Can you get a close-up of the diff pair?Hi Ian,
William offers good advice as always. He's one of the most experienced people around regarding JFET front ends on Aleph-X's.
Regarding your questions:
1) William addresses this well. It's also the reason why I picked the BL parts. I don't know if Nelson or any other manufacturers Idss grade their JFETs within the Toshiba color grades like BL for instance or whether they they design their circuits so that they can just take parts out of the bag. I really doubt that they hand select front end JFET's for sweet spot performance on an amplifier by amplifier basis. I am just taking J109BL's out of the bag.
2) The 10p cap was originally put in because it was there on the original Aleph-X. I have tried no cap and I have tried twisted wires (in fact that's what's in the pictures at the beginning of the thread if your eyes are good enough to see it). I couldn't hear any difference so I've deleted these parts.
Thank you for the photo. A very professional looking job. I deliberately used scatter wiring to connect the driver board to the output section to avoid potential coupling problems.
The Aleph-X, either the homebrew or commercial version, is an exquisite balance of bare sufficiency and excess. On the excess side it is a high power class A circuit with excess current and heat compared to most power amplifiers. On the sufficiency side the signal passes through only two transistors between the input and output, and there are very few parts that directly touch the signal. The input stage design is carefully balanced against the output stage design. So much bias current here vs. so much input cap there vs. so much voltage swing etc, etc. The whole amplifier circuit is a collective sweet spot. Change almost anything and you risk leaving that place.
OK. That being said, I figure that if you want to change the input section you need to make it look just like the old one to the rest of the circuit. That means 1) driving the output stage with the same voltage and current parameters, and 2) changing the input drive requirements as little as possible. That's where the parallel J109BL idea came from. It provides most or all of the requirements of this argument AFAIK. I'm not sweating the transconductance thing for now because the Aleph-X has two gain stages, not one like a UGS or a ZV9 etc.
I don't know if the AX100J circuit will work. There's only one way to find out and that's to try it. Although Nelsons comment above is encouraging, he really only refers only to a parallel JFET implementation being better than the MOSFET implementation. He doesn't actually refer to this specific implementation. Something could still be amiss.
Cheers,
Graeme
William offers good advice as always. He's one of the most experienced people around regarding JFET front ends on Aleph-X's.
Regarding your questions:
1) William addresses this well. It's also the reason why I picked the BL parts. I don't know if Nelson or any other manufacturers Idss grade their JFETs within the Toshiba color grades like BL for instance or whether they they design their circuits so that they can just take parts out of the bag. I really doubt that they hand select front end JFET's for sweet spot performance on an amplifier by amplifier basis. I am just taking J109BL's out of the bag.
2) The 10p cap was originally put in because it was there on the original Aleph-X. I have tried no cap and I have tried twisted wires (in fact that's what's in the pictures at the beginning of the thread if your eyes are good enough to see it). I couldn't hear any difference so I've deleted these parts.
Thank you for the photo. A very professional looking job. I deliberately used scatter wiring to connect the driver board to the output section to avoid potential coupling problems.
The Aleph-X, either the homebrew or commercial version, is an exquisite balance of bare sufficiency and excess. On the excess side it is a high power class A circuit with excess current and heat compared to most power amplifiers. On the sufficiency side the signal passes through only two transistors between the input and output, and there are very few parts that directly touch the signal. The input stage design is carefully balanced against the output stage design. So much bias current here vs. so much input cap there vs. so much voltage swing etc, etc. The whole amplifier circuit is a collective sweet spot. Change almost anything and you risk leaving that place.
OK. That being said, I figure that if you want to change the input section you need to make it look just like the old one to the rest of the circuit. That means 1) driving the output stage with the same voltage and current parameters, and 2) changing the input drive requirements as little as possible. That's where the parallel J109BL idea came from. It provides most or all of the requirements of this argument AFAIK. I'm not sweating the transconductance thing for now because the Aleph-X has two gain stages, not one like a UGS or a ZV9 etc.
I don't know if the AX100J circuit will work. There's only one way to find out and that's to try it. Although Nelsons comment above is encouraging, he really only refers only to a parallel JFET implementation being better than the MOSFET implementation. He doesn't actually refer to this specific implementation. Something could still be amiss.
