Hi Ian,
It sounds like you're making good progress. The re-scaled resistors will increase the input impedance to something more real-world and practical. The increased Macmillans will reduce the loading on the JFET sources and (IMO) improve the sound in terms of imaging.
That's a good result in finding that the VAS load resistors aren't strictly needed. I had thought from the "UGS adventures" thread that they would be. Excellent!
Cheers,
Graeme
It sounds like you're making good progress. The re-scaled resistors will increase the input impedance to something more real-world and practical. The increased Macmillans will reduce the loading on the JFET sources and (IMO) improve the sound in terms of imaging.
That's a good result in finding that the VAS load resistors aren't strictly needed. I had thought from the "UGS adventures" thread that they would be. Excellent!
Cheers,
Graeme
gl said:
I agree these values make more sense given that they seem to work OK. I note that the Pass Labs user manual for the XA100.5 gives an input impedance of 15K SE and 30 balances so they must be using values at least as high as 10K.
I am still interested in what advantages they might be in using current mirrors and whether it is worth investigating further. Can anyone help out with some theories here?
Ian.
Hi Ian,
1) My choice would be to use the current mirrors. The folks in the UGS adventures thread (Cheff) found that they made a big improvement over the other ideas they tried. And NP chooses to use them here too. In my view that makes mirrors the game to beat - not the other way round. I would use BJT's so that you lose only .7 volts of output swing instead of 3 and I would use 5 transistors in parallel on the output side. That would mean that you could use TO92's for an easier build and not lose the voltage on the output side because your emitter load resistor was 5 times bigger. It also means that you do away with the .5 gain stage (IMO parallel transistors don't count as additional stages). Look at the UGS adventures thread. Hifizen has presented a similar arrangement recently. The two transistors I'm using for this on a current project are the ZTX796A and ZTX696B.
2) I would also like to suggest that you change R21 and R37 to 220K (assuming you have changed the input resistors to 10K) which would lower the gain to about 26dB which I think is more appropriate for a project amp.
I realize that your circuit is just a sim to see what can be done. But perhaps it might be interesting to start scaling it into a practical amplifier. What do you think would be a desirable nominal output for an Aleph X V2?
Cheers,
Graeme
1) My choice would be to use the current mirrors. The folks in the UGS adventures thread (Cheff) found that they made a big improvement over the other ideas they tried. And NP chooses to use them here too. In my view that makes mirrors the game to beat - not the other way round. I would use BJT's so that you lose only .7 volts of output swing instead of 3 and I would use 5 transistors in parallel on the output side. That would mean that you could use TO92's for an easier build and not lose the voltage on the output side because your emitter load resistor was 5 times bigger. It also means that you do away with the .5 gain stage (IMO parallel transistors don't count as additional stages). Look at the UGS adventures thread. Hifizen has presented a similar arrangement recently. The two transistors I'm using for this on a current project are the ZTX796A and ZTX696B.
2) I would also like to suggest that you change R21 and R37 to 220K (assuming you have changed the input resistors to 10K) which would lower the gain to about 26dB which I think is more appropriate for a project amp.
I realize that your circuit is just a sim to see what can be done. But perhaps it might be interesting to start scaling it into a practical amplifier. What do you think would be a desirable nominal output for an Aleph X V2?
Cheers,
Graeme
Ian,
I printed your schematic and gave it a quick glance during supper. It's close enough to what I would have posted that I won't bother knocking together a schematic unless I come up with something I think will add to the mix.
Speaking purely for myself, I'm not sure that the current mirror thing is such a foregone conclusion. There's enough latitude in the design of these things that--unless you're hell-bent on trying to have something identical to Nelson's production unit--most any reasonable balanced front end will do. Don't choke yourself on little-bitty details that may or may not be in the real thing. Yes, I know there are scads of people who will disagree. To them I give a heartfelt "Harrumph!" Is DIY about slavishly copying the same thing a thousand others are doing or is it about carefully hand-crafting a custom piece of art?
Express yourself. Dare to be different.
After five years and enough pages to fill a novel (a decent-sized trilogy if you include the related threads), the principles should be well enough understood that most anyone can build one of these things and have it work.
Grey
I printed your schematic and gave it a quick glance during supper. It's close enough to what I would have posted that I won't bother knocking together a schematic unless I come up with something I think will add to the mix.
