Recommend JFETS for audio

[tk@halmi]

Tom,

No trouble at all, I will let you know when I am ordering.
"Well matching" is a bit vague goal for me. With transistors that live
on the same substrate, i.e. dual transistors, matching is better than
2% to 3% for most parameters. This is rather difficult to achieve
through matching individual devices, IMHO. I have decided not
to obsess with it and settled for 5% or less. For the devices with
the same designation that come from the same batch you get a good
number of matches. This is one reason for ordering a large number
of the same devices at once. The input pair is perhaps the most
critical in this regard so you can always reserve the best matches
for that. Then use the rest for VAS and current source tasks.

My experience with complementary devices has been exactly
the opposite where getting smaller lots frequently helps you get better
matching between P and N channel transistors on the long run.

There is a bit of caution on using the 2SK30 instead of the 2SK170.
The gain of the 2SK30 is less than a tenth of the 2SK170 or the 2SK389.
This leaves little room to "linearize" the input pair and degeneration if used
may reduce its gain to nothing. If you account for this in your design
you will be very happy with the 2SK30 as well.

Cheers,
Tamas

 

[TomWaterman]

Thanks Tamas, you're a star!

I was really hoping on using the 2SK389 for my input pairs, a couple in parallel in fact, but if I can match the 170s close enough I will use them.
I'll hopefully get some 389s before they are all gone.

I'm still learning a lot and have no concrete design as of yet but want to play around to learn more...

As a side note, have you played with LEDs a constant current sources? I remember reading Winston O'Boogie saying he used them?

Anyone know where he is BTW?

Cheers Tom

 

[Guest]

[quote author="TomWaterman"]As a side note, have you played with LEDs a constant current sources? I remember reading Winston O'Boogie saying [/quote]
LED current sources have low noise.
Here is very high end gramo preamp which uses it in discrete OP-AMP.
This preamp have synthetic resistor for pick-up damping (load)
It has been developed by my lecturer dr. Sykora in 1980-th. and
have noise near theoretical minimum with 2SK389 or 170 s.
(All op amps are 5534, some RVHP numbering here is obsolette.)

(you must copy this link to the browser)

http://mujweb.cz/www/xvlkxvlk/sykora.gif

xvlk

 

[TomWaterman]

Hi xvlk,

I heard they made a good low noise source - I also read over at diyaudio.com that the agilent series of L.E.Ds have lower noise than most?
That could be a myth but its something I will investigate further.

One question, the opamps, what are they used for, some kind of filter circuit or complicated DC servo?
That circuit is a little above my knowledge right now.... :grin:

Cheers Tom

 

[Guest]

[quote author="TomWaterman"]¨
some kind of filter circuit or complicated DC servo?

That circuit is a little above my knowledge right now.... :grin:
[/quote]
Yes, properly engineered circuits are above knowledge of everybody.
It is pity that no every audio circuits are properly engineered,
schematics are then monotonous and reading of it boring.

Some OP amps in feedback are DC servos, but first feedback OP amp
is invertor with gain which forms some as Miller effect in capacity.
But it is with resistance and this megaohm resistor is transformed to
normal pick-up 47 kOhm load. But current noise is still determined by
its physical megaohms and then this "synthetic" resistance have lower
noise than normal 47 kOhm.
Van der Ziel called it "resistance with noise ratio <1 "

Because in pick up is dominant rather big inductancy at H.F., normal 47 kOhm
resistance forms with this highpass and this resistance can be in the highs
dominant in noise. Synthetic resistance have smaller noise and input of
the FET is still dominant.

About noise properties of HP leds I never heard, but generally
how cleaner process to make semiconductors is, than lower
some kinds of noises (mostly 1/f and popcorn) are.
And HP can have acuratte process for producing measurement technology.

xvlk