JFET input for BJT opamps, design choice discussion

[Rob Strand]

The 1/f noise is normally lumped in with the input noise voltage, as shown in the note you linked to. Input noise current is normally considered to be white and is exceedingly small in JEFTs, exactly as PRR stated, and confirmed in the same note you linked to: "Practically  all  JFETs  being  manufactured  today  have i(N) insufficiently low that it can be neglected for source impedance values up to 10Meg".

I actually thought he meant "input hiss" as a *voltage* but he clearly said *current*.  I don't know how I missed that.

[antonis]

FETs aren't by default better than BJT at impedance matching

Of course, but Andy wish to deal with high-Z sources (if I got it right..) and FETs are superior on high impedance matching level.. - although I'm a BJT lover, I've to admit it..
(Source follower output impedance is 1/g_m where Emitter follower output impedance is r_e + R_source/h_FE   -  for 1mA, say, working current, g_m= 4mS & h_FE=200 we get 250Ohms for FET, independed of signal source resistance, compared to R_source/200 + 25, resulting in FET's better impedance matching for R_source > 47k..)

[Steben]

True.

However, I wonder how many players actually use the lower 50% of the volume control pot. At full volume, guitars are never in that +47k region.

[bool]

In old times turning off the input of a 5534 and replacing it with a fet-dual like a defunct 2sk389 was a legitimate design choice. Today it's wasting time.

If you want to pair a hi-z input with a 5532 (a hard-to-beat price/performace opamp), just use a standard jfet gainstage in front of the 5532; and use capacitive coupling to the opamp. Much more consistency this way.

[PRR]

Yes, input current is moot in JFETs in most audio. (The really fat JFETs do less well in condenser mike heads with hundreds-Meg impedance.)

That Siliconix paper fig 6 is of interest.

Jellybean BJTs at two different current, versus a good/common JFET at Idss. Over the range of likely guitar noise resistance, all three devices can give few-dB noise figure, very good. But the BJT optimum is at the very low current of 10uA. This may complicate later amplifier stages, and also super-sonic response (and NFB stability). The JFET gives, for these impedances, same/better NF at *several* mA, offering good slam to the next stage.

OTOH, below 1K Ohms these or other low-price BJTs will give better NF. And over 100k the JFET whups the BJTs at any practical current.

In passing: the classic general-purpose chip opamps run input BJTs at like 20uA, and so incidentally happen to be in the zone for low hiss from e-guitar. Spoiled in '741 by a quad-BJT structure which more than doubles the hiss voltage; also generic chip process makes lower-hFE (higher Ib and Ihiss) NPNs than you can get from a jellybean BJT-only process.

[PRR]

oooh, that sounds intriguing! But I don't quite understand how to connect the JFETs.

A non-inverting amplifier assuming bipolar power rails:

The JFETs will need to be fairly well matched.

A related trick, with BJTs, from a very old (NatSemi AN299) app-note, now TI's snva530, until TI renumbers all their docs AGAIN.

Since there is added gain inside NFB, we normally need custom compensation.

[amptramp]

^^^

The LM118/LM218/LM318 is a good choice for this trick because of its wide bandwidth.  I determined that it had an input voltage noise of 11 nV/(Hz)^0.5 before they published this as their spec so for some uses, it actually works well on its own.

[Steben]

https://electronics.stackexchange.com/questions/164165/would-this-jfetop-amp-circuit-work

[bool]

One more time: If you *must* use a jfet; then cap-couple to the opamp. Otherwise you will spend a lot of time getting bias-related problems under control.

One more point to think about: you can get a fet-input opamp with low-enough noise (like a dual OPA134) for the same-ish price than f.e. a single 2sk170 fet if you look around.. not much to be gained here.

If you want to work with smd, then look for a BF862 if you can find some; not costly. You can pair with a 5532 smd...

[Steben]

Those jfet tl07x ones beat the classic omnipresent 4558 big time. Its all ..... perspective.

[PRR]

> One more time: If you *must* use a jfet; then cap-couple to the opamp. Otherwise you will spend a lot of time getting bias-related problems under control.

The opamp CAN be used to control JFET DC bias. This is often more tolerant of JFET parameters than letting the JFET bias itself.

[bool]

...
The opamp CAN be used to control JFET DC bias. This is often more tolerant of JFET parameters than letting the JFET bias itself.
...

It can do for sure.

But with low 9V rails, that's a bit of a gamble - unless you use somewhat preselected fets with low Vgsoff to leave enough room for "dc correction"; in short this only works practically/satisfactory within a narrow window of rail voltage fluctuation (!), imho not very battery-power friendly. And if you wrap the nfb around the opamp, you need be very carefull with possible oscillation issues when rail voltage changes. (been there, fond memories)

When you factor-in all the hassle, a humble old opa134/2134 is a win-win-win ... winner. Or something like a TLC2272 if that's your thing...

But if you really must roll your own, buy all of the BF862 you can...