Speaking purely for myself, I'm not sure that the current mirror thing is such a foregone conclusion. There's enough latitude in the design of these things that--unless you're hell-bent on trying to have something identical to Nelson's production unit--most any reasonable balanced front end will do. Don't choke yourself on little-bitty details that may or may not be in the real thing. Yes, I know there are scads of people who will disagree. To them I give a heartfelt "Harrumph!" Is DIY about slavishly copying the same thing a thousand others are doing or is it about carefully hand-crafting a custom piece of art?
Express yourself. Dare to be different.
After five years and enough pages to fill a novel (a decent-sized trilogy if you include the related threads), the principles should be well enough understood that most anyone can build one of these things and have it work.
Grey
gl said:
If I may squeeze in a comment, I think that a 100 watt'er would be a solid baseline to go for. There are lots of small wattage F's and Zen's to choose from😉 A solid well working, managable (read adjustable😉 ) 100 watt'er would be the dream🙂
Steen😎
Hi Steen,
A 100W amp sounds good. It was my choice when building an Aleph-X so you know I was going to agree. The trouble is a 100W class A amp may be too intimidating for most DIY'ers plus they may not need the power. From what I've seen most of the Aleph-X's built were in the 40 watt range of the original and people were quite happy with that much output power.
All I'm suggesting is that a "nominal" basic amplifier be defined. It could be the default that most people, especially inexperienced people, could build. The more experienced people could scale that design up as required.
OK, I was thinking that the way to go here would be to choose a commonly available transfromer secondary voltage like say 18V X 2, choose a conservative output wattage based on the voltage swing available, figure out the VA rating from that, set a dissipation per output transistor goal, calculate the bias-per-transistor, calculate the number of output transistors, and done. I'm thinking something in the 50W to 75W range? Comments?
Oh yes, and there would need to be a ratio set of SE bias vs PP bias.
Cheers,
Graeme
A 100W amp sounds good. It was my choice when building an Aleph-X so you know I was going to agree. The trouble is a 100W class A amp may be too intimidating for most DIY'ers plus they may not need the power. From what I've seen most of the Aleph-X's built were in the 40 watt range of the original and people were quite happy with that much output power.
All I'm suggesting is that a "nominal" basic amplifier be defined. It could be the default that most people, especially inexperienced people, could build. The more experienced people could scale that design up as required.
OK, I was thinking that the way to go here would be to choose a commonly available transfromer secondary voltage like say 18V X 2, choose a conservative output wattage based on the voltage swing available, figure out the VA rating from that, set a dissipation per output transistor goal, calculate the bias-per-transistor, calculate the number of output transistors, and done. I'm thinking something in the 50W to 75W range? Comments?
Oh yes, and there would need to be a ratio set of SE bias vs PP bias.
Cheers,
Graeme
Graeme, if people dont want a "lot" of power, I would suggest that they build an F-amp or a ZenV?-something!! There is a lot to choose from. If you did read my questionable reviews, you will know that some of them sounded good too!! The ZenV9 sounds really nice!! In my point of view, the A-X is for those wanting more power than the avarage NP-amp (or Firstwatt, as it were) Its not for the newbie DIY'er, whatsoever!!! They would better stick with a tried and tested design!! The F4 being mentioned as one!!
Steen😎
BTW Graeme, I still have a bunch of steelheadflies sitting here on my bench, with your name on it???
Steen😎
BTW Graeme, I still have a bunch of steelheadflies sitting here on my bench, with your name on it???
Seems like I have a few things to reply to. Here goes…
GL: Current mirrors? I can still find little advantage to using these as the Vas stage and the obvious drawback is additional complexity. Based on a simulation, which I accept may well be wrong, the open loop gain and frequency response are hardly altered by the addition of mirrors. Not surprising really as the current swing in both diff pair and Vas is relatively small and hence the variation in Vgs of the Vas is also small. Under these conditions, the ratio of current between the two will remain largely constant, i.e. like a current mirror. I also would prefer to avoid additional bipolar devices in the signal path as others have reported that these can add a ‘haze’ to the sound (including respected folks like Erno Borbely). I agree this could be subjective and down to personal preference but there you have it. Simple, if done well, is usually best. I’m also fairly sure that the XA100.5 does not use mirrors although this is based only on an inspection of a relatively poor resolution photo of the UGS5. Basically a quick device count rules out mirrors unless semiconductors are also hidden under the PCB. Unless someone knows differently…
I’ve changed all the resistor values as suggested and attach a revised schematic.
