Aleph-X
Ian,
Going back over your Cct, the feedback network via R46,47 etc seems to feed directly to those Oh So Precious 2sks which don't seeem to have any protection against DC offset at Amp start-up, etc.
ALL my fets (i/p & o/p) are mounted to the heatsinks and temp controlled (55*C) and they STILL give start-up DC offset that is a fact of life for these amps, I think.
Something about Q11 (21) appears incomplete - not sure what.
Good efforts on Q2,3 should give direct improvements as they drive Q12,17 etc. Perhaps a reference zener and a capacitor across the R6, like the quiet 4040-10 reference zener used on Terry Abens X-bosoz cascodes (Twisted Pair) or a simple Z buffer as per Rod Elliot - seems a bit excessive when it's written like this - pity I can't create those nice neat Ccts that you guys seem so good at.
On a personal note - on my 200w/8R Amp o/p stage that started life as an Audio Research D-100 (?) clone (biased to 10w A), the 39,000uF/80v ITW rail caps (with PP bipass), did a good job but when I added a C multiplier after this, the amp started to REALLY sound good - and when I added a series reg, it killed the sound dead - all the small signal and transient response - gone! couldn't believe it - tried to sort it out for months and just dumped it - I sometimes wonder if the u-beoot Krell P/S setup has a problem like this also - b. nice base tho!
... jh
Ian,
Going back over your Cct, the feedback network via R46,47 etc seems to feed directly to those Oh So Precious 2sks which don't seeem to have any protection against DC offset at Amp start-up, etc.
ALL my fets (i/p & o/p) are mounted to the heatsinks and temp controlled (55*C) and they STILL give start-up DC offset that is a fact of life for these amps, I think.
Something about Q11 (21) appears incomplete - not sure what.
Good efforts on Q2,3 should give direct improvements as they drive Q12,17 etc. Perhaps a reference zener and a capacitor across the R6, like the quiet 4040-10 reference zener used on Terry Abens X-bosoz cascodes (Twisted Pair) or a simple Z buffer as per Rod Elliot - seems a bit excessive when it's written like this - pity I can't create those nice neat Ccts that you guys seem so good at.
On a personal note - on my 200w/8R Amp o/p stage that started life as an Audio Research D-100 (?) clone (biased to 10w A), the 39,000uF/80v ITW rail caps (with PP bipass), did a good job but when I added a C multiplier after this, the amp started to REALLY sound good - and when I added a series reg, it killed the sound dead - all the small signal and transient response - gone! couldn't believe it - tried to sort it out for months and just dumped it - I sometimes wonder if the u-beoot Krell P/S setup has a problem like this also - b. nice base tho!
... jh
James,
Yup, I didn't bother with embellishments like protection in my early drafts and I agree that we should add this before building a proto in earnest.
Not sure what is incomplete about Q11 (21) - looks OK to me.
Yes, using a quiet reference is a fine idea for the cascode bias. I see that the X-BOSOZ also references this to the diff pair sources rather than signal earth. I couldn't decide whether this was better or not. I believe Erno Borbely also uses this arrangement. We could also replace the 'feed' resistor from the supply rail with a CCS or perhaps just split the resistor in two and connect the centre to earch with a capacitor for filtering? Not sure how much benefit there would be? Anyone tried experimenting with this?
I'm still not sure about a Cmx on a potentially class AB amp but maybe I am misguided.
Ian.
Yup, I didn't bother with embellishments like protection in my early drafts and I agree that we should add this before building a proto in earnest.
Not sure what is incomplete about Q11 (21) - looks OK to me.
Yes, using a quiet reference is a fine idea for the cascode bias. I see that the X-BOSOZ also references this to the diff pair sources rather than signal earth. I couldn't decide whether this was better or not. I believe Erno Borbely also uses this arrangement. We could also replace the 'feed' resistor from the supply rail with a CCS or perhaps just split the resistor in two and connect the centre to earch with a capacitor for filtering? Not sure how much benefit there would be? Anyone tried experimenting with this?
I'm still not sure about a Cmx on a potentially class AB amp but maybe I am misguided.
Ian.
