GRollins said:
When they cobbled together the Class definitions in the old days,
they didn't anticipate potential varieties and the level of consumer interest.
A Class AB amplifier starts out Class A, then becomes Class B.
This one is part SE, part PP Class A, then becomes B eventually.
At some point all these amps leave Class A (in spirit if not in
technicality), and it is better to examine the actual operating
curves and their influence on the performance than to describe
them simplistically.
Unless you're selling them, of course.
😎
Member
Joined 2002
Eric said:
I wish i had the money you had for that test equipment DROOL! Good work on the Aleph x.
Nelson Pass said:
I've never been comfortable with some of the the classes. AB1, AB2, and some of the odd ones (e.g. class T and the rest of the alphabet) seem to me to be answers to questions that should never have been asked, not unlike wanting to know how many angels can dance on the head of a pin. Yes, the question can be formulated, but is it a reasonable question to ask?
Circuits like the one under discussion invite the use of prodigious quantities of letters and acronyms. Someone inclined to abuse the language could tag Single Ended, Constant Current Source, Push Pull, class A, class B, and X together and come up with a truly painful name for the circuit: SE-A-CCS-PP-AB-X. That would be difficult to pronounce even for a Klingon. Needs more vowels.
If someone wants to undertake the circuit I'll try to chip in from time to time, but it shouldn't be too difficult to do.
In the meantime, I've got to hie myself hence. I've got a hefty to-do list, beginning with putting up a new mailbox. The DOT (Department of Transportation--yes, another acronym) knocked over my old one Thursday with a backhoe whilst repaving the street. The problem is that my mailbox wasn't mounted on a 4" x 4" post like everyone else's--it was in a monstrous pile of brick (think monolith from 2001: A Space Oddessy, but with a sloped top) that weighed at least half a ton if it weighs an ounce. I not only have to get a new mailbox up, but I've got to remove the rubble left from the old one; it's beyond repair. My back hurts already, just thinking about it.
My next post here may be from the hospital.
Grey
GRollins said:
And there's Class XD(?!!) as defined by Douglas Self. Which turns out after some scrutiny from NP is very similar to the Aleph 0's (v1.0?) output stage. It turns out DS took all the credit did not mention work of designers that came before him...including NPass!
Post#8; Also read thru the pages...
http://www.diyaudio.com/forums/showthread.php?s=&threadid=87916&perpage=10&highlight=&pagenumber=1
Mr. Pass, speaking of the A0 output for the XA.5...does it have current modulation?
Everybody seems to want to define a new class and call it their own. If I were sufficiently ego driven, I'd invent a class G (named after myself, natch). I'm not sufficiently motivated to do so and if I recall correctly somebody (one of the Japanese companies?) has already done so, anyway.
Is class "R" still unclaimed?
Just teasing.
Someday we'll run out of letters in the alphabet, though. Stake your claim now. Tomorrow may be too late.
Here's one of a half-dozen ideas that I considered to see if I could juice the basic topology up enough to get myself interested:
--Okay, so an X-style amp has two banks of outputs, right? Make one side the way Nelson did it in the original Aleph 0. Now make the opposing side "upside down" with P-ch where the N-ch are and vice versa. The fixed current source goes up top on this side. That way the entire amp is going push-pull, offering complementary action from side to side as well as top to bottom.
While I was slaving away in the heat today, I tried to remember what I did for that SSF front end I mentioned above. No go. The best I could do was rouse a faint memory of poor bandwidth being the reason I dropped it. Now before you say "too bad," remember that I'm pretty extreme in my views about bandwidth, and that anything less than, say, 50kHz open loop gets tossed. Clearly that won't bother most people...it's just me being goofy about these things because I don't want to use a lot of feedback to get where I'm going (my target closed loop bandwidth is on the order of 250kHz, minimum, and I want it with less than 15-20dB of feedback; the closer to 0dB, the better). So to me, something like 10kHz open loop is a non-starter, even though you could get it up to 100kHz with about 20dB of NFB. That's still pretty modest feedback by other peoples' standards. It's roughly what the Alephs use, for instance, if my memory cells weren't baked too badly by the sun today.
Of course, the Alephs were known for their poor sound quality...
(ahem)
If I could only remember what I did.
Grey
Is class "R" still unclaimed?
Just teasing.