Power output for a proto? Difficult one as it will be a compromise based on cost, weight (heatsinks & transformer), etc and the amount of power required to drive the target speakers. With my inefficient speakers, I will need at least 100W if I am to be able to make a realistic subjective (read listening) judgment of the performance of any proto. The needs of others may well be less. The values shown are roughly right for a 100W version although it could be that the power supplies can be reduced a little in voltage (we will need plenty of current). For that matter we haven’t decided whether a split power supply is worth the effort. Nor have we discussed what kind or smoothing/regulation we might at least for the front end. Ah, so many decisions…
Grey: I take your point about not slavishly trying to copy someone else’s work. However, Nelson is both more experienced and has more time to weed out the duff ideas and hence his direction is usually a good one to follow. Pass Labs just happens to make the best sounding amplifiers around too. A personal view I appreciate.
Ian.
GL: Current mirrors? I can still find little advantage to using these as the Vas stage and the obvious drawback is additional complexity. Based on a simulation, which I accept may well be wrong, the open loop gain and frequency response are hardly altered by the addition of mirrors. Not surprising really as the current swing in both diff pair and Vas is relatively small and hence the variation in Vgs of the Vas is also small. Under these conditions, the ratio of current between the two will remain largely constant, i.e. like a current mirror. I also would prefer to avoid additional bipolar devices in the signal path as others have reported that these can add a ‘haze’ to the sound (including respected folks like Erno Borbely). I agree this could be subjective and down to personal preference but there you have it. Simple, if done well, is usually best. I’m also fairly sure that the XA100.5 does not use mirrors although this is based only on an inspection of a relatively poor resolution photo of the UGS5. Basically a quick device count rules out mirrors unless semiconductors are also hidden under the PCB. Unless someone knows differently…
I’ve changed all the resistor values as suggested and attach a revised schematic.
Power output for a proto? Difficult one as it will be a compromise based on cost, weight (heatsinks & transformer), etc and the amount of power required to drive the target speakers. With my inefficient speakers, I will need at least 100W if I am to be able to make a realistic subjective (read listening) judgment of the performance of any proto. The needs of others may well be less. The values shown are roughly right for a 100W version although it could be that the power supplies can be reduced a little in voltage (we will need plenty of current). For that matter we haven’t decided whether a split power supply is worth the effort. Nor have we discussed what kind or smoothing/regulation we might at least for the front end. Ah, so many decisions…
Grey: I take your point about not slavishly trying to copy someone else’s work. However, Nelson is both more experienced and has more time to weed out the duff ideas and hence his direction is usually a good one to follow. Pass Labs just happens to make the best sounding amplifiers around too. A personal view I appreciate.
Ian.
Hi Ian,
OK. Well..... I stand on my comments regarding mirrors. I believe that that's what's in the XA100.5. The big clue is in NP's characterization of the circuit as having 2.5 stages. But let's move on.
I disagree with the blanket comment regarding BJT's and haze. In the early days "haze" was liberally applied to the sound of power mosfets. I've also seen certain valves given this label. One can't generalize. Again, read the UGS adventures thread. There's a lot of experience there that was gained by listening.
Your circuit as it stands has enough voltage swing for an amp of over 200W. But for a 100W amp you would have to be running 2.5A of bias through Q15 and Q22 yielding a dissipation on each transistor of 77W. That's too much IMO.
Cheers,
Graeme
P.S. Steen - Ouch! You got me good there. Email coming tomorrow. I promise. Again.
OK. Well..... I stand on my comments regarding mirrors. I believe that that's what's in the XA100.5. The big clue is in NP's characterization of the circuit as having 2.5 stages. But let's move on.
I disagree with the blanket comment regarding BJT's and haze. In the early days "haze" was liberally applied to the sound of power mosfets. I've also seen certain valves given this label. One can't generalize. Again, read the UGS adventures thread. There's a lot of experience there that was gained by listening.
Your circuit as it stands has enough voltage swing for an amp of over 200W. But for a 100W amp you would have to be running 2.5A of bias through Q15 and Q22 yielding a dissipation on each transistor of 77W. That's too much IMO.
