diyAudio (
-   Pass Labs (
-   -   The Beast with a Dozen Jfets (

wrenchone 4th May 2013 02:09 AM

The Beast with a Dozen Jfets
The beast is coming... There's a board, and it's partially stuffed.

Next is deciding what jfet to actually use and match up a bunch. I have 3 choices: J310, J110, or my old friend the PN4393. Schematic will follow when the thing is actually stuffed and running. The concept looks very much like a classic 2-stage SE tube amp, but with a follower in between input stage and output to drive all those gates. The amp will be transformer coupled and run on ~+300V and +30V supplies, just like my "Das ist Aber Schade" and "SiC Pupppy" amps. I'm giving it my pair of 3k 15W Transcendar output transformers - I hope the performance will be worthy of the iron.

wrenchone 5th May 2013 12:57 AM

The first thing I encountered rummaging through my stuff today was an unbreached bag of On Semi J310s. Such is fate - I'll see if I can make them work in this context.

wrenchone 6th May 2013 12:18 AM

I'm getting old - I need to squint harder. The jfets I encountered were actually J110s. They were remarkably consistent right out of the bag, so it wasn't much effort to size up two groups of ten (I should have bought a lot more of these while they were still around). Each jfet will be based to 10ma, so bias for each channel will be 100ma, within the ratings of the output transformers. More later, but this amp is shaping up faster than expected. Once things are biased up, I'll post a schematic.

wrenchone 6th May 2013 04:44 AM

1 Attachment(s)
Here's the first rev schematic of the Dozen JFET Beast - a dozen JFETs, as advertised. The overall form will probably stay the same, but some values may get tweaked. The compensation is a guess for right now. It may be interesting to do a version of this that has a pair of yellow-green GaP LEDs in series with each output JFET source instead of the RC network for biasing. Lotsa LEDS - SY might like that....

I need to scare up the big MOSFETs, the output transformers, and a suitable heat sink (I have them, but where?) before I can finish the beast and bring it up with bench supplies.

kasey197 6th May 2013 08:04 AM

cool stuff - one quick question what is the role of R40 ? Also I guess tx secondary needs one leg grounded ?

wrenchone 6th May 2013 09:53 AM

Look up some threads on partial feedback/"Schade" amps (a fair number of threads in the tube section), and you'll see the purpose for R40.

Yeah, I wondered if anyone would catch the grounding omission. I realized I'd left it out after I posted the schematic. I generally run a shielded cable all the way from the output terminals to the summing node so I can park the feedback components right where they belong, which is right at the summing node at the input.

wrenchone 7th May 2013 08:36 PM

I powered up one channel with a dummy fet in place of Q13, with its drain pulled up to the +30V supply in order to check out the biasing of Q2 and jfets Q3-Q12. Bias current on Q2 is as planned (~3ma), and Q3-Q12 are all pretty closely matched at around 10ma apiece. As an added bonus, the voltage across the source resistors is such that they could easily be replaced by a pair of GaP green LEDs. I may do this with one channel and compare the two.

wrenchone 7th May 2013 11:09 PM

1 Attachment(s)
Here's the board, it still needs the big fets, the ins and outs, and a big heat sink. These will follow in due course.

wrenchone 8th May 2013 08:52 PM

1 Attachment(s)
Here's a Beast board populated with the LED biasing option. The bias current in each output jfet is a little high, and as can be expected, consistency of bias current from fet to fet is worse, as there's only ~30 ohms of incremental impedance for the string of two LEDs, as opposed to the 430 ohms used in the RC biasing scheme. Adding some ballast resistance in series with each LED string would help consistency, at the expense of reducing open loop gain. Adding a negative supply to the drain of the P-channel csource follower and its gate resistor would allow dialing in the output jfet bias more precisely, at the cost of a few more parts. Betcha the LED bias sounds tighter than the RC bias, though.

6L6 8th May 2013 10:38 PM

I know there haven't been any comments yet, so it might seem kinda lonely out there...

People, myself included, are watching this with great interest!

BTW, what is the purpose of Q13? I'm still very new to all this...

All times are GMT. The time now is 03:35 PM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright 1999-2017 diyAudio

Content Relevant URLs by vBSEO 3.3.2