Beast with 1000 JFETs redux?

[mkc]

Well, as far as I understand from previous videos with Nelson and also Jon this is not using the Toshiba parts, but the J113.

 

[EUVL]

Really ? Where can I find that ?
Care to post a link ?


Cheers,
Patrick

 

[mkc]

Hi Patrick,

I can find this one where Nelson states that he is working with substitute parts. I believe I have either read or seen another video where Nelson says the same although concluding it is the J113 might be my interpretation given the amount Nelson seems to have bought of them. I can'r remember if Nelson has mentioned it or written it specifically here on diyaudio,

Cheers,
Mogens

 

[Biosonar]

Below is a higher resolution picture of the new Beast of a thousand JFETS. The new Beast uses the Fairchild J113 JFETS. As configured, the Beast can supply 20 watts into 8 ohms with balanced inputs.

Nelson has made smaller versions called the mini-beast and micro-beast. Nelson gave me a 48 JFET Mico-Beast which I use as a headphone amplifier. It provides about a watt into 32 ohms with a single-ended inputs. See below.

 

[mhenschel]

There's this too.

 

[Zen Mod]

aha

which gives me an idea about my next amp, possible due to extremely clever operation of OPLDF:

Beast with Thousand Pass SIT-1

 

[EUVL]

I am pretty sure this is not what Nelson has built.
In any case, if restricted to J113 only, this is how I would build,.
Fully balanced. 500 pairs per channel. 7mA 35V per JFET.

And very low distortion, if you believe in Spice.
e.g. 0.1% 64Wrms 1kHz into 8R.
0.001% 1Wrms into 8R.

Of course the devil is in the details.
For example DC bias drift. But can be solved.




Patrick

 

[EUVL]

Here you go, schematics in pdf.
Of course lots more info in Spice file.

R11,12, V4 only needed once per board, but then 499R and not 20k.
R1~4 have to be one set per JFET pair.
JFETs have to be matched for Idss for equal current sharing.
Will get hot at 250mW each.

RL1 is of course the load (speaker).


Happy matching,
Patrick

 

[EUVL]

And yes, you can have lower power without heatsinking.
For example 24V supplies and 4mA bias per JFET.
That will give you 20W into 8R at 0.1%.
Dissipation is 100mW per JFET, can be left as stand-alone.


Patrick

 

[Nelson Pass]

As the Beast is actually 4 banks of parallel jfets per channel you can wire it up as you like.

 

[EUVL]

So how well did you match between individual JFETs ?


Patrick

 

[Nelson Pass]

Matched for Vp to .1V

 

[EUVL]

Thanks for the hint.


Patrick

 

[vince76]

I'm interested in building a beast-like amp

 

[EUVL]

First step is to match 2000x J113 in TO92 package.
Plenty available at Mouser or Digikey.
Cost about 280€ for 2000x.

And that will keep you very busy for a long time.




Patrick

 

[vince76]

First step is to match 2000x J113 in TO92 package.
Plenty available at Mouser or Digikey.
Cost about 280€ for 2000x.

And that will keep you very busy for a long time.




Patrick

I'm considering to do this.
but I'm wondering:is j113 the only candidate?
Another thing I'm considering is: 64W is plenty of power for me, maybe I could scale down the thing to less jfets in parallel or it is better to have less bias current?

 

[EUVL]

If you want to use it without heatsink, then 100mW per JFET is about right.
That means 24V 4mA bias for the circlotron.


Patrick

 

[Nelson Pass]

The Vp distribution of the J113's is between 0.9 and 2.8 volts, so you will need a serious population to get that quality of match.

Me, I have about 300,000 and I pay my son to measure them - keeps him well employed during Covid....

 

[Zen Mod]

.... I have about 300,000 ....

Muuuuhahahahahaaaaaa....

 

[Blues]

I would like to build a mini or a micro-beast and I have enough matched j74s and k170s (muuwahahaha!)...Nelson is the official writeup/circuit board coming soon?