Well, the 2009 Toshiba version was 0.001% 1W into 8R.
And that was with 28V, 5A bias per channel.
https://www.firstwatt.com/pdf/art_beast.pdf
I guess we expect a bit worse with J113's, not ?
Does that fit your definition of extremely low distortion ?
Patrick
And that was with 28V, 5A bias per channel.
https://www.firstwatt.com/pdf/art_beast.pdf
I guess we expect a bit worse with J113's, not ?
Does that fit your definition of extremely low distortion ?
Patrick
This is from Nelson's article on the optimised J113 source follower :
https://www.firstwatt.com/pdf/art_lxmini crossover.pdf
"without part selection / matching and some additional tweaking it will not give particularly good numbers – distortion figures from about 0.03% to 0.1%.
Matching the devices improves that number, but they still fall short of the Toshiba / Linear Systems numbers, although they are very usable."
So with a load of 3k ohm per J113 pair, I read THD of ~ 0.015% THD at 2.8Vrms, when optimised individually with loading.
Patrick
https://www.firstwatt.com/pdf/art_lxmini crossover.pdf
"without part selection / matching and some additional tweaking it will not give particularly good numbers – distortion figures from about 0.03% to 0.1%.
Matching the devices improves that number, but they still fall short of the Toshiba / Linear Systems numbers, although they are very usable."
So with a load of 3k ohm per J113 pair, I read THD of ~ 0.015% THD at 2.8Vrms, when optimised individually with loading.
Patrick
We want to do a quick sanity check, just to see how bad the real-life J113 is compared to the spice models.
This is because of the bad experience we had using the Onsemi J113 in the UDNeSS, which was eventually replaced by dual 2SK209GR.
The tested is with a J113 Circlotron single cell (1 pair J113 Idss matched)
4Vp-p output 1kHz into 4k ohm (corresponds to 1W into 8R when 500x in parallel)
Rail Bias H2 H3
24V 6mA 0.032% 0.055%
24V 8mA 0.030% 0.049%
24V 10mA 0.028% 0.045%
33V 6mA 0.030% 0.059%
33V 8mA 0.030% 0.057%
33V 10mA 0.029% 0.056%
So these number corresponds those in the red curve above.
The second harmonics should be significantly lower when the J113s would be curve tracer matched.
You can compare them with 2SK209GR numbers we published before :
https://www.diyaudio.com/community/threads/njfets-for-source-follower-applications.329131/
My sincere thanks to a DIYA member in the EU for performing the above tested.
Cheers,
Patrick
.
This is because of the bad experience we had using the Onsemi J113 in the UDNeSS, which was eventually replaced by dual 2SK209GR.
The tested is with a J113 Circlotron single cell (1 pair J113 Idss matched)
4Vp-p output 1kHz into 4k ohm (corresponds to 1W into 8R when 500x in parallel)
Rail Bias H2 H3
24V 6mA 0.032% 0.055%
24V 8mA 0.030% 0.049%
24V 10mA 0.028% 0.045%
33V 6mA 0.030% 0.059%
33V 8mA 0.030% 0.057%
33V 10mA 0.029% 0.056%
So these number corresponds those in the red curve above.
The second harmonics should be significantly lower when the J113s would be curve tracer matched.
You can compare them with 2SK209GR numbers we published before :
https://www.diyaudio.com/community/threads/njfets-for-source-follower-applications.329131/
My sincere thanks to a DIYA member in the EU for performing the above tested.
Cheers,
Patrick
.
Attachments
Did this project get any further?Cascode Single Ended B1 9 cells
This one is intended to be either a preamp or a headphone amp. 9 cells.
2-sided. It was more difficult to lay out but will cost less per area.
New versions of the Beast coming off drawing boards, including EUVL's suggestion of Circlo version.


There is a Jfet beast shaped hole in my life that needs filling 😆
Hi Woofertester,
Very cool idea to give the 1000 FETs beast another shot. I once made a headphone amp with 10x BF862’s per channel based on EUVL’s design. I think it could be scaled up easily with 2SK209’s.