There doesn't seem to be a great deal of interest in this project - is it insufficiently radical or too similar to the original Aleph-X perhaps? Is there another reason? I appreciate that I have not yet posted an explanation of how it works but I'd be surprised if that was the problem. Anyway, I'll try and rectify this over the weekend if I get a few spare minutes. I will also aim to post an updated schematic to include the latest suggestions.
The next step? Decide on the voltages and power supply details and get on with building something! Anyone else care to help out?
Ian.
The next step? Decide on the voltages and power supply details and get on with building something! Anyone else care to help out?
Ian.
Hi Ian,
I have to admit that I have asked myself the same question regarding interest. One thing to consider is the fact that the 2sk389 and 2sj109 are unobtainable for almost all of the potential builders here. It reduces the whole exercise to just that- an exercise. It's much easier to develop a passion for a project when there's a substantial chance that it could end up being added to ones own system.
Cheers,
Graeme
I have to admit that I have asked myself the same question regarding interest. One thing to consider is the fact that the 2sk389 and 2sj109 are unobtainable for almost all of the potential builders here. It reduces the whole exercise to just that- an exercise. It's much easier to develop a passion for a project when there's a substantial chance that it could end up being added to ones own system.
Cheers,
Graeme
Graeme,
You may well be right but I don't these particular dual FETs are mandatory - just desirable. One could always build with selected single devices and bond the two together with some heatsink compound as suggested by someone else. There are some other duals around too that would probably do.
Ian.
You may well be right but I don't these particular dual FETs are mandatory - just desirable. One could always build with selected single devices and bond the two together with some heatsink compound as suggested by someone else. There are some other duals around too that would probably do.
Ian.
I think it is a great project, but the smaller amps are just so much easier to build. How many output devices will be needed for a 60W version?
About the parts - like most others, I'm sick and tired of searching for them...
How is the transconductance requirements of this setup? Would 2n5459/2n5462 be an option?
About the parts - like most others, I'm sick and tired of searching for them...
How is the transconductance requirements of this setup? Would 2n5459/2n5462 be an option?
cviller said:
How is the transconductance requirements of this setup? Would 2n5459/2n5462 be an option?
I haven't used these parts so can't comment without looking up the datasheet. That said, there are no strong demands of the front end imposed by this design just as long as the devices can handle the (fairly moderate) voltage and have an Idss of 10mA or more. The strongest requirement is that they be matched as closely as possible.
Ian.
cviller said:I think it is a great project, but the smaller amps are just so much easier to build. How many output devices will be needed for a 60W version?
About the parts - like most others, I'm sick and tired of searching for them...
How is the transconductance requirements of this setup? Would 2n5459/2n5462 be an option?
What if IRFP150 parts were used for this amp? Wouldn't it allow for double the power output with fewer parts?
Kinf of like what Choky did for the Babblefish.
-David
There doesn't seem to be a great deal of interest in this project - is it insufficiently radical or too similar to the original Aleph-X perhaps?
Don't take my silence as lack of interest! I have been watching with great interest to see how everything develops. I am in the process (slowly) of building two big monoblock chasis. I will probably ditch my aleph-X boards and build whatever develops here.
Overall, I really like the circuit. I am a bit more interested in a single ended front end, but at the end of the day, if a group buy develops for this circuit, I will most likely go with it.
PS--Thank you for all the effort going into this.
JJ
dw8083 said:
What if IRFP150 parts were used for this amp? Wouldn't it allow for double the power output with fewer parts?
Kinf of like what Choky did for the Babblefish.
Quite possibly if the other parameters (e.g. gate capacitance) are compatible. I have no experience with the sonics of this device but no doubt others can comment. Getting rid of the heat (twice the dissipation from the same area) may also be a problem but I haven't tried.
Ian.
