JFET input for BJT opamps, design choice discussion

[Fancy Lime]

Hi there,

we have recently been discussing an interesting gain stage design:
https://www.diystompboxes.com/smfforum/index.php?topic=122749.0
The original design the devices discussed there are derived from the TC Electronic Integrated Preamp. Original schematic found here:
https://www.diystompboxes.com/smfforum/index.php?topic=89712.0

The main point of the extra transistor before the opamp seems to be noise reduction and impedance matching. For guitar use we would likely fare better with a JFET input as suggested by Paul here:
https://www.diystompboxes.com/smfforum/index.php?topic=122279.msg1152716#msg1152716

I am trying to understand the pros and cons of different ways of adding a JFET to the front of a BJT-input opamp. Doing that should bring down noise when feeding the stage with high-z signals. But only if done right. Principle schematics for discussion:



In addition to the "weird" stage discussed in the posts mentioned above ("A" in the schematic), I can think of two other ways that are about the same level of complexity and can be made to basically do the same thing, namely lots of non-inverting gain and high input impedance. "B" is a JFET common drain amplifier followed by an inverting opamp stage. "C" is a common drain buffer with a non inverting opamp stage. A and B save one resistor compared to C because they use the output impedance of Q1 as the input resistance of the opamp. Other than that, the circuits are equally simple.

I was wondering if someone can enlighten me with regard to the pros and cons of each of the designs.

1) How would they compare noise wise? Or, if there is no general answer (when is there ever?), how do I calculate the noise performances from the obtainable info? Does it even matter? My previous tests have shown that I could never tell the difference between different opamps in simple non-inverting stages with high gain, because the noise of the input stage was always swamped by the noise from the pickups.

2) What other advantages and disadvantages do these stages have? I feel like "A" seems a bit like a cascode in that it presents a lowish resistance to the drain of Q1, which should help reduce the Miller effect, right? Not that that matters to our puny audio signals, though. "C" has the least finicky biasing but is the only one that cannot be made to have gains lower than 1. R2 probably needs fine tuning in A and B, which would also influence the gain by changing the output impedance of Q1. "A" may not be suitable for single supply 9V operation, if I understand the biasing requirements correctly. What else?

Thanks and cheers,
Andy

[amptramp]

You might have some trouble getting schematic A to work because the operating point of the JFET is affected by the position of the feedback control.  The DC level at the output of the op amp has to rise or fall in order to keep the JFET biased correctly and this may affect the amount of signal you can put through it.  When you add a stage within the feedback loop, you may run into the need for additional compensation since there would be an additional phase shift for the feedback to take into account.

Since the JFET in schematic B acts as a very high impedance at the output, the stage gain is set by the drain resistor.  Typical numbers are in the several megohm region and this is usually well above the resistance of the drain resistor, so the gain will be quite predictable.

[Rob Strand]

I am trying to understand the pros and cons of different ways of adding a JFET to the front of a BJT-input opamp. Doing that should bring down noise when feeding the stage with high-z signals. But only if done right. Principle schematics for discussion:

You can argue this with general principles.

Case C:  Pros: prevents the noise currents from a BJT opamp from contributing noise.
              Cons: Introduces a small amount of distortion and signal loss.

Case B:  Pros: prevents the noise currents from a BJT opamp from contributing noise.
              The first stage provides gain.  This improves the s/n of BJT opamp as the
              input signal is higher.
              Cons:   Introduces distortion.

Case A:  Pros: Attempts to achieve the same goal as B but by putting the JFET in the feedback loop
              it reduces distortion.
              Cons:   It's a pain to get the biasing to work out.
                          (Circuits B and C separate the biasing issue altogether).

[PRR]

> ways of adding a JFET to the front of a BJT-input opamp.

If only they made JFET-input opamps. Maybe someday.

[antonis]

Haven't something handy for the moment to post but there are quite enough textbooks puplished by well establised brands (Texas Instruments, National Semiconductors, Fairchild etc..) with fair in-depth ways of noise analysis, Andy..
(although, I personally prefer to pretend immune to noise ear than mess up with all those #%^#@ formulae..)

[amptramp]

There is another way to get a FET input to an op amp if you choose something like the LM318 or the NE5534.  With both of these op amps, you can take the inverting and non-inverting inputs to the negative supply, cutting off the input transistors, and add a differential pair of input transistors across the compensation / balance inputs.  National Semiconductor put out the LH0061CH, a hybrid with JFET's replacing the inputs of an LM318 and I know about this because I was the first customer in Canada for these devices back in 1973.