Very cool idea to give the 1000 FETs beast another shot. I once made a headphone amp with 10x BF862’s per channel based on EUVL’s design. I think it could be scaled up easily with 2SK209’s.
That particular circuit (which originated from that of DIYA member easphyx) was originally published here in 2016 (with Spice files):
https://www.diyaudio.com/community/threads/tin-of-fets-headphone-amplifier.298026/post-4895179
It is not particular good in bias stability, PSRR, and therefore must have output coupling capacitor and very low noise PSU.
The circuits posted above in #367 are of course simplified schematics with no details required for practical implementation.
They only serve to illustrate the ideas behind the design.
Cheers,
Patrick
https://www.diyaudio.com/community/threads/tin-of-fets-headphone-amplifier.298026/post-4895179
It is not particular good in bias stability, PSRR, and therefore must have output coupling capacitor and very low noise PSU.
The circuits posted above in #367 are of course simplified schematics with no details required for practical implementation.
They only serve to illustrate the ideas behind the design.
Cheers,
Patrick
I am not very good at this at all but i like to experiment and try and learn..
I tried one experiment with LU1014 but surprising (at least to me) gives dominant third harmonic.
I tried another one but with 2sk209's (just more of them) but this time its dominant 2nd harmonic, also with a lot of higher orders in either cases (i guess it doesnt like the load very much)
Schematic is same as first, just more cells. I got LU1014 model from one of ZM's files and if its good for him its good for me.
Anyway somehow LU's have even order cancellation but 2sk209's dont? The topology is the same so is this a model quirk or something else? I guess i have some reading&research to do...
I tried one experiment with LU1014 but surprising (at least to me) gives dominant third harmonic.
I tried another one but with 2sk209's (just more of them) but this time its dominant 2nd harmonic, also with a lot of higher orders in either cases (i guess it doesnt like the load very much)
Schematic is same as first, just more cells. I got LU1014 model from one of ZM's files and if its good for him its good for me.
Anyway somehow LU's have even order cancellation but 2sk209's dont? The topology is the same so is this a model quirk or something else? I guess i have some reading&research to do...
you can bet that at 15V rail LU is not going to do much good
study Papa's F3, if you want to squeeze goods from it, or series of EUVL's schematics, most of them for cans
though, be reminded that LU, in my head remembered as "always cascoded" is ending as superior pentode part - call it super mosfet - if not having source degeneration , and nice triode if having it, in hefty amount
study Papa's F3, if you want to squeeze goods from it, or series of EUVL's schematics, most of them for cans
though, be reminded that LU, in my head remembered as "always cascoded" is ending as superior pentode part - call it super mosfet - if not having source degeneration , and nice triode if having it, in hefty amount
Well then, here's some more.
I've been playing with this lately, and here's something that is going off to a show
next month….
I've been playing with this lately, and here's something that is going off to a show
next month….
Last edited by a moderator:
It lives!!! Let the speculation begin 🙂
I see footprints to accommodate different pinouts for the 8 rows of 28 devices and 2xN and 2xP in the front. I smell an article coming ...
Matching needed?
Hmmm... what shows are coming up in Nov? Capital Audiofest in Maryland and High End Society in Vienna.
I see footprints to accommodate different pinouts for the 8 rows of 28 devices and 2xN and 2xP in the front. I smell an article coming ...
Matching needed?
Hmmm... what shows are coming up in Nov? Capital Audiofest in Maryland and High End Society in Vienna.
The show Nelson mentioned above is the upcoming Capitol Audiofest. Refined Audio will be demonstrating Nelson's latest JFET Beast with gain suitable for driving efficient speakers and headphones.
In addition, we will be exhibiting a Beast Negative Output Impedance Electrostatic Headphone amplifier:
In addition, we will be exhibiting a Beast Negative Output Impedance Electrostatic Headphone amplifier:
This was at AXPONA 2023 in the Refined Audio room. I've been waiting since April to share this, had to wait until Nelson's announcement. The gentlemen who showed this to me, and allowed me to listen to it, told me the transformers were hand made by one of them. I felt so lucky that they shared all this to me, great folks these DIY people!
Attachments
- Home
- Amplifiers
- Pass Labs
- Beast with 1000 JFETs redux?