Cheers,
Graeme
P.S. Steen - Ouch! You got me good there. Email coming tomorrow. I promise. Again.
The middle stage of the XA is just a quad-symmetric set of
common-source devices with so much degeneration that they
don't have much gain. Topologically, it's not any more
complicated than that, so don't get too excited.
😎
common-source devices with so much degeneration that they
don't have much gain. Topologically, it's not any more
complicated than that, so don't get too excited.
😎
It's not like it's all that difficult to achieve +-25V swing. You fellas are acting as though you're building a 10kW amp and all the gain has to be done in one step.
In this case, the front end can easily have a separate, higher rail (there goes Grey again with all that complicated rail nonsense...), unlike the Alephs and Aleph-X where the negative rail is integral to the output stage bias. Not that other schemes can't be managed, but referencing the bias to the rail is a real elegant solution to the problem. The ability to have a higher rail allows for a lot of voltage swing. It's just not an issue.
Grey
In this case, the front end can easily have a separate, higher rail (there goes Grey again with all that complicated rail nonsense...), unlike the Alephs and Aleph-X where the negative rail is integral to the output stage bias. Not that other schemes can't be managed, but referencing the bias to the rail is a real elegant solution to the problem. The ability to have a higher rail allows for a lot of voltage swing. It's just not an issue.
Grey
Aleph-XIO
Hi Grey,
I hesitate to disagree with you about Power Supply "simplicity", but my two cents worth may spark and/or promote some p/s discussion and variety.
By using a seperate supply for the front end instead of relying on the balanced topology to cancel out common noise, ripple, o/p load modulation, etc the Av gain stage will nearly always (?!) "function better" and CMR cancellation, etc is a BONUS to Cct performance, not a necessity.
It also encourages different Cct variations and arrangements as rails can be of any old voltage at all.
I don't think it's "news" that different P/S systems (and quality) alter the sound of the active Ccts, and it is a bit surprising that there has been little suggestion of using Cmx, series or shunt rail regulation, and not only for the A-X.
However, it CAN BE more complicated and expensive with seperate Tx windings, parts count, higher power consumption, power-up rail delay times,and so on, but if "done well", can achieve remarkable improvements with the very same Circuit.
Don't shout at me for saying this but the "tube" and "hybrid" fraternity accept really "weird" (!) variations of rail voltages (and regulation) as routine but the attraction of P/S simplicity seems to have created unneccessary design restrictions here.
My X-Bosoz (Russ / Brian) has undergone some strange P/S rearrangements plus Steens (?) j-fet alteration and shows amazing variations of sound, including that most unwelcome "blue smoke" variety!
My reference is the "Passlabs X-2.5" into the "F3" into .....
... James
Hi Grey,
I hesitate to disagree with you about Power Supply "simplicity", but my two cents worth may spark and/or promote some p/s discussion and variety.
By using a seperate supply for the front end instead of relying on the balanced topology to cancel out common noise, ripple, o/p load modulation, etc the Av gain stage will nearly always (?!) "function better" and CMR cancellation, etc is a BONUS to Cct performance, not a necessity.
It also encourages different Cct variations and arrangements as rails can be of any old voltage at all.
I don't think it's "news" that different P/S systems (and quality) alter the sound of the active Ccts, and it is a bit surprising that there has been little suggestion of using Cmx, series or shunt rail regulation, and not only for the A-X.
However, it CAN BE more complicated and expensive with seperate Tx windings, parts count, higher power consumption, power-up rail delay times,and so on, but if "done well", can achieve remarkable improvements with the very same Circuit.
Don't shout at me for saying this but the "tube" and "hybrid" fraternity accept really "weird" (!) variations of rail voltages (and regulation) as routine but the attraction of P/S simplicity seems to have created unneccessary design restrictions here.
My X-Bosoz (Russ / Brian) has undergone some strange P/S rearrangements plus Steens (?) j-fet alteration and shows amazing variations of sound, including that most unwelcome "blue smoke" variety!
My reference is the "Passlabs X-2.5" into the "F3" into .....
