FirstWatt J2

6L6

Moderator
Joined 2010
Paid Member
There has been a rule of thumb that if H2 is dominant and the rest are in a smooth depending stair-step below that, the amp will sound good, natural, non-fatiguing and overall very good.

Jean Hiraga's observation was that H2 only needed to be 60db down from the fundamental, which in the grand scheme of things isn't all that low.

My observations and experiments in this regard hold this rule of thumb to be true. I strongly prefer a dominant 2nd to a significantly lower overall distortion where H3 is bigger than H2
 
Last edited:
Please consider this modified input version. :)

I am confident it will deliver the goods.
Not a lot of space on my own amp...probably new PCB is needed.
 

Attachments

  • DIYJ2 bootstrapped bootstrapped-compound CFP.asc
    6.8 KB · Views: 68
  • J2 bootstrapped bootstrapped-compound CFP.png
    J2 bootstrapped bootstrapped-compound CFP.png
    22.6 KB · Views: 893
The one and only
Joined 2001
Paid Member
There has been a rule of thumb that if H2 is dominant and the rest are in a smooth depending stair-step below that, the amp will sound good, natural, non-fatiguing and overall very good.

Jean Hiraga's observation was that H2 only needed to be 60db down from the fundamental, which in the grand scheme of things isn't all that low.

My observations and experiments in this regard hold this rule of thumb to be true. I strongly prefer a dominant 2nd to a significantly lower overall distortion where H3 is bigger than H2

Nice to see Julian Hirsch (Stereo Review) validated after years of jeers.

:cheers:
 
Lots of parts in that J2.I'd find it hard to imagine it better than my version of the F7.After many iterations I have finally settled on "the one".Alas,I can only imagine a J2 in my dreams as I do not know anyone with one and cannot afford one (i'd have to start saving my coin and do the diy).However at the present time I have no interest, from what i've read it is a very special beast.
 
Here's another question for you guys:

In my sim of a 40-watt J2 variant I end up with the JFET drain resistors in "spice flames" if I short the output when there's a signal on the input.

I had to increase the JFET source resistor values to get a more civilised response. This is more degeneration/local feedback, right? I thought this would reduce the gain, but it doesn't seem to (at least not until I run out of voltage). What's the draw-back of more degeneration?

attachment.php
 

Attachments

  • j40-schematic.jpeg
    j40-schematic.jpeg
    123 KB · Views: 1,694
Quite amazingly, after 10 years, the FW J2 still remains rather secretive.
The recently published, reverse-engineered schematics from post #1151 to #1154 do not quite represent the original.
Above all, rail voltages and bias are, as so intended, much higher than the original.
This results in a different distortion signature, as well as very high dissipation on the power devices.

If one wants to copy and paste 1:1 of the original, the best deal IMHO is to pay the 4000USD and get an original.
4 matched Semisouths, 4x 2SJ109s and 4x 2SK170s will easily costs you 500USD.
By the time you have bought all the other components, and a decent case, you are most probably close to 2000USD.
And then you still have to build, not sure it will match the original performance, and have no resale value.

But if you want to build because you love DIY, then it is a different matter.
So let's have a look to see if we can come close without a 2k budget.

AND, I hope Nelson would not object. :)


Patrick
 
So let's take one of the exisitng DIY J2 schematics, and change them to come close to the rail voltage and bias of the originals.
Here I chose +7-24V and 1,2A bias, giving a dissipation of ~30W per power device, which is quite safe.

Keeping the paralleled JFET frontend, we get ~0.03% THD for 1W into 8 ohm at 1kHz.
Very close to the figure published in the Stereophile review.
I chose to have a bandwidth of just over 100kHz, but you can easily change that with the cap on the feedback loop.

You need to unzip the file and place the +.lib file in the same directory as the spice files to run.


Patrick

.
 

Attachments

  • J2 Clone Dual IP Amp.asc
    6.9 KB · Views: 274
  • J2 Clone Dual IP Freq.asc
    6.9 KB · Views: 169
  • SemiSouth-1.zip
    1.4 KB · Views: 200
last time i bought jfet was around $16/pair from alweit so it will only cost you $70, while ccs part can use irfp240 then you will only need to get 2 single SS let say $100 each. with other small parts maybe around $50.

in total maybe you'll only need $300 excluded psu & heatsink which presumed you already have. this project won't be the interest of new builder anyway, just for greedy boyz :)

worth to wait for official schematic? why not but in the mean time you can build any available copy design just like F7 situation.

for new builder then it's preferred to go on available kit on the store
 
Can we only use a pair of matched JFETs instead of 2x 2SJ109s for the frontend ?
Some one mentioned in post #1084 of a bias of 8mA.
So must be possible.

But does that make a lot of difference, e.g. in bandwidth ?
Apparently not, at least in Spice.
So perhaps Nelson will tell us one day why he uses parallel devices.


Patrick

.
 

Attachments

  • J2 Clone Sing IP Amp.asc
    6.1 KB · Views: 158
  • J2 Clone Sing IP Freq.asc
    6.1 KB · Views: 108
But we are still using 4x matched Semisouths, which is a major cost factor, if they are available at all.
And as we found out, they are not very consistent in their manufacturing tolerances as e.g. Toshiba.
https://www.diyaudio.com/forums/gro...tracer-matched-pairs-quads-4.html#post5892320
(Post #188).

So can we use some other devices for the modulated current source instead ?
Post #1154 was already suggesting that, using the commonly available IRFP240.
But these have much lower transconductances than the Semisouths and hence will change the distortion spectrum.

Instead, we want to try out 2SK3497, which has a Yfs of about 7S.
Their high positive tempco would not matter here, due to the clever active
-bias control using the opto-coupler.

Again, in Spice, there are hardly any differences in distortion spectrum and bandwidth than a Semisouth CS.


Patrick

.
 

Attachments

  • J2 Clone K3497 Amp.asc
    6.1 KB · Views: 212
  • J2 Clone K3497 Freq.asc
    6.1 KB · Views: 116
So here you are.

It is not the original.
If that is what you want, I am sure Nelson will be happy to sell you one.
But what is proposed here would be fun to build without breaking the bank.

And finally the disclaimer :
I have only simulated and not built.
But if I were to build, this is probably what I would try out first.
One can always add parallel JFETs and swap the CS MOSFET later on, when all is working.


Cheers,
Patrick
 
I got the leftover board from a member who has compared DIY version with matched SS vs Original one and he admitted that can never matched the original one :D maybe that's why i only needed to pay the board very cheap

so if you want original sound, buy original J2 hehe i'm just happy enough with what i have even though not sure what to do with leftover 1pair SS on my drawer
 
So let's take one of the exisitng DIY J2 schematics, and change them to come close to the rail voltage and bias of the originals.
Here I chose +7-24V and 1,2A bias, giving a dissipation of ~30W per power device, which is quite safe.

Keeping the paralleled JFET frontend, we get ~0.03% THD for 1W into 8 ohm at 1kHz.
Very close to the figure published in the Stereophile review.
I chose to have a bandwidth of just over 100kHz, but you can easily change that with the cap on the feedback loop.

You need to unzip the file and place the +.lib file in the same directory as the spice files to run.

Patrick
.

Thanks for the info.
I set mine to 1.5A since I do not tolerate heat well :eek: and installed fans. Now I ordered low noise fans. Anyway, I'll probably sell it or lend it to a friend here, since I don't quite use it...

Cheers,
M.
 
  • Like
Reactions: 1 user