Here's a quote from Nelson regarding the use of different IRFP parts:
-David
Originally posted by Nelson Pass
Power Mosfet Choices Post #1
I received the following email question, which I think makes for worthwhile discussion:
(Quote) I have posted several questions on diyaudio Aleph-X thread concerning the more esoteric issues in selecting a mosfet for this amp. It had occurred to me that Nelson may be uncomfortable giving an outright answer on the optimum rail voltage and current bias in an open forum so I have decided to ask this privately. When I look at the transconductance figures of the IRF family when compared to their input capacitance, the IRFP240 looks like a relatively poor choice. As per Nelsons suggestions, I read the Zen v5 article and payed particular attention to the fig. 5-7 distortion graphs. The IRFP240 is a 200V part and so it is logical to me that it would provide much lower distortion figures as you increase the voltage rails but would there be commensurate effect with the other IRFP mosfets? To me, the IRFP044N, IRFP150N and IRFP150V have huge a huge transconductance to Ciss ratio compared to the IRFP240 and were designed for a lower voltage application so I would like Nelson (or any of his highly competent engineers) to just give it to me straight. What combination of rails and bias make the best sounding Aleph-X for each of the following parts?
IRFP044N
IRFP150V
IRFP150N
IRFP044
IRFP150 (end quote)
For the record, I experience no discomfort. It is true that I tend to pick IRF244's, partly because I own a zillion of them, having cornered the market on TO-3's. Of course the reason I did that was that I like the 244's sound in general.
Each of these Mosfets brings something different to the table, and each results in a different sound. Some people like the lower transconductance models, such as the early Mosfet like Hitachi's, and some like it higher. A lot higher.
It is my experience that picking the optimal part (or parallel group of parts) is an exercise in experiment and experience, and it is my opinion that for a given amplifier design there is an optimal characteristic for the transconductance.
I recommend the 044 for Zens because they benefit from maximal transconductance, which you get with the lower voltage types. For a Zen amp, high transconductance is a good thing, however it is traded off against Gate capacitance, which is not.
IRFP150's are nice parts. They look to me like two IRFP140's in parallel (the same applies to 250/240 comparison). The advantage in that case to the 140 is that two of them have greater dissipation capability, being in two packages.
According to my book, the 150 has about 1.5 times the Ciss of the 140, but also about 1.5 times more transconductance (this at 10 amps).
Just about any decent Mosfet will work in any of the Zens, X's, and XA's. Besides reliability, it's a question of sonic character.
As the aggregate transconductance gets higher in simple amps, the bottom end becomes more pronounced and you do get better control, but it seems that you are trading off against sparkle on the top, and vice versa.
You will also experience differences in the sound between Mosfets of the same type - remember that the published curves are typical examples. If you really want to nail the thing down, you have to substitute parts in and out and evaluate each.
So we pick a middle ground (or not), and try to enjoy what we've got.
Is this a good argument for bi/tri/quad amping? Yes, I think it is, as you can tailor the amps for best performance in a given spectral area.
It is also a fine example of something a DIYer can do that a manufacturer will never do? I think so.
Ian.
-David
David,
Then I think you have your answer. For a predominently class A amp like the AX, the 140 is preferred over the 150 on reliability grounds (easier to get rid of the heat). Other considerations such as what each sounds like may well be a subjective judgement best left to each individual. I intend to go with the 140 as it is obviously tried and tested, and will hopefully sound good if not necessarily the best in all situations.
Ian.
Then I think you have your answer. For a predominently class A amp like the AX, the 140 is preferred over the 150 on reliability grounds (easier to get rid of the heat). Other considerations such as what each sounds like may well be a subjective judgement best left to each individual. I intend to go with the 140 as it is obviously tried and tested, and will hopefully sound good if not necessarily the best in all situations.
Ian.
As I've noted about thirty times, you can use most any front end you want on this thing. If you feel so motivated, you can use MOSFETs; something like the A-75 front end can be mirror imaged to produce a balanced circuit. Bipolars? Sure. JFETs are nice for any number of reasons, and it's clear that they're the flavor of the month, but they are not a requirement. If you want something in a JFET, track down the 2N5457 and its higher Idss siblings--there's a complementary P-ch series (2N5460, et. al. if I recall correctly). I've made a point of mentioning the J310 and J271 recently. If you're feeling adventurous, try a few of the Fairchild units. They're cheap and available. The Vishay parts work nicely, although the J271 is difficult to locate. I did finally manage to lay hands on several hundred, but it took some doing. Yes, the Toshiba parts are nice, but don't get obsessed.