Cheers,
Graeme
> The whole amplifier circuit is a collective sweet spot. Change almost anything and you risk leaving that place.
But then there are XA-30, XA-60, XA-100, XA-160, XA-200, all with different number of output devices, different open loop gains and bandwidths, different voltages & biases on the output devices, .....
So which one is the sweet spot, or is the sweet spot large enough that you can still change a few things ??
> I'm not sweating the transconductance thing for now because the Aleph-X has two gain stages, not one like a UGS or a ZV9 etc.
That I do not understand. Would you care to explain ?
What do you mean by sweating the transconductance thing ?
> I couldn't hear any difference so I've deleted these parts.
I wonder whether you see any difference in your 10kHz square wave response......
Patrick
But then there are XA-30, XA-60, XA-100, XA-160, XA-200, all with different number of output devices, different open loop gains and bandwidths, different voltages & biases on the output devices, .....
So which one is the sweet spot, or is the sweet spot large enough that you can still change a few things ??
> I'm not sweating the transconductance thing for now because the Aleph-X has two gain stages, not one like a UGS or a ZV9 etc.
That I do not understand. Would you care to explain ?
What do you mean by sweating the transconductance thing ?
> I couldn't hear any difference so I've deleted these parts.
I wonder whether you see any difference in your 10kHz square wave response......
Patrick
Hi EUVL,
Exactly the point. Each of those XA amps has it's own collective sweet spot.
I was saying that I believe that there is enough transconductance in the parallel JFET input stage shown to do the job here.
Yes I have seen the the very small difference. But I can't hear the difference so I'm deleting them. Others have said the same.
GL
Exactly the point. Each of those XA amps has it's own collective sweet spot.
I was saying that I believe that there is enough transconductance in the parallel JFET input stage shown to do the job here.
Yes I have seen the the very small difference. But I can't hear the difference so I'm deleting them. Others have said the same.
GL
In terms of tweaking I have found the 220uf capacitors in the current share loop critical for that special result.
This was discussed ages ago but I tried tonight Panasonics FC 220 uf 63 volt with and without Panasonic PP film 0.0068 uf bypassed. Nice with the bypass if perhaps a bit prominant in the extreme top end. Without quite plain and uninteresting.
Tomorrow I will try bypass with 0.046 Panasonic PP film and see how that goes. I also have some Elna Silmic and some KZ Gold Nichicon Muse to try.
I know some people hav tried BG series but they never really captured my imagination.
macka
Macka
This was discussed ages ago but I tried tonight Panasonics FC 220 uf 63 volt with and without Panasonic PP film 0.0068 uf bypassed. Nice with the bypass if perhaps a bit prominant in the extreme top end. Without quite plain and uninteresting.
Tomorrow I will try bypass with 0.046 Panasonic PP film and see how that goes. I also have some Elna Silmic and some KZ Gold Nichicon Muse to try.
I know some people hav tried BG series but they never really captured my imagination.
macka
Macka
Hi Ian,
The Nelson Pass paper on DIY opamps on the PassDIY site has instructions for matching MOSFETs, JFETs, and BJT's.
I'm very interested in your experiments on bypassing those particular electrolytics. Like you say there has been lots of discussion on this in the past. But I don't recall any conclusions or even consensus. I also don't recall ever seeing any Pass Labs or First Watt schematic that showed these particular caps being bypassed either. However, there are Pass Labs schematics that show electrolytics that are in series with the signal being bypassed with small film caps.
Are you allowing time for the electrolytics to break in before adding the bypasses and are you listening before and after?
Graeme
The Nelson Pass paper on DIY opamps on the PassDIY site has instructions for matching MOSFETs, JFETs, and BJT's.
I'm very interested in your experiments on bypassing those particular electrolytics. Like you say there has been lots of discussion on this in the past. But I don't recall any conclusions or even consensus. I also don't recall ever seeing any Pass Labs or First Watt schematic that showed these particular caps being bypassed either. However, there are Pass Labs schematics that show electrolytics that are in series with the signal being bypassed with small film caps.
Are you allowing time for the electrolytics to break in before adding the bypasses and are you listening before and after?
Graeme
Graeme,
Yes it is interesting but I like to apply a little science rather than adhoc what sounds best. The caps have been used for some weeks now.