... James
Re: Aleph-XIO
Thanks James, but the J-fet alteration is the work of Metalman and Zen Mod. I was only building it😉
Steen🙂
jameshillj said:
Thanks James, but the J-fet alteration is the work of Metalman and Zen Mod. I was only building it😉
Steen🙂
Well it seems that the question regarding the mirrors is answered and shall be sticking with the current design as it would appear to be most similar to shipping AX.5. No doubt others will try the mirror and some may even decide that is better. Me, I like to keep it simple.
It also sounds like we are agreed on using a dual rail PS as per the schematic. The question is now what voltages, how much current and perhaps more important, what kind of regulation for the front end supply? I'm currently favouring either a current source or capacitor multiplier approach as the current draw should be fairly constant (front end only). I'm sure some people will want to throw iron at it in the form of inductors but I think there will be enought weight with the transformer and heatsinks.
I appreciate that the voltages I have shown are more than required for 100W but I was working on the principle that we after 100W class A but something beyond that in AB. 100W is not a lot to allow for realistic peaks even though it will more than cater for average levels. Does this make sense or am I treading too much into the 'standard' X.5 space?
Ian.
It also sounds like we are agreed on using a dual rail PS as per the schematic. The question is now what voltages, how much current and perhaps more important, what kind of regulation for the front end supply? I'm currently favouring either a current source or capacitor multiplier approach as the current draw should be fairly constant (front end only). I'm sure some people will want to throw iron at it in the form of inductors but I think there will be enought weight with the transformer and heatsinks.
I appreciate that the voltages I have shown are more than required for 100W but I was working on the principle that we after 100W class A but something beyond that in AB. 100W is not a lot to allow for realistic peaks even though it will more than cater for average levels. Does this make sense or am I treading too much into the 'standard' X.5 space?
Ian.
Oops, forgot to address GLs concern over 77W per transistor which I agree is too much. I was planning on replicating the output stage by 6, i.e. a total of 40 devices including the SE current sources. This should limit the current to 1A or less per device which I think is fine for IRF240.
Ian.
Ian.
Aleph-X
Ian,
I went back over your design AFTER I posted my comments about power supplies, and see that you have already done just this - my apologies.
I would encourage you to really go to town on the I/P stage regs with medium ripple caps(3300uF) into Cmx for fast and solid bass, then current source (maybe) into a good non-feedback shunt reg to tailor the sound for whatever particular balance of bass, mids and tops that you want, not just the minimum.
The series regs don't seem to offer the same possibilities and / or flexibility but are quite an improvement on the usual CRC (or CLC) setup even when overdone like most of us like to do - they are also a lot simpler.
On your Cct diag, I would suggest to beef-up the isolation of the junction of R6,24 and bases of those Q's (plus R20,26 etc) and perhaps also include an extra R across those VRs - just being a bit picky, please excuse, but if it's kept in mind for possible later pcb design.....
It's taking shape nicely - lots of gain, too. ... jh
Ian,
I went back over your design AFTER I posted my comments about power supplies, and see that you have already done just this - my apologies.
I would encourage you to really go to town on the I/P stage regs with medium ripple caps(3300uF) into Cmx for fast and solid bass, then current source (maybe) into a good non-feedback shunt reg to tailor the sound for whatever particular balance of bass, mids and tops that you want, not just the minimum.
The series regs don't seem to offer the same possibilities and / or flexibility but are quite an improvement on the usual CRC (or CLC) setup even when overdone like most of us like to do - they are also a lot simpler.
On your Cct diag, I would suggest to beef-up the isolation of the junction of R6,24 and bases of those Q's (plus R20,26 etc) and perhaps also include an extra R across those VRs - just being a bit picky, please excuse, but if it's kept in mind for possible later pcb design.....
It's taking shape nicely - lots of gain, too. ... jh
James
All feedback and suggestions are appreciated, I have no problem with you being picky. One thing about DIY is that we don't have to worry about commercial cost considerations and can quite happily go to town with things!
I don't much like series regulators either so I wouldn't have chosen that route. Current source (mainly for good filtering) followed by shunt reg is fine. Forgive my ignorance but Cmx stands for what?
I wasn't sure about the simple voltage divider for the cascodes either but showed them biased this way for simplicity and to get more serious issues out of the way first. Given that we don't (yet) have any serious issues I agree we can think about improving this arrangement. What had you in mind, capacitors, current source or something else entirely?