I've been dropping hints for the last week or two that you shouldn't get all wrung in the withers if your circuit doesn't match some hypothetical circuit that two or three people decide is "what Nelson's doing." They may be right, but they may also be wrong. Use your imagination. Focus on what can be done, not what can't. Survey the parts available to you and see if you can put together a front end that will do the job. One possibility (make it a competition for those who feel that such things are fun) is to have "continental" versions of the front end. North America has relatively easy access to Mouser and Digikey. Europe has...what?...Farnell? Borbely? Asia will presumably have some of the Toshiba parts and who knows what else?
Build front ends. Perfect them. Then send the front end boards around the world for other to listen to.
Just an idea. Or two. Or three.
Me? I'm up to my ankles in alligators, but I'll try to drop hints, ideas, tips, and so forth.
Like I said, don't get rigid about this. Be flexible. Yes, Nelson knows a wee bit about electronics and he has his reasons for building the circuits he builds. But some of those reasons might just be different from yours. You might be able to build something that suits your needs better than his circuits.
Grey
I've been dropping hints for the last week or two that you shouldn't get all wrung in the withers if your circuit doesn't match some hypothetical circuit that two or three people decide is "what Nelson's doing." They may be right, but they may also be wrong. Use your imagination. Focus on what can be done, not what can't. Survey the parts available to you and see if you can put together a front end that will do the job. One possibility (make it a competition for those who feel that such things are fun) is to have "continental" versions of the front end. North America has relatively easy access to Mouser and Digikey. Europe has...what?...Farnell? Borbely? Asia will presumably have some of the Toshiba parts and who knows what else?
Build front ends. Perfect them. Then send the front end boards around the world for other to listen to.
Just an idea. Or two. Or three.
Me? I'm up to my ankles in alligators, but I'll try to drop hints, ideas, tips, and so forth.
Like I said, don't get rigid about this. Be flexible. Yes, Nelson knows a wee bit about electronics and he has his reasons for building the circuits he builds. But some of those reasons might just be different from yours. You might be able to build something that suits your needs better than his circuits.
Grey
Grey,
I think it would spoil the fun to let the cat out of the bag just yet.
To satisfy my own curiousity on all this I had look at the internals of my X250.5 from various angles this afternoon.
All I am going to say is Nelson is a very smart cookie and leave it at that.
Ian
I think it would spoil the fun to let the cat out of the bag just yet.
To satisfy my own curiousity on all this I had look at the internals of my X250.5 from various angles this afternoon.
All I am going to say is Nelson is a very smart cookie and leave it at that.
Ian
macka said:Grey,
I think it would spoil the fun to let the cat out of the bag just yet.
To satisfy my own curiousity on all this I had look at the internals of my X250.5 from various angles this afternoon.
All I am going to say is Nelson is a very smart cookie and leave it at that.
Ian
What cat? Which bag? More hints, please!
Originally posted by GRollins
a current source running in parallel with the P-ch outputs. Given that lower wattage signals will be handled primarily by the heavier bias from up top working against the current source, that part is presumed to be kinda sorta class A. Since the current source is defined (unlike the "Aleph current source"), once the output demand goes beyond its contribution, the P-ch devices kick in and you're in something that's arguably class B, but with the current source still contributing. Want more class A? Increase the current source contribution.
I'm certain that Nelson has thrown something kinky into the topology, but anything that I'd build would be pretty predictable given what I outlined above.
The trick is that you'll need the output drive to end up at ground potential, not at V- plus the Vgs of the output devices. Once you've got that requirement in mind, you can actually use any number of front ends; doesn't even have to be a "UGS."
Hint: I've mentioned in the past that I look at "X" circuits as either Same-Side-Feedback (SSF) or Opposite-Side-Feedback (OSF). If you use an SSF front end, you won't have all the DC offset weirdness that drove so many people crazy at the beginning of this thread.
Its all spelt out here. Just don't get lost in the fine details.
1. Forget all the fine detail for the moment and imagine a basic PP class A/B amp or even an opamp like in the passdiy amps article. How do you bias it up? Clue see Passdiy opamps article.
2. A Class A/B amp is biased into class A to start with or as much as the heatsink real estate as you can afford. This is termed forward biasing via the gates. How do you bias it ? (see Borbely article on class A bias calculations.
3. When output goes beyond the P-ch devices kick in and you're in something that's arguably class B. How do we do that?
Clue :Look to the original Aleph 0 in this case, not the later.