In that respect I think some diy efforts fail because people dont get the fundermentals right like feedback, gain and compensation.
Could you explain the rational behind used the 220K feedback resisters and 10 pf? Why not use 5pf for the same time constant as the standard Aleph uses 10 pf and 100K?
I know you decided on 26 db of gain but what is the amount of feeback? Less feedback means lower damping factor in my book. I calculated damping factor of around only 25 or so for the current design.
But I have to say so far I like the 220K feedback. In my earlier X Aleph efforts with 10 pf & 100K. It sounded a bit bright but was tight and punchy. Using the other compensation caps in other areas did not help. From what I know the susy is better with low feedback. The point is what is meant by low?
macka
Yes it is interesting but I like to apply a little science rather than adhoc what sounds best. The caps have been used for some weeks now.
In that respect I think some diy efforts fail because people dont get the fundermentals right like feedback, gain and compensation.
Could you explain the rational behind used the 220K feedback resisters and 10 pf? Why not use 5pf for the same time constant as the standard Aleph uses 10 pf and 100K?
I know you decided on 26 db of gain but what is the amount of feeback? Less feedback means lower damping factor in my book. I calculated damping factor of around only 25 or so for the current design.
But I have to say so far I like the 220K feedback. In my earlier X Aleph efforts with 10 pf & 100K. It sounded a bit bright but was tight and punchy. Using the other compensation caps in other areas did not help. From what I know the susy is better with low feedback. The point is what is meant by low?
macka
Hi Ian,
The 220K feedback resistors change the closed loop gain to 26dB. This is more useful to me in my system. Nelson has written on at least one occasion that about 10dB of SuSy feedback is a reasonable number, but I don't think there's a hard and fast rule. I don't remember his exact words.
The factory XA amps were set to 26dB of closed loop gain.
I don't know what the open loop gain is of the AX100 so I don't know how much SuSy feedback is being applied. From my experience listening to the unit and from comments made by you and by Dave about the sonics, I think we're safe. In other words I think that the AX100 has adequate open loop gain to provide 26dB of closed loop gain without problems.
Your earlier Aleph-X had a lot of circuit differences from the one you have now. Who's to say what's causing the sonic changes you're hearing? It's more likely to be things like the number and type of output transistor, or even the power supply, and not the main feedback resistors or compensation caps.
The 10pf compensation cap is not really a big deal. Many people have discussed this in the past. Nelson eventually eliminated it from some or all of the factory units because there was enough distributed cap in the internal wiring to do the job. I recall that he or someone else indicated that these caps were included to remove any potential ringing that might occur on square wave test signals. I recall the word "polish" being used to describe this. If I was going to submit the AX100 to a magazine to be reviewed, and I knew they were going to publish scope photos of square wave test signals I would definitely include the 10pf caps. But I'm not so I won't. I have tried 0pf, 5pf and 10 pf and I hear no difference.
By the way I am going to PM you a new schematic sometime in the next few days. I will post here to let you know to look for it.
Cheers,
Graeme
The 220K feedback resistors change the closed loop gain to 26dB. This is more useful to me in my system. Nelson has written on at least one occasion that about 10dB of SuSy feedback is a reasonable number, but I don't think there's a hard and fast rule. I don't remember his exact words.
The factory XA amps were set to 26dB of closed loop gain.
I don't know what the open loop gain is of the AX100 so I don't know how much SuSy feedback is being applied. From my experience listening to the unit and from comments made by you and by Dave about the sonics, I think we're safe. In other words I think that the AX100 has adequate open loop gain to provide 26dB of closed loop gain without problems.
Your earlier Aleph-X had a lot of circuit differences from the one you have now. Who's to say what's causing the sonic changes you're hearing? It's more likely to be things like the number and type of output transistor, or even the power supply, and not the main feedback resistors or compensation caps.
The 10pf compensation cap is not really a big deal. Many people have discussed this in the past. Nelson eventually eliminated it from some or all of the factory units because there was enough distributed cap in the internal wiring to do the job. I recall that he or someone else indicated that these caps were included to remove any potential ringing that might occur on square wave test signals. I recall the word "polish" being used to describe this. If I was going to submit the AX100 to a magazine to be reviewed, and I knew they were going to publish scope photos of square wave test signals I would definitely include the 10pf caps. But I'm not so I won't. I have tried 0pf, 5pf and 10 pf and I hear no difference.