Agree w.r.t resistors across VRs. If we are looking at producing PCBs then we might also want to include allowance for frequency compensation capacitors although I doubt that they will be required in practice.
Keep the suggestions coming...
Ian.
All feedback and suggestions are appreciated, I have no problem with you being picky. One thing about DIY is that we don't have to worry about commercial cost considerations and can quite happily go to town with things!
I don't much like series regulators either so I wouldn't have chosen that route. Current source (mainly for good filtering) followed by shunt reg is fine. Forgive my ignorance but Cmx stands for what?
I wasn't sure about the simple voltage divider for the cascodes either but showed them biased this way for simplicity and to get more serious issues out of the way first. Given that we don't (yet) have any serious issues I agree we can think about improving this arrangement. What had you in mind, capacitors, current source or something else entirely?
Agree w.r.t resistors across VRs. If we are looking at producing PCBs then we might also want to include allowance for frequency compensation capacitors although I doubt that they will be required in practice.
Keep the suggestions coming...
Ian.
Aleph-X
Ian,
Cmx is just short for C multiplier - Sorry - and I'm always having trouble trying to figure out other peoples short-cuts!
I'm having a bit of fun with trannie version of "Mr Evil"s "Ultimate C Multiplier"- he's from your side of the pond - new website - I'm sure he won't mind me airing his comments and site here-
<MrEvil@blueyonder.co.uk>
Indeed I did finish it. I did make a second post after I had finished
it, which can still be found:
http://www.diyaudio.com/forums/showthread.php?threadid=53359
It surpassed my expectations, working perfectly first time. It's still
working great.
Miles.
I'm quite interested in the idea of really high gain without any stability problems and so can amplify small, good propylenes to give even better results and reduce size of ripple caps and hence (!) better quality, etc - I tend to ramble on!
Mind you, I have used these things mainly on valve gear and preamps and have even gone back to using the simple C multiplier on the o/p stage rails on my poor suffering basic A-X - (tried, tested, patched tracks and still working well!)
Actually, now that I mention this, I'm looking for a new set of Aleph-X boards (the ones with the skinny o/p stage pcbs) if you know of any not in use.....
... james
Ian,
Cmx is just short for C multiplier - Sorry - and I'm always having trouble trying to figure out other peoples short-cuts!
I'm having a bit of fun with trannie version of "Mr Evil"s "Ultimate C Multiplier"- he's from your side of the pond - new website - I'm sure he won't mind me airing his comments and site here-
<MrEvil@blueyonder.co.uk>
Indeed I did finish it. I did make a second post after I had finished
it, which can still be found:
http://www.diyaudio.com/forums/showthread.php?threadid=53359
It surpassed my expectations, working perfectly first time. It's still
working great.
Miles.
I'm quite interested in the idea of really high gain without any stability problems and so can amplify small, good propylenes to give even better results and reduce size of ripple caps and hence (!) better quality, etc - I tend to ramble on!
Mind you, I have used these things mainly on valve gear and preamps and have even gone back to using the simple C multiplier on the o/p stage rails on my poor suffering basic A-X - (tried, tested, patched tracks and still working well!)
Actually, now that I mention this, I'm looking for a new set of Aleph-X boards (the ones with the skinny o/p stage pcbs) if you know of any not in use.....
... james
James
Thank you for the clarification 🙂
In my albeit limited experience, high gain seldom comes without instability problems. I'm leaning more in the direction of CCS and shunt reg without too much feedback. Actually I'm not even sure that the front end of this design really requires much regulation as the current draw will be pretty much constant.
The output stage is another matter, particularly if we are expecting to make much use of the region beyond class A. I'm inclined to stick to something simple here such as a bank of not-too-big capacitors. Not sure about the use of Cmx in this situation.
Sorry I don't have any PCBs.
Ian.
Thank you for the clarification 🙂
In my albeit limited experience, high gain seldom comes without instability problems. I'm leaning more in the direction of CCS and shunt reg without too much feedback. Actually I'm not even sure that the front end of this design really requires much regulation as the current draw will be pretty much constant.
The output stage is another matter, particularly if we are expecting to make much use of the region beyond class A. I'm inclined to stick to something simple here such as a bank of not-too-big capacitors. Not sure about the use of Cmx in this situation.
Sorry I don't have any PCBs.
Ian.