In the case of the AX as I see it one of the above is common to both halfs, the other two are not. You have to use some conceptual imagination then scale the various techniques for the job at hand.
As Grey pointed out the UGS module sort of is what it is..hint!From whay I can see the more recent version has more beefed up fets and a slighty different layout.
Ian
a current source running in parallel with the P-ch outputs. Given that lower wattage signals will be handled primarily by the heavier bias from up top working against the current source, that part is presumed to be kinda sorta class A. Since the current source is defined (unlike the "Aleph current source"), once the output demand goes beyond its contribution, the P-ch devices kick in and you're in something that's arguably class B, but with the current source still contributing. Want more class A? Increase the current source contribution.
I'm certain that Nelson has thrown something kinky into the topology, but anything that I'd build would be pretty predictable given what I outlined above.
The trick is that you'll need the output drive to end up at ground potential, not at V- plus the Vgs of the output devices. Once you've got that requirement in mind, you can actually use any number of front ends; doesn't even have to be a "UGS."
Hint: I've mentioned in the past that I look at "X" circuits as either Same-Side-Feedback (SSF) or Opposite-Side-Feedback (OSF). If you use an SSF front end, you won't have all the DC offset weirdness that drove so many people crazy at the beginning of this thread.
Nelson Pass said:
Paragraph 1: The P channels are also forward biased at idle, so
what you see is a mix of SE A, and PP A at idle. For each half, you
see a slightly asymmetric load line, and the transition in
proportion from SE to PP to "PP AB" occurs at different points
in the two halves of an X circuit.
Paragraph 2: No, it's a variation on the same old stuff - look to
the original Aleph 0 in this case, not the later.
Paragraph 3: Actually the SSF and OSF have the same issues
with absolute offset. That's because there's no common mode
feedback in an X amp. Anything that's identical (in phase) on
both sides is not processed by the feedback loop. X amps
process asymmetrical error.
😎
Its all spelt out here. Just don't get lost in the fine details.
1. Forget all the fine detail for the moment and imagine a basic PP class A/B amp or even an opamp like in the passdiy amps article. How do you bias it up? Clue see Passdiy opamps article.
2. A Class A/B amp is biased into class A to start with or as much as the heatsink real estate as you can afford. This is termed forward biasing via the gates. How do you bias it ? (see Borbely article on class A bias calculations.
3. When output goes beyond the P-ch devices kick in and you're in something that's arguably class B. How do we do that?
Clue :Look to the original Aleph 0 in this case, not the later.
In the case of the AX as I see it one of the above is common to both halfs, the other two are not. You have to use some conceptual imagination then scale the various techniques for the job at hand.
As Grey pointed out the UGS module sort of is what it is..hint!From whay I can see the more recent version has more beefed up fets and a slighty different layout.
Ian
Look here , they seem to have lots of out of stock Toshiba transistors available , including 2SK389 & 2SJ109 ( >5000 pieces of each between BL & GR is what I got when I did a search ). Definitely worthwhile checking. They quote a minimum order of 300$ . Not located in US , is impossible to organize a group buy, but if there are any volunteers to call them and ask for a quote , count me in ...
Okay, I just asked for a price-quote for 100 of each. BL's, since I consider them suitable?.
Not that I intend to buy so many for myself, but it will give a pointer to the pricelevel. I just hope that its not one of those "15 Euro's a piece" kind of shops😉 That's seen before, and it concerns me a bit that they put so much effort into telling that they are specific Toshiba dealers. Kind of gives me the clue that the price will be hefty... Since they know what they are sitting on.
Steen🙂
Not that I intend to buy so many for myself, but it will give a pointer to the pricelevel. I just hope that its not one of those "15 Euro's a piece" kind of shops😉 That's seen before, and it concerns me a bit that they put so much effort into telling that they are specific Toshiba dealers. Kind of gives me the clue that the price will be hefty... Since they know what they are sitting on.
Steen🙂
steenoe said:one of those
Would you like the answer now or rather wait for the miracle response ?

jacco vermeulen said:
Would you like the answer now or rather wait for the miracle response ?![]()
shoot
- Home
- Amplifiers
- Pass Labs
- The Aleph-X