By the way I am going to PM you a new schematic sometime in the next few days. I will post here to let you know to look for it.
Cheers,
Graeme
Thanks Graeme for the explanation.
No other new but on the Jfet matching.
I have looked at other posts and it would seem people have to order a large quanity because of variations with Jfets to do good matching.
I think this is counter productive because A/ cost and B/ it is wasteful of scare parts. I have therefore arranged delivery of matched devices with some spare sets. The cost is half that of buying 40 parts and hoping for a match and I will have 8 matched sets of four Jfets and some other parts.
Macka
It will probably take a week or two for this to arrange.
No other new but on the Jfet matching.
I have looked at other posts and it would seem people have to order a large quanity because of variations with Jfets to do good matching.
I think this is counter productive because A/ cost and B/ it is wasteful of scare parts. I have therefore arranged delivery of matched devices with some spare sets. The cost is half that of buying 40 parts and hoping for a match and I will have 8 matched sets of four Jfets and some other parts.
Macka
It will probably take a week or two for this to arrange.
aleph1.2 to AX100
GL and NP and Grey
I have had a chance to listen to my new XA120+ for a good length of time. The following are my observations as compared with the A75 and what I remember my Aleph 1.2's sounding like driving my line arrays from 2600hz and up. The Pass 350.5 drives the mid bass from 2600hz down to about 45hz where the line arrays are crossed over (in parallel) with a stereo set of Buggtussel transmission line subs.
1. the soundstage is immense, I have never heard it this big. Line arrays have this ability but the Pass amps really show it off.
2. Clarity, I may have to turn the hi frequencies down another .5db or so as the AX really does well with them and the drivers are super efficient.
3. Tonal balance is excellent but I can never be sure how good the master tape was recorded and then records produced. A live saxaphone always seems brighter on most recordings than live.
The system now comes a bit closer to this sense of warmth.
4. Very low distortion-I can't here any! With multiple drivers they don't work very hard.
5. Images are really life like with good height and with line arrays this doesn't change when you are standing.
Lessons learned:
Matching devices is very tedious and the out of circuit tests don't predict what the diff pair will do in circuit. Next time I put in sockets for the transistors and just plug them in and out and if when I get a good set I will then solder them in place.
I am frustrated with noisy transformers, I guess I will have to buck up for a really good audio/hospital grade type that is potted.
More and More heatsink-these run hot and you can only touch the heat sinks for a few seconds after they have been operating for 5 or six hours.
Due to the fact that I am using 24 output devices my class A current is over 6 amps vs the 5amps for GL's. I don't know if this is the reason my 1kva trannies are noisy. The transformer bolt is only attached to the chassis on one end so I am not getting any shorting from the metal hold down bolt.
I am not sure how to figure class A output watts for single ended devices but I get 48volts rms (pk-pk) on sine waves before clipping. I have to add the two curves on my dual trace scope.
With the add function on the scope I don't get a completely st line so I see that as distortion. I cannot adjust the gain pot to completely get rid of it. I am using the 10pf cap across the feedback resistor. I can never be sure what will set an amp oscillating and these do not. I probably couldn't hear the difference. I did bypass the PS caps including some on the boards.
Regarding going to J-fets, I will save that until GL gets finished and can give an update on the sound. I feel J-fets will be an improvement based on the fact that Audio Research uses them as input diff pairs on their very expensive tube amps.
I will listen more this weekend. Thanks to all and now my xformer search begins (maybe NP has extra B stock laying around)--chow, dave
GL and NP and Grey
I have had a chance to listen to my new XA120+ for a good length of time. The following are my observations as compared with the A75 and what I remember my Aleph 1.2's sounding like driving my line arrays from 2600hz and up. The Pass 350.5 drives the mid bass from 2600hz down to about 45hz where the line arrays are crossed over (in parallel) with a stereo set of Buggtussel transmission line subs.
1. the soundstage is immense, I have never heard it this big. Line arrays have this ability but the Pass amps really show it off.
