The Aleph-X

[HarryHaller]

Yes it will for reasons of transconductance and N vs. P type devices as well. Who said any thing about just dropping it in?

H.H.

[rtirion]

A 2sj109 solves one problem. The rest is up to you.:-)

[jam]

Harry,

I agree that a matched j-fet pair may better offset characteristics but mosfets will sound better.

Jam

P.S. That's for everyone that thinks that I agree with everything you say. O' Great One .

[GRollins]

As it happens, I launched into the R&D on this thing without a decently matched set of IRF9610s on hand. (Poor at the moment...kinda stung to place orders for the parts for this thing.) Some of the imbalance I was fighting in the beginning was due purely to a notably imperfect front end. I remember grousing to Nelson in an e-mail that I needed a decent differential, and mentioned that the '389 might make a good choice. Another point in its favor is the inherently tight coupling, thermally, harking back to the point we were talking about earlier about having the differentials on the same heatsink.
Incidentally, for those who might be wondering, Q5 & Q7 (the input differential) don't really have to be on a heatsink at all. The power dissipation is really quite modest for a TO-220 device and they're quite happy to be naked. If someone chooses to run this circuit at a higher voltage, then the power dissipation will eventually become an issue. The only reason to heatsink them at this point is to keep the two devices tracking the same as ambient temperatures change. And change they will...this thing puts out a fair amount of heat.
I don't have any 2SK389s on hand, and have pretty much hit the limit monetarily. Amongst other things, I'm planning on Caddock resistors for the MOSFET sources and outputs and what with the new IRFP044s etc., alternate front ends will have to wait...probably for quite some time. Sad, but there it is. Another problem is that people seem to feel that there are a great many counterfeit Japanese devices being sold out there; for every person who says that they are happy with the devices they bought from business A, another comes along and says that A sold them bogus parts, but they are happy with parts obtained from B, then another person comes along and says that B sold them bad parts, and that he likes vendor C...and so on. It would be nice if the 2SK389 was available from someone like Digikey or Mouser (where do they come up with these names?) who you could count on.
Maybe someday.
I still don't know what I'm going to do for bulk power supply caps. I thought I had a box with a bunch of 25 & 35V caps in it. Turned out they were 7.5 and 10V. Ooops! For the time being, I'll toss in some mongrel 60V caps until I get a chance to knock over a little old lady and steal her lunch money.
And to think it used to be the young, pretty girls who ran when they saw me coming...my, how the times have changed.

Grey

[HarryHaller]

Jfets are not drop in replacements for the mosfets. You may be comparing apples and oranges if you are using them in the same circuit. They are great diff pairs and I have heard them sound great in several products as well as my own designs. I heard a rumour that Nelson Pass even uses them in some products....... You know a couple of matched J109s in parallel with an Idss of greater than 8mA would just about drop into Grey's.......... neverminds, everyone forget what I just said.



"It would be nice if the 2SK389 was available from someone like Digikey or Mouser (where do they come up with these names?) who you could count on." They are MCM and Consolodated Electronics.

By the way I have never seen any counterfit japanese transistors, sounds like a case of audiophile nervosa to me. But that could never happen, right?

H.H.

[Joe Berry]

Originally posted by GRollins
I know at one point not too long ago, Joe Berry sent me a schematic of an Aleph-Xish kind of circuit. It wasn’t like this one, but by the second or third try he had the differential cross-over trick in hand, but nothing in place to handle the DC offset. I believe he was modeling the circuit, rather than building it. What he ever did about the DC, I don’t know, as he was one of the people I quit responding to ‘cause I just couldn’t afford the time.

Hi Gray -- just to update you:

That first schematic was to show that two independent single-ended Aleph channels could be bridged to give su-sy behavior with no DC offset issues. It worked well in the simulator, but on closer examination, I found that it needed to have a balanced input signal, so I became a bit disenchanted with the approach.

