Aleph-X with 2SJ109 as input pair

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi folks!

Being thrilled with the great Aleph-X design (thanks Grey!!) I want to combine it with another thing that has my attention at the moment: JFet inputs.

After reading wuffwaffs thread on it (see here if you missed it ) I think the dual package 2sj109bl with the same transconductance as the 2sa74 should fit perfectly, no?

Using a dual package should be the best in terms of matching, no?

Up to which rail voltage I don't need to cascode the jfets? I refer mainly to EUVL's post here , where he says 20V would be too close to VDS - but up to what would it be ok?

I want to build a stock 38W Aleph-X (+-15V rails), but I don't get matched IRFP044, just IRFP244 so I read that I've to use higher rails (> 20V) to lower distortion of these IRFP244.

If you wonder, what the main point in my post is, then..well I want to collect all design recommendations and maybe some time in the future I can stick them all together and get a nicely working Jfet input stage (dreaming).

Or what changes would be needed when I simply replace the IRF9619 pair with the 2SJ109? Maybe I'm lucky and somebody has too much spare time and cares to help me :rolleyes: ?

If I need to cascode, what parts would fit perfectly?

Of course if somebody of you gentlemen drops a schematic, I'll gladly take that!

Yes I know many questions!

Thanks a lot for helping! Hannes

PS: info will be added to the Aleph-X wiki, so don't desparate, nothing will be lost and other people can look it up without bothering you again!
 
Very nice schemetics for JFET input is in service manual for Passlabs ONO. I have built one (bit simplified) and it was working perfectly. So I decided to build up also power amp on the same principple - and it is also working perfectly. It is difficult for me to say how it sounds because I have never heard original Aleph. I was looking for less analytical and more musical sound and this my requirement was fulfilled.

:)
 
You mean the 2SK389 input that also the pearl has?

Interesting suggestion - thanks a lot!

However I do think this is a bit far from what I can use without doing actual measurements, since the Ono and the pearl as well use the 2SK389 only cascoded (the pearl uses the ztx450 if I remember correctly and the Ono the 2sc1844).

What about the current through the 2SJ109? What do you guys think is best?

Cheers, Hannes

PS: hehe I know never ask what's best...then you get dozens of replies - just what I want ;-))
 
If you wish to build the Grey Rollins circuit with 4 FETs, then I would recommend you to use IRFP044. It has more transconductance then the 244, and you need that with 4 FETs. If you really have problems getting IRFP044s, you may conatct me for matched sets of 4's. But you should have no problem getting them from Conrad, Reichelt, RS, Farnell, etc.

Voltage would still be +/-15V, so 2SK109 is fine without cascode. If you use 1k resistor at the drain of the JFETs, you would need Idss of about 8 to 10mA. But you may find 2J109BLs very difficult to get.

If not in a rush, I suggest you wait until Nelson publishes his Aleph-J service manual.


Patrick
 
Hi Patrick!

Thanks for your reply!

Indeed getting the IRFP044 is not the problem, just getting matched ones are! Originally I wanted to build a stock Aleph 30, so I ordered nicely matched IRFP244 (unfortunately this source does not offer matched IRFP044 :bawling: ); however since I just learned developing pcbs I can build now whatever I want and not only amps for which pcbs are offered!

So Aleph-X I'm coming :D

And yes, I'm not in a rush (actually I've to finish 3 other projects first, hehe), so indeed I want to have a look at Nelson's JFET input in the Aleph J service manual for which I'm daily visiting his site to see wether it's already online ;)

I just love collecting information to build a great amp!

Thanks for all the input! Cheers, Hannes
 
Well I'm currently wondering about 2 things:

1. gain: the transconductance of the 2SJ109 is a factor of 4-5 lower than of the IRF9610...so what to do about that? Just live with lower gain? C'mon!