2. Clarity, I may have to turn the hi frequencies down another .5db or so as the AX really does well with them and the drivers are super efficient.
3. Tonal balance is excellent but I can never be sure how good the master tape was recorded and then records produced. A live saxaphone always seems brighter on most recordings than live.
The system now comes a bit closer to this sense of warmth.
4. Very low distortion-I can't here any! With multiple drivers they don't work very hard.
5. Images are really life like with good height and with line arrays this doesn't change when you are standing.
Lessons learned:
Matching devices is very tedious and the out of circuit tests don't predict what the diff pair will do in circuit. Next time I put in sockets for the transistors and just plug them in and out and if when I get a good set I will then solder them in place.
I am frustrated with noisy transformers, I guess I will have to buck up for a really good audio/hospital grade type that is potted.
More and More heatsink-these run hot and you can only touch the heat sinks for a few seconds after they have been operating for 5 or six hours.
Due to the fact that I am using 24 output devices my class A current is over 6 amps vs the 5amps for GL's. I don't know if this is the reason my 1kva trannies are noisy. The transformer bolt is only attached to the chassis on one end so I am not getting any shorting from the metal hold down bolt.
I am not sure how to figure class A output watts for single ended devices but I get 48volts rms (pk-pk) on sine waves before clipping. I have to add the two curves on my dual trace scope.
With the add function on the scope I don't get a completely st line so I see that as distortion. I cannot adjust the gain pot to completely get rid of it. I am using the 10pf cap across the feedback resistor. I can never be sure what will set an amp oscillating and these do not. I probably couldn't hear the difference. I did bypass the PS caps including some on the boards.
Regarding going to J-fets, I will save that until GL gets finished and can give an update on the sound. I feel J-fets will be an improvement based on the fact that Audio Research uses them as input diff pairs on their very expensive tube amps.
I will listen more this weekend. Thanks to all and now my xformer search begins (maybe NP has extra B stock laying around)--chow, dave
Hi,
nice to hear it sounds good!
Getting frustrated with noisy transformers is something I´ve gotten used too. My expensive potted ones are very noisy until the absolute dc offset gets near zero. After that they are just noisy.
I´ve build a pair of Aleph3´s for a friend with some quite cheap open transformers (Reichelt). One was silent, one just a bit noisy.
Seems you need a bit of luck too!
William
nice to hear it sounds good!
Getting frustrated with noisy transformers is something I´ve gotten used too. My expensive potted ones are very noisy until the absolute dc offset gets near zero. After that they are just noisy.
I´ve build a pair of Aleph3´s for a friend with some quite cheap open transformers (Reichelt). One was silent, one just a bit noisy.
Seems you need a bit of luck too!
William
Hi Dave,
Thank you for the results of your listening tests. I am really happy you're getting the performance you had hoped for. I think a lot of people have learned a great deal from watching your amps come together.
I would very much like to see you start a thread on your speaker project. So when you have the time.......?
Cheers,
Graeme
Thank you for the results of your listening tests. I am really happy you're getting the performance you had hoped for. I think a lot of people have learned a great deal from watching your amps come together.
I would very much like to see you start a thread on your speaker project. So when you have the time.......?
Cheers,
Graeme
aleph1.2 to AX100 and the Doobie Brothers Wine
NP, Gl, William, thanks for the feedback, continued listening has been equally enjoyable as I Played the system all day yesterday while working around the house. I could never have imagined that I would be where I am at without the help and coaching from this thread. A speaker Thread, well that could be done if I could ever get off the airplane for awhile.
I will pass on some information regarding wine. My son buys me every year for Christmas a box set of Red wine from B.R. Cohn wineries in Napa Valley. This winery is owned by the Manager of the Doobie Brothers. Every summer the Doobie Brothers put on a concert at the vineyard. A 1000 box kits of the new wine is labeled with different pictures of their album covers. I only open a bottle for special occassions (three bottle box) and I did last night. If your ever down that way sign up for a set--it was good.
Don't worry I have 11 bottles left.
Bottom line is Pass Amps go good with Doobie Red
dave
NP, Gl, William, thanks for the feedback, continued listening has been equally enjoyable as I Played the system all day yesterday while working around the house. I could never have imagined that I would be where I am at without the help and coaching from this thread. A speaker Thread, well that could be done if I could ever get off the airplane for awhile.