Someday we'll run out of letters in the alphabet, though. Stake your claim now. Tomorrow may be too late.
Here's one of a half-dozen ideas that I considered to see if I could juice the basic topology up enough to get myself interested:
--Okay, so an X-style amp has two banks of outputs, right? Make one side the way Nelson did it in the original Aleph 0. Now make the opposing side "upside down" with P-ch where the N-ch are and vice versa. The fixed current source goes up top on this side. That way the entire amp is going push-pull, offering complementary action from side to side as well as top to bottom.
While I was slaving away in the heat today, I tried to remember what I did for that SSF front end I mentioned above. No go. The best I could do was rouse a faint memory of poor bandwidth being the reason I dropped it. Now before you say "too bad," remember that I'm pretty extreme in my views about bandwidth, and that anything less than, say, 50kHz open loop gets tossed. Clearly that won't bother most people...it's just me being goofy about these things because I don't want to use a lot of feedback to get where I'm going (my target closed loop bandwidth is on the order of 250kHz, minimum, and I want it with less than 15-20dB of feedback; the closer to 0dB, the better). So to me, something like 10kHz open loop is a non-starter, even though you could get it up to 100kHz with about 20dB of NFB. That's still pretty modest feedback by other peoples' standards. It's roughly what the Alephs use, for instance, if my memory cells weren't baked too badly by the sun today.
Of course, the Alephs were known for their poor sound quality...
(ahem)
If I could only remember what I did.
Grey
Grey,
Thanks for your reply to my earlier question about the Aleph0 output stage. Unfortunately it seems we were talking about different versions – you presumably were referring to the earlier one, whilst I was interested in the post 1.3 version with the active current source as per my posted schematic. I believe I know how it works in principle but not the all important fine details. It is interesting that this current source works quite differently from the most recent Aleph current source in that it delivers a largely constant value below the bias point but increases thereafter, i.e. it is not linear with output current as per the most recent type. Anyway, although I remain interested, it is perhaps moot given that we are now lead to believe that the latest AX.5 uses the earlier Aleph0 fixed current source and most of us are primarily interested in this direction.
I’m not sure that I entirely agree that the early Aleph0 is unexciting although I take your point about it being relatively straightforward to cobble a UGS and a couple of Aleph0 output stages together. As always though the devil is in the details and I suspect, as I believe do you, that Nelson has indeed included something “kinky” in the XA.5. Anyway I have a good reason to pursue further, apart from just sheer interest with which I also afflicted. My rather inefficient Martin Logan’s need an amp with plenty of current available but I would like to retain the sweetness and clarity of the XA/Aleph-X topology.
Another thing I have pondered is what is the important difference between current X.5 and XA.5? Nelson says they are similar but presumably the difference must amount to more than just cranking up the class A bias for the XA.5? Aleph and original X differed in being 2 and 3 stage amplifiers respectively so perhaps this remains significant? Perhaps the XA.5 is more like your “double-cross” Aleph-X variant but with the addition of the SE current source? Just thinking out loud really.
Another question that often bugs me: what exactly contributes to the sonic signature of the XA versus the original X? Is it the SE output compared with PP, the 2 rather than 3 gain stages (treating the followers as active stages here) or something else? Perhaps this is just one of life’s mysteries.
Ian.
Thanks for your reply to my earlier question about the Aleph0 output stage. Unfortunately it seems we were talking about different versions – you presumably were referring to the earlier one, whilst I was interested in the post 1.3 version with the active current source as per my posted schematic. I believe I know how it works in principle but not the all important fine details. It is interesting that this current source works quite differently from the most recent Aleph current source in that it delivers a largely constant value below the bias point but increases thereafter, i.e. it is not linear with output current as per the most recent type. Anyway, although I remain interested, it is perhaps moot given that we are now lead to believe that the latest AX.5 uses the earlier Aleph0 fixed current source and most of us are primarily interested in this direction.
I’m not sure that I entirely agree that the early Aleph0 is unexciting although I take your point about it being relatively straightforward to cobble a UGS and a couple of Aleph0 output stages together. As always though the devil is in the details and I suspect, as I believe do you, that Nelson has indeed included something “kinky” in the XA.5. Anyway I have a good reason to pursue further, apart from just sheer interest with which I also afflicted. My rather inefficient Martin Logan’s need an amp with plenty of current available but I would like to retain the sweetness and clarity of the XA/Aleph-X topology.