2. I started reading the initial Aleph-X thread and already on page 2 I found

rtirion said:
The current source above Q5 and Q7 provides +/- 20 mA. This means +/- 10 mA and 150 to 170 mW dissipation per IRF9610

300 to 350mW for a 2sk389 is a little to much. :)

(2sj109 datasheet says 200mW per package (?) max, anyway constant current can be lowered)

and even more important

Lowering the current in this pair is an option but will wreak havoc on the DC setting. Using a 2sk389 will need a little more work than just a replacement I think.

It would be great if somebody could comment on this!

Thanks a lot for your reply!

Cheers, Hannes
 
Official Court Jester
Joined 2003
Paid Member
h_a said:
Well I'm currently wondering about 2 things:

1. gain: the transconductance of the 2SJ109 is a factor of 4-5 lower than of the IRF9610...so what to do about that? Just live with lower gain? C'mon!

2. I started reading the initial Aleph-X thread and already on page 2 I found



(2sj109 datasheet says 200mW per package (?) max, anyway constant current can be lowered)

and even more important



It would be great if somebody could comment on this!

Thanks a lot for your reply!

Cheers, Hannes

read my pdf
 
Dir Sir,

I read your enlightening pdf a multiple times, but I don't have a clue, how such a difference in transconductance does not lead to a different gain.

My humble greetings, Hannes

EDIT: is this because the current through input pair is also reduced to about the half?

The relative change in current (with a signal on gate) is then not the same, but at least only 2 times smaller.
 
Official Court Jester
Joined 2003
Paid Member
h_a said:
Dir Sir,

I read your enlightening pdf a multiple times, but I don't have a clue, how such a difference in transconductance does not lead to a different gain.

My humble greetings, Hannes

EDIT: is this because the current through input pair is also reduced to about the half?

The relative change in current (with a signal on gate) is then not the same, but at least only 2 times smaller.


yes sir,but you didn't think in in "open loop" and "closed loop" boundaries;
download Aleph J manual (I hope that srvman will be soon available) and read.....

with decreased open loop gain.............with same closed loop gain as in case of Aleph 30....consequence is that feedback amount is decreased........

what you need to look is "enough " xconductance and gain...........not absolute values
 
The reason people don't put 2SJ109s in as the front end of Alephs and Aleph-Xs every day of the week is because it's not a drop-in replacement for the IRF9610. I would have thought this was obvious, but perhaps not.
The 2SJ109 has:
--dramatically lower current capability
--far less transconductance
--less power dissipation
--less voltage capability
etc.
Basically the 2SJ109 only has one thing going for it...it's a nice, dual, low noise JFET. After that, it's all downhill.
Okay...
That said, you can work the problems if you try hard enough.
Cascoding can help with the voltage and indirectly help with the power dissipation.
The lower current capability isn't too bad a problem as long as you're willing to limit the number of output devices you're trying to drive. Depending on how you go about it, this may alleviate the need to cascode as well.
That leaves the transconductance. This is a bit of a problem, as getting the gain up is going to require a redesign. One possibility is to use active loads for the front end. Of course that might leave you with some biasing problems for the back end, but offhand I'd think the feedback could probably handle it. Another possibility is to add another stage. On the one hand, that begins to depart from Nelson's original topology. On the other, it opens several doors. One thing you'll find is that the most obvious front end change is going to be a 2SK389 working into, say, a folded cascode, with the cascode's load being the drive for the output. Yes, there are other possibilities, but that one is short, sweet, and to the point.
Me? I'd go for something like a Mini-A with a 2SJ109 and no cascode. You'll have to flog the rest of the circuit to get it as near perfect as possible, because there's not going to be enough feedback to cure residual ills. It will take a delicate balancing act to get things to run properly. You'll need enough current to drive the output(s), a load resistor value that will react with the front end Iq to give something like 4V at the output Gate(s), and enough gain so that you can still use feedback to tidy things up.
There was another thread recently where someone dropped some JFETs into an Aleph-X (or was it a straight Aleph? I don't remember). Needless to say, they discovered that things weren't quite so simple.
Can it be done? Yes. Sorta. But don't just toss out the idea and expect people to fall all over themselves exclaiming what a wonderful concept it is. It's a limited application thing. You'll have to accept some compromises along the way.
Or you could simply ponder the fact that Nelson released his JFET Aleph as a First Watt model (meaning it's a small amp), rather than as a full-sized Pass Labs circuit.