I will pass on some information regarding wine. My son buys me every year for Christmas a box set of Red wine from B.R. Cohn wineries in Napa Valley. This winery is owned by the Manager of the Doobie Brothers. Every summer the Doobie Brothers put on a concert at the vineyard. A 1000 box kits of the new wine is labeled with different pictures of their album covers. I only open a bottle for special occassions (three bottle box) and I did last night. If your ever down that way sign up for a set--it was good.
Don't worry I have 11 bottles left.
Bottom line is Pass Amps go good with Doobie Red
dave
Ian,
One AX100J monobllock is completed. The absolute offset has been adjusted over a period of two hours and appears to have settled. I ran out of adjustment range on V2 with the 5% resistor values so I went back to the 11% values.
The absolute offset starts off at about 8 to 9 volts and drops to about 5 volts after a few minutes. After 20 to 25 minutes it is down to 1.5 volts. It's down to around zero after two hours (maybe less, I wasn't keeping really close track).
The relaitive offset, however, starts off very low - just a few millivolts and rises to about 100 mv after two hours. I find this strange. This is with no connections at the inputs and outputs.
I have not installed the input coupling caps and I find that with the speaker terminals open, the minus shorting plug installed and no connection at the positive input, that the relative offset performance during the settling period is really bad. I don't know if this is normal. I didn't see this with the mosfet inputs but then I had the coupling caps in place.
I have not done any wave form tests or listening tests. There are a few things I want to try first to see if I can improve the relative offset. The thing that really puzzles me is why it grows as the amp heats up and the absolute offset settles. Perhaps there is something about using JFET's instead of MOSFETs in this specific location.
Here are some thoughts:
1) I matched all the resistors in the front end of the AX100 to less than .1% before I installed them. I didn't do this with the AX100J. I will match up a set of resistors and install them. I don't know how much difference this will make but it certainly can't hurt.
2) I am going to try different values of gate to ground resistors to see if the relative offset is affected.
3) Maybe there is a special trick that we aren't aware of yet.
Maybe there is something you or someone else can think of. All suggestions are welcome. I am rather busy the next few days so progress here will be somewhere between slow and non-existant.
How are your amps doing? Did you make much progress this weekend or are you waiting for JFETs? If you recall you asked me to hold you to your committment.
Cheers,
Graeme
Hey Dave - if you were closer I would definitely help you celebrate!
One AX100J monobllock is completed. The absolute offset has been adjusted over a period of two hours and appears to have settled. I ran out of adjustment range on V2 with the 5% resistor values so I went back to the 11% values.
The absolute offset starts off at about 8 to 9 volts and drops to about 5 volts after a few minutes. After 20 to 25 minutes it is down to 1.5 volts. It's down to around zero after two hours (maybe less, I wasn't keeping really close track).
The relaitive offset, however, starts off very low - just a few millivolts and rises to about 100 mv after two hours. I find this strange. This is with no connections at the inputs and outputs.
I have not installed the input coupling caps and I find that with the speaker terminals open, the minus shorting plug installed and no connection at the positive input, that the relative offset performance during the settling period is really bad. I don't know if this is normal. I didn't see this with the mosfet inputs but then I had the coupling caps in place.
I have not done any wave form tests or listening tests. There are a few things I want to try first to see if I can improve the relative offset. The thing that really puzzles me is why it grows as the amp heats up and the absolute offset settles. Perhaps there is something about using JFET's instead of MOSFETs in this specific location.
Here are some thoughts:
1) I matched all the resistors in the front end of the AX100 to less than .1% before I installed them. I didn't do this with the AX100J. I will match up a set of resistors and install them. I don't know how much difference this will make but it certainly can't hurt.
2) I am going to try different values of gate to ground resistors to see if the relative offset is affected.
3) Maybe there is a special trick that we aren't aware of yet.
Maybe there is something you or someone else can think of. All suggestions are welcome. I am rather busy the next few days so progress here will be somewhere between slow and non-existant.
How are your amps doing? Did you make much progress this weekend or are you waiting for JFETs? If you recall you asked me to hold you to your committment.
Cheers,
Graeme
Hey Dave - if you were closer I would definitely help you celebrate!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.