The advantage of doing it this way is since you are not adding a stage, just replacing one, you do not have to change any compensation or add compensation to an op amp that does not need it.

[Fancy Lime]

> ways of adding a JFET to the front of a BJT-input opamp.

If only they made JFET-input opamps. Maybe someday.

Well, that's not entirley the point. For many applications a JFET input opamp would probably be best/simplest when designing something new in 2019. But what I started this thread for, is trying to understand why people in the past have done things one way or the other. And maybe if there is a legitimate reason to want to do it like this still, in certain situations. Given that any JFET input opamps that have a noise voltage approaching that of the humble old NE5532 are not only 10-20 times as expensive as the NE5532 but also not available from any of my usual sources unless I buy in bulk. Especially considering the noise characteristics of something like a 2SK170 JFET or even a chap 2SK117. Sure, there is always the TL072, which is usually good enough for our purposes. But as I said: I'm not so much after practical design advice here but rather trying to understand some basic theory in order to then be able to derive the practical recommendations by myself.

Haven't something handy for the moment to post but there are quite enough textbooks puplished by well establised brands (Texas Instruments, National Semiconductors, Fairchild etc..) with fair in-depth ways of noise analysis, Andy..
(although, I personally prefer to pretend immune to noise ear than mess up with all those #%^#@ formulae..)

For certain simple stages, like a single non-inverting opamp, yes, there is plenty of material. But as soon as it gets any more complicated, all the sources I found get rather thin-lipped and use a lot of simplifications which only apply to the exact example circuit they discuss. Nice if you want to build exactly that, useless for anything else and particularly unhelpful when trying to compare several different designs. Also: The fact that I still have no clue what the name of topology "A" is, does not help with searching for information about it. "JFET input opamp" obviously does not return what I am looking for. Neither do any of the dozens of variants I tried so far. I'm happy with hard dry theory. My problem is not that that is too complicated in the design notes and whatnot but that they tend to dumb it down to the point where it is no longer universally applicable but only works under certain assumptions. I'm sure there are good books on this, I just don't know any. Would love some recommendations!

Cheers and thanks,
Andy

[Fancy Lime]

There is another way to get a FET input to an op amp if you choose something like the LM318 or the NE5534.  With both of these op amps, you can take the inverting and non-inverting inputs to the negative supply, cutting off the input transistors, and add a differential pair of input transistors across the compensation / balance inputs.  National Semiconductor put out the LH0061CH, a hybrid with JFET's replacing the inputs of an LM318 and I know about this because I was the first customer in Canada for these devices back in 1973.

The advantage of doing it this way is since you are not adding a stage, just replacing one, you do not have to change any compensation or add compensation to an op amp that does not need it.

Hi Ron,
oooh, that sounds intriguing! But I don't quite understand how to connect the JFETs. Sources to pins 1 and 8, respectively and drains to Vcc via resistors (assuming a NE5534)?
Thanks,
Andy

[PRR]

...National Semiconductor put out the LH0061CH, a hybrid with JFET's replacing the inputs of an LM318...

Moot because we can't buy one today, but: LH0061 isn't a JFET-input part.
http://pdf.datasheetcatalog.com/datasheets2/61/61489_1.pdf

An all-BJT wideband high current TO-3 for servos, yoke and capstan drivers. It says "audio" but was much too expensive for that.

[Steben]

FETs aren't by default better than BJT at impedance matching, nor at noise reduction. They require less passives and behave more like tubes, that's true.

(brace for impact)

[PRR]

> FETs aren't by default better than BJT

FETs have "no" input hiss current. This is moot in some cases, critical in others.

Define Your Application Before You Talk About Hiss!!

In audio we face hiss from 0.1 Ohm ribbon mikes and 500 MegOhm condenser mikes. Troubles and solutions will be very different.

Here is old but still insightful data:
https://www.onsemi.com/pub/Collateral/AN-6601.pdf.pdf

The distinction between NPN and PNP is moot. Theory suggests slight advantage to PNP but geometry and processing makes far more difference. The 2N930 is an *ancient* part, low-hiss but equalled by most recent jellybeans. The shift of OSI with BJT current 1mA vs 0.1mA is a real effect: trade Vhiss against Ihiss.