Grey
 
@Grey: thanks for all your input! I didn't expect all people to become excited about Jfet inputs, I just read the thread you mentioned and the consequence was that I became excited. It seemed that with a few adjustments one could do it (at least for a single output pair per side), so I asked the more experiences guys, that's it. I had hoped you've gone for the IRF9610 to make it as general as possible.

And - as always - things are not so simple :whazzat:

@Nelson Pass: I think I heared you laughing when you wrote your reply :bawling:

Ok seems I've to bury the idea (no scope, no audio precision, no toys to play with such an experimental circuit, not even enough experience)

But thanks a lot for your valuable input!

Cheers, Hannes
 
Nelson Pass said:
The XA100, XA60, and XA30 models all used 2SJ109's.

:cool:


The poop sheets from the manufacturer are, shall we say, scant on interesting details (ahem) and I haven't seen any pictures of the innards, however, I'll hazard a guess and say that the XA30 uses one pair of devices per side per channel, not unlike the Aleph-X. The XA60 could be done with one pair per side, run harder, or two pairs running more conservatively. In either case the rail voltages and drive currents required are easily within the 2SJ109's ratings.
The XA100 strikes me as a more interesting case. We're talking, what, two, maybe three output pairs per side? Hmmm. At least a BL rated part called for. And the rails are still within reason. Yeah, okay, I can see that.
And yes, Hannes, I tried to make the Aleph-X buildable, both in terms of parts availablity and in terms of scaling the thing up. No matter how big you make it, someone's always going to want to make it bigger. The IRF9610 can supply enough current to drive any rational number of output devices. The 2SJ109 is more limited. If you post a schematic based on the '109, you're going to have to post another for larger circuits. It's difficult enough, time-wise, for me to do one circuit. To do two versions of the same circuit really puts me up against the wall. Bear in mind that by the time I'm finishing one circuit and (sometimes) posting it, I'm already overlapping work on a newer, different one. That would mean I'd be messing with three circuits at the same time. Can't do it. Nelson, I estimate, runs R&D on three to five projects at a time, minimum. But he's got more space, parts, resources (read: money), and brain cells than I've got.
Oh, well. I'll muddle through one way or another. C'mon brain cells...get in harness...there's work to be done.

Grey

P.S.: Hannes, I wasn't trying to discourage you from building such a circuit. The only time I actively, consciously try to discourage people from building things is if it's going to be dangerous. If I sound grumpier than usual, bear in mind that I haven't gotten more than four or five hours of sleep per night in the last week. By all means, build the circuit, just bear in mind that it will take some fiddling. As I said above, the problems are workable as long as your goal is reasonable.
 
Thanks for your reply!

Well though I'm learning hard I'm not an electronic's hero, so I want to keep it simple. 38Wrms are perfect for my ears. So only 1 output pair.

The 2SJ109 BL handles these stock rails.

Maybe you could elaborate a bit more in detail what my problem will be in this special case, so that I can think, read and think even more about it.

To come then back and steal again your time!

Cheers, Hannes
 
Ok...I read the first 100 pages of the monster thread :D

I think that thread is the first one I ever saw here on DIYAudio that contained on the first ~60 pages only conflicts between people, followed by some argueing about legal use of patents in DIY and not many technical discussions (apart from Nelsons kind contributions!).

Very interesting to read that already on page 2 (I think) somebody wants that JFET input and then follows most part of technical discussions about why this won't work.

Although I know it better now ;)

However I think I've gathered a lot of bit's n' pieces to feed the wiki!

Once more I want to say thank you to Nelson Pass for his kind support! His helpfulness is so outstanding. Thank you.

Cheers, Hannes

PS: if anyone wants to see pictures of Nelson's staff, Wayne (Xono, Pearl, preamps), Karen (I think she wrote the tips on soldering) and Cyclotronguy, they're also buried in that thread...maybe I say where, hehe.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.