Another thing I have pondered is what is the important difference between current X.5 and XA.5? Nelson says they are similar but presumably the difference must amount to more than just cranking up the class A bias for the XA.5? Aleph and original X differed in being 2 and 3 stage amplifiers respectively so perhaps this remains significant? Perhaps the XA.5 is more like your “double-cross” Aleph-X variant but with the addition of the SE current source? Just thinking out loud really.
Another question that often bugs me: what exactly contributes to the sonic signature of the XA versus the original X? Is it the SE output compared with PP, the 2 rather than 3 gain stages (treating the followers as active stages here) or something else? Perhaps this is just one of life’s mysteries.
Ian.
Pass DIY Addict
Joined 2000
Paid Member
jleaman said:
Yeah, me too! One of the benefits of a career in academia is access to a host of nice equipment while the students are away on summer break. I call over to my buddy who runs the EE labs and he gives me access for the day.
Next, I have to see if they have a suitable variac for slowly powering up the amps.
Eric
Ian Macmillan said:
No, theres nothing kinky (not that there's anything wrong with
that). The changes have all been straightforward improvements
and tweaks. You won't find any surprises in the schematics,
but there are some subtle things embedded that have been
worked out over the years that make them sound better.
😎
Ian Macmillan said:
Please do not take my posts as trying to damp anyone's enthusiasm for the circuit. That was not my intention. I was speaking purely in the personal sense. View it as my failure of imagination, rather than anything negative about the circuit.
Counting stages is a perilous endeavor. It's all in the eye of the beholder. That said, I anticipate that (by my count--others will no doubt view things differently) the .5 versions will have at least one extra stage. I believe Nelson said somewhere that the front end is JFET, which pretty much forces at least one additional stage just to handle voltage swing and drive current for the output stage. Like I said above, I see something like a UGS-ish front end. As far as bias goes, I imagine that quiescent bias is actually lower, proportionate to the output wattage, as the amp goes AB above some arbitrary power level before it reaches full power.
The circuits I refer to as Double Cross and Difference Engine don't have much in common with my--admittedly guesswork--perception of what Nelson's doing with the new XAs. But stranger things have occurred. For one thing, both the Double Cross and the Difference Engine have transconductance output stages, whereas the older Aleph is at least party push-pull, albeit with a CCS.
One of the things that I find interesting about the older Aleph output stage (dare I call it Aleph Classic?) is that the characteristics almost have to change as the amp delivers higher power. At low wattages, it will be predominately follower/CCS, but at high power outputs, it will be more push-pull. The transition is graceful, not switch-like, but the idea is an interesting one.
Grey
Interesting discussion.
As I recall the X first super symmetry heavily biased into class A and then A/b beyond. The X.5 has all sorts of refinements with some clever approaches to improved filtering and SE biasing.
The lower powered versions use the Jfet front modules as far as I am aware. The initial information suggests doubling the PS capability and scaling the power and hence full bias to suit the prevailing chassis.
Given the Master's historical desire for loading on the bias it would not surprise me of the AX.5 was in many ways a X.5 on class A steroids.
Ian
As I recall the X first super symmetry heavily biased into class A and then A/b beyond. The X.5 has all sorts of refinements with some clever approaches to improved filtering and SE biasing.
The lower powered versions use the Jfet front modules as far as I am aware. The initial information suggests doubling the PS capability and scaling the power and hence full bias to suit the prevailing chassis.
Given the Master's historical desire for loading on the bias it would not surprise me of the AX.5 was in many ways a X.5 on class A steroids.
Ian
Did the one-handed typing thing while I was feeding the boys and went over to Pass Labs. Looks as though power dissipation on the XA.5s is still roughly 3 x RMS power--about the same as the older XA amps and Alephs.
Grey
Grey
Seems that is at least two of us that thought that the new XA.5 would be a heavily biased X.5 - seems that we might both be wrong though 🙁 Given what the One and Only says, we would appear to have to chose our UGS and remainder of the circuit very carefully as the key difference is in the fine detail. Are we sure the JFET front end only applies to the lower power XA.5? This will potentially make a big difference to the result.
Make that three of us.
http://www.diyaudio.com/forums/showthread.php?postid=1217666#post1217666
Graeme
http://www.diyaudio.com/forums/showthread.php?postid=1217666#post1217666
Graeme
Ian Macmillan said:
There are more similarities before, and this consolidation will
tend to continue over the short term.