JFETs traditionally have higher hiss voltage but "no" hiss current. So less-good for low impedance sources, more-good for high impedance sources. However a few (rarer, costly) JFETs challenge BJTs even in 100r MC phono apps. However quality JFETs of all types get scarcer every month.

[merlinb]

1) How would they compare noise wise?

'A' normally has gain in the JFET so the EIN will basically be the voltage noise of the JFET alone. Linearity will be the best of the three. It is self servoing, but could be tricky to configure in a 9V pedal environment.
'B' is a kinda like a cascode. The noise will again be that of the JFET, assuming R4 isn't very large, and linearity will be the worst of the three circuits.
'C' effectively eliminates the current noise of the opamp but has less than unity gain from the JFET. EIN will be that of the JFET plus that of the opamp (added quadratically of course). This is likely to be the noisiest of the three, analytically speaking.

[merlinb]

The fact that I still have no clue what the name of topology "A" is, does not help with searching for information about it. "JFET input opamp" obviously does not return what I am looking for.

I've been calling it a JFET augmented opamp, but I too don't get much from Google using that name.

[Rob Strand]

FETs have "no" input hiss current. This is moot in some cases, critical in others.

Actually they do have input hiss.  If it were only thermal noise (and shot noise) we wouldn't see 1/f noise.   The 1/f noise comes from surface effects.
https://www.vishay.com/docs/70599/70599.pdf

Interestingly researchers in the field have been arguing about the true noise mechanisms in MOSFETs for about 40 years or so!

[Rob Strand]

I've been calling it a JFET augmented opamp, but I too don't get much from Google using that name.

Here's a National Semiconductor reference calling it a compound amplifier,



I had a feeling it was called a compound amplifier.  I don't know where I first heard the term it would have been 30 years ago, not from that reference.

[amptramp]

I've been calling it a JFET augmented opamp, but I too don't get much from Google using that name.

Here's a National Semiconductor reference calling it a compound amplifier,



I had a feeling it was called a compound amplifier.  I don't know where I first heard the term it would have been 30 years ago, not from that reference.

As you can see, R1 and C1 in series are required to ensure stability because otherwise you would have enough phase shifts to cause oscillation so you need the compensation network to ensure the gain goes below -1 in magnitude before the phase goes past 180 degrees for the inverting input.  But if you look at the input stage of a NE5534 or an LM318, taking the inputs to the negative supply shuts off the input stage entirely.  Then you add a differential JFET input with the drains connected to the balance and bal/comp inputs and turn it on.  Since there is no extra stage as in the example above, there is no need for any more compensation than you would have with the original devices.

BTW I was sure the LH0061 was FET input but maybe that was another device.  National dropped all their LH (hybrid) op amps later on, so they does not show up in any later databooks.

[Rob Strand]

As you can see, R1 and C1 in series are required to ensure stability because otherwise you would have enough phase shifts to cause oscillation so you need the compensation network to ensure the gain goes below -1 in magnitude before the phase goes past 180 degrees for the inverting input.

Yes bypassing the first stage altogether is better for stability.   The other way I've seen those compound amps stabilized is with an RC network between the drain/collectors, basically a lag compensator.  I believe the feedback method shown is better.

BTW I was sure the LH0061 was FET input but maybe that was another device.  National dropped all their LH (hybrid) op amps later on, so they does not show up in any later databooks.

They were fairly expensive and specialized devices at the end of the day.  Cool parts nonetheless.  I've got a couple of LH0070/0071 reference chips.  They are accurate to all digits on a 4.5 digit DMM.  One of the chips is quite old (IIRC 70s?) so it has exceeded the long term stability spec in the datasheet by a large factor.

[merlinb]

FETs have "no" input hiss current. This is moot in some cases, critical in others.

Actually they do have input hiss.  If it were only thermal noise (and shot noise) we wouldn't see 1/f noise.   The 1/f noise comes from surface effects.
https://www.vishay.com/docs/70599/70599.pdf

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".

[merlinb]

There is another way to get a FET input to an op amp if you choose something like the LM318 or the NE5534.  With both of these op amps, you can take the inverting and non-inverting inputs to the negative supply, cutting off the input transistors, and add a differential pair of input transistors across the compensation / balance inputs. 

That's really interesting!

[merlinb]

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.