuit!(perhaps my higher biased front stage also has something to do with it).
As to the sound comparison of one monblock AX to Aleph5:I usually connect both speakers on one channel for a kind of mono operation.
It seems the AX is not just better on points but is a whole level or two better.Bass much heftier ,mids are similar , treble separation and extension much better.Also even in mono there is more apparent depth and separation of instruments.
Aleph X (low vs. high power)
Hi,
I got the board from Hifizen. The original config. is too low powered for me.
I need 50 W at 4 Ohm load. I consider 15v rails and increasing bias to 5-6 A.
I plan to use 4 Fets per Output/Current Source. This leads to 0.33 Ohm Source Resistors.
R42/43 need to be changed to 0.33 Ohm as well. R 44 / 45 are then 250 Ohm each.
am i right ? What is your opinion on Powerrating of both source, R43/43 and R44/45 ?
Thanks for reply.
Best Regards
Frederik
Hi,
I got the board from Hifizen. The original config. is too low powered for me.
I need 50 W at 4 Ohm load. I consider 15v rails and increasing bias to 5-6 A.
I plan to use 4 Fets per Output/Current Source. This leads to 0.33 Ohm Source Resistors.
R42/43 need to be changed to 0.33 Ohm as well. R 44 / 45 are then 250 Ohm each.
am i right ? What is your opinion on Powerrating of both source, R43/43 and R44/45 ?
Thanks for reply.
Best Regards
Frederik
You could also manage with 3,5A , 66% ac current gain and 26w pd with only 4 fets giving 53W into 4ohms.
I still would like some more information about ac current gain and its limits re max current and any sonic compromises.I think even Nelson suggests going up to 66% for low impedance loads.But what happens at 75 or even 90%?Chads spreadsheet suggests dramatic increases in current available , what's the catch though?
I still would like some more information about ac current gain and its limits re max current and any sonic compromises.I think even Nelson suggests going up to 66% for low impedance loads.But what happens at 75 or even 90%?Chads spreadsheet suggests dramatic increases in current available , what's the catch though?
Protos: I was also pretty surprised how mobile the absolute DC offset can be. You can make the adjustment less sensitive to the pot by using a smaller pot and a bigger fixed resistor. But the DC swing is caused by two things: self-heating of Q6 (referring to hifiZen schematic) and self-heating of Q2/Q11. To solve the Q6 problem you need to bias the gate of Q6 from an opamp that holds V(R24) constant. To solve the both problems simultaneously you need to drive the gate of Q6 from an opamp which holds the DC output at ground.
Hi jwb
Mobile is an understatement in my case.I thought yesterday I had adjusted it at a reasonable level.Imagine my surprise when I switched on today the offset went up to 9v and only very very slowly over the course of an hour or two did it stabilise to under 1v.At 9v I am sure the amp is not working properly.
This is a very finicky amp.I could live with several hundred mv offset but swings of several volts is way too much.
Are other people experiencing these wide variations or am I the exception?
Mobile is an understatement in my case.I thought yesterday I had adjusted it at a reasonable level.Imagine my surprise when I switched on today the offset went up to 9v and only very very slowly over the course of an hour or two did it stabilise to under 1v.At 9v I am sure the amp is not working properly.
This is a very finicky amp.I could live with several hundred mv offset but swings of several volts is way too much.
Are other people experiencing these wide variations or am I the exception?
http://www.colomar.com/Shavano/inductor_info.html
Dont go lower than 1.9mm magnet wire (good for approx 600W) as it will get quite hot.
Magura
Dont go lower than 1.9mm magnet wire (good for approx 600W) as it will get quite hot.
Magura