I have a prototype with J113's. Kind of tricky to get the performance, but
when you look at $20 for 2SK170's/J74's it starts looking pretty reasonable.
when you look at $20 for 2SK170's/J74's it starts looking pretty reasonable.
Is it correct understood that you can't make a linear JFET SE amp if you operate it in the constant current region? …..because in that region the JFET in non-linear (square-law)? …..therefor you need matched pairs?
Depends on what you think of as linear. You can get pretty decent distortion
numbers with a single-ended Jfet, and if you want better you can operate
balanced and/or use feedback.
Careful selection and matching is a big help...
I routinely see SE J113's getting down to .001% under line level conditions, and
I have examples of 1 watt SE amplifiers at .01%. Since it is primarily 2nd
harmonic, it can be largely cancelled with balanced operation.
The P channel version of the J113 would probably be the J176, although it does
not have the nicely complementary properties of the 2SK170/2SJ74.
numbers with a single-ended Jfet, and if you want better you can operate
balanced and/or use feedback.
Careful selection and matching is a big help...
I routinely see SE J113's getting down to .001% under line level conditions, and
I have examples of 1 watt SE amplifiers at .01%. Since it is primarily 2nd
harmonic, it can be largely cancelled with balanced operation.
The P channel version of the J113 would probably be the J176, although it does
not have the nicely complementary properties of the 2SK170/2SJ74.
What I was thinking of linear was that a change in Vgs will always give same change in Id regardless if the change is from -1V to -1.5V or from -4V to -4.5V.....so there was a linear relation instead of a "square law" relation. But it may be a wrong way of thinking? ….but if you only use a very small part of the curve it may be linear "enough"? …..I am just trying to understand the curves……and got confused by the "square law" relation between Vgs and Id 🙂
It could be fun with a DIY SE J113 project using 1000 of these pr channel? …..the nice thing is that you don't need a heatsink.....
It's pretty much square law, but if the change in current isn't a high
percentage of the bias then you get fairly linear operation.
And of course to the extent that it is square law, un-degenerated push-pull
operation can perfectly cancel the nonlinearity.
percentage of the bias then you get fairly linear operation.
And of course to the extent that it is square law, un-degenerated push-pull
operation can perfectly cancel the nonlinearity.
Is there an appropriate JFET in a surface mount config? this could be replicated as a production item! Or have a small batch of populated boards made. So that’s easy probably, but then there’s buying 1000ish JFETS so The SE j113 would prob be necessary, which might quash the whole deal
Also Of course loses a bit of the outrageous factor, and knowing Nelson the JFETS are probably matched!
Also Of course loses a bit of the outrageous factor, and knowing Nelson the JFETS are probably matched!
Last edited:
Now this sounds interesting. Digikey/Mouser sell 1000 pcs of J113 for around $100.
And 2000 pcs $170. with no heat sinks needed. I'm in.
Last edited:
Can a SMD JFET dissipate same amount of heat like the TO-92?
It would be good practice to solder 2 x 1000 SMD JFETs…...the PCB could be fairly compact.
It would be good practice to solder 2 x 1000 SMD JFETs…...the PCB could be fairly compact.
Too compact would require a fan, but smaller pcbs would be much cheaper.
And smt packages could be on both sides of the pcb, if the heat allows.
There are also cheaper standard 10k J113 OEM packages that can be shared.
Don't forget you will need a big drive voltage, due to the follower topology.
Last edited:
Only if you are a glutton for punishment 😉
That's what summer interns are for. I think a PCB might be laid out that could support the power dissipation, but it would not save any space vs the TO-92 devices. Each SOT-23 would need some surface area for its drain connection.
The very short "legs" SMD has should give less inductance and maybe capacitance? ….if we have 1000.....it could mean something?
Soldering 2 x 1000 SMDs…..I would give it some days/weeks….e.g. solder 10 every day from Monday - Friday and 20 Saturday and Sunday. It could be a good mental exercise every evening preparing for a good nights sleep…….
Soldering 2 x 1000 SMDs…..I would give it some days/weeks….e.g. solder 10 every day from Monday - Friday and 20 Saturday and Sunday. It could be a good mental exercise every evening preparing for a good nights sleep…….
Ok....project is probably only realistic using TO-92 types. I wonder how many J113 should be purchased to get 2000 for the project. I guess some kind of Idss matching is necessary?
If one JFET fails either during assembly or during operation is it easy to find which one?
If one JFET fails either during assembly or during operation is it easy to find which one?