😎
Every product has a lifespan. Something like the original Volkswagen Beetle might last for quite a long time indeed, but most consumer items are not quite so fortunate. If you look at cars, which generally are updated every year, then high end audio products have quite long beards by comparison. The original XA topology is roughly the same age as this thread, meaning about five years. It's not that surprising that the circuit might be due for a face lift.
Doubling the power into a 4 Ohm load is likely to be a popular step, but it's not easily done with the original circuit unless you want some really huge amps sitting on the floor; the power supply and heat sink requirements are rather unforgiving. The price tag would climb, as well. So it's logical to look to class AB to fill in the gap.
Grey
Doubling the power into a 4 Ohm load is likely to be a popular step, but it's not easily done with the original circuit unless you want some really huge amps sitting on the floor; the power supply and heat sink requirements are rather unforgiving. The price tag would climb, as well. So it's logical to look to class AB to fill in the gap.
Grey
So where does this leave us in terms of our journey towards a v2 Aleph-X that is similar to an XA.5?
The output stage is to be similar to the classic Aleph0 (coining the phrase that Grey used) which is complementary PP with a parallel SE current sink to the negative rail. The front end is believed to use JFETs in a manner similar to UGS and will need an additional stage (beyond the diff pair) to swing the voltage and provide sufficient current. Does that mean we will end up with a three stage amplifier? I’m guessing it does unless we use a folded cascade (or similar) and count it as a single stage with the diff pair although this may be problematic in terms of the voltage across the JFETs. In addition, I seem to recall the One and Only saying that SuSy only works for simple circuits and the early X amps excluded the output followers from the feedback loop for this reason. Will it be safe applying overall feedback around a three stage amplifier as envisaged? I confess to not completely understanding the issue here, although I appreciate that there may well be more common-mode noise. The later X amps also seem to use overall feedback. Anything else? I’m guessing that we should stick to FETs (MOS or otherwise) in the signal path.
Seems we have quite a bit of the puzzle figured out but not quite enough to proceed. Anyone else like to chip in with their pet theories or ideas? Personally I’m looking for something akin to an XA200.5 in terms of power output.
Ian.
The output stage is to be similar to the classic Aleph0 (coining the phrase that Grey used) which is complementary PP with a parallel SE current sink to the negative rail. The front end is believed to use JFETs in a manner similar to UGS and will need an additional stage (beyond the diff pair) to swing the voltage and provide sufficient current. Does that mean we will end up with a three stage amplifier? I’m guessing it does unless we use a folded cascade (or similar) and count it as a single stage with the diff pair although this may be problematic in terms of the voltage across the JFETs. In addition, I seem to recall the One and Only saying that SuSy only works for simple circuits and the early X amps excluded the output followers from the feedback loop for this reason. Will it be safe applying overall feedback around a three stage amplifier as envisaged? I confess to not completely understanding the issue here, although I appreciate that there may well be more common-mode noise. The later X amps also seem to use overall feedback. Anything else? I’m guessing that we should stick to FETs (MOS or otherwise) in the signal path.
Seems we have quite a bit of the puzzle figured out but not quite enough to proceed. Anyone else like to chip in with their pet theories or ideas? Personally I’m looking for something akin to an XA200.5 in terms of power output.
Ian.
Ian Macmillan said:
If I can scrape up a few minutes this evening (an iffy proposition) I'll try to give a better description, but in the meantime start with the schematic for the output stage of the 40W Aleph 0 (if you bias it properly, this will give you 160W/8 Ohms), cut, mirror-image, and paste. Drop a UGS in between. The only thing missing at this point is the bias circuit, which goes between the opposing Drains of the folded cascode output of the UGS. One bias circuit for each side.
Oh, and you'll need to attend to the DC offset since I have failed to remember what I did with that SSF front end. Sorry.
Grey
EDIT: Depending on who is doing the counting, this is either two stages or three.
Ian Macmillan said:
With Mosfets for inputs, we can use folded cascodes or unity
gain CS easily enough for level shifting as the Mosfets have a
transconductance of about .1S in that circuit. Unfortunately the
JFETs have about 1/5 the transconductance, so I've had to make
that gain up with a gain of 5 on the CS
I would call it 2.5 stages. 😉
😎