In exploring a second possibility through email exchanges with NP, I also ended up with essentially the same schematic as the one you've posted here. My initial reaction to 30 ohms from each output to ground was also similar to yours, but it helped me to imagine them as finned aluminum units bolted to the inside rear panel at the speaker binding posts, rather than taking up lots of PC board real estate. Nelson himself suggested that the resistors could even be incandescent lamps a la Zen Lite. ;-)

I did model several variations of the circuit, and used thermal DC sweeps to play with the effect of different values of output resistance. I ended up adding a couple of regular diodes in series with D1 -- one to counteract the thermal trend of Q3/Q8, and a second to cancel out D1's thermal drift. This improved absolute offset stability in the model, but Nelson seemed to see this as an unnecessary complication, so it may not matter much in real life.

I also proposed a relatively simple DC servo design as an alternative to the power resistors at the output. This seemed to work quite well in the model. But again, NP seemed to think that this was an needlessly complex solution.

I look forward to hearing more about your progress with an actual prototype. I'm quite sure the results will repay the effort!

[grataku]

Grey,
I have been waiting for this since the CES ended, cool!
What about a mini-aleph-X set up to give you the same power as the mini-me, sorry miniA with +/- 7.5 V? You would get to use all your 10V electrolytics.

HH
never seen fake Japs? Try buying 2sc2922 from MCM and see what you get!
Another thing that spoils the fun when buying japs is the morons that work in the stacks at places like MCM and such. You see, all the japs transistors come in very carefully selected gain ranges with super low gain distribution in the groups, so you may get the real deal but what good does it do if they give you different gain series?
I would gladly pay 10 extra cents to get semis from the same gain groups but no vendor seems to have the time and power to accomodate my humble request.

[hifi]

How large psu is required for that thing and can it live from a floating supply of say 45V ? how much heat will it produce? then and how much power will it develop?

/micke

[lohk]

Very nice work, Grey. Thank you. I just went through this thread.

Some observations:
Looking inside the X600 I cannot spot any 2sk389s ( There are two of them in the X0.2 at least). Only IRF610, IRF9610, IRFP240, IRFP9240 etc. There are separate bias and offset pots (5k) for each "side" of the amp and another one for the input (?). Except of the output section and a - thermally coupled - pair at the input (?) all other MOSFETs are sitting on their individual heatsinks.
And I do not see any Caddocks etc.

You may wonder from where I poor boy know all this: I am - sigh! - not looking at my own amps, just marvelling over the glossy images printed in the German mag Audiophile 3/2002...

Klaus

[GRollins]

Joe,
Good to hear that you managed to beat the thing into shape. I make no claims that this is the be-all, end-all...that's why I called it Version 1.0.
Incidentally, my intention--for mine, at least, you guys can do as you like--is to build something along the lines of the old Mark Levinson ML-2. 25W into 8 ohms, 50W/4, 100/2...given that the Magneplanar ribbon tweeters are around 2.5 ohms, the Aleph 2 is a less than perfect match. The bias isn't high enough. The solution is lower rails and mucho bias current. As I said above, +-12V rails will give the right general output voltage, and I'll be putting in .15 ohm source resistors, which will bias the outputs to roughly 3 amps per side. Also reduce R16 to 1.0k. Yes, I intend to regulate the rails, although I don't know if I'll make it quite so fancy as the ML-2's regulators.
Harry,
Note grataku's post about parts. Like I said...I'd love for Digikey or Mouser to carry the Toshiba JFET pieces. I'm too tired/lazy (four hours sleep last night) to look, but I believe that Digikey already carries some Toshiba parts (perhaps Mouser does also, don't remember). Surely, it wouldn't be too much trouble for them to stock JFETs as well.
grataku,
I guess the ultimate goal here would be to go for a 1W amp (I've also got--I think--some 5 and 6V caps if some one wants to trade). Sadly, none of the speakers I've got on hand are efficient enough to take advantage of such a thing. Horn speaker fans, take note.
micke,
45V? Whoa! Monster amp! Keep in mind that the rails on Nelson's 200W version will be on the order of 32V or so. According to some rough calculations I just did, 45V rails would put you in the vicinity of 400W into an 8 ohm load, but you'd be running 1300W of pure heat (looks like about 14A total bias). Now, that would be a serious project, indeed.
You might want to go water-cooled on that...
Klaus,
The Caddocks are just something I wanted to do as a DIY thing. Nelson's got to strike a balance between sound quality and parts cost; he's got to show a profit. Still, might be fun to take one of Nelson's production pieces and retrofit it with Caddock stuff to tune it up.

Grey