Discret JFET preamp quieter than TL072?

[tempus]

Hey all;

I'm building a preamp for my electric guitar to use with an amp simulator on my computer. I want to boost the signal to close to line level, which I've deduced will require a gain of about 1.7. Since a high gain ampsim will boost the noise level considerably, I'd like to keep the noise level from the preamp as low as possible (I suppose that's good practice in any case). Since I don't need much gain, I could get away with using a single JFET. Would the JFET produce less noise than a TL072 (my other option)?

Thanks

[jonny.reckless]

I suggest a J113 with a 1k source resistor and 2k2 drain resistor assuming a supply of 9 - 18VDC. You'll get a voltage gain around 2. This will be about 8dB quieter than a comparable gain preamp using a TL072, assuming a 10k source impedance. You can get lower noise with a low noise op amp such as the NE5534 or LM4562 of course. I happen to like the character a JFET front end imparts to a guitar signal chain - a bit of second harmonic warms up the tone a little. Keep the gate stopper resistor small to maintain low noise - I find 1k is about right for the first stage.

[amptramp]

You don't have a circuit or device selected in your example, but if you had the JFET used in the TL072 input stage operating at the same source current as in the TL072, a single amplifier stage would give you 1/(SQRT 2) of the noise of the differential input stage used in the TL072 because the input stage is differential and the noise adds as the square root of the sum of the squares for two FETs in a differential amplifier.  In fact, there is a TL062 which operates at lower current but has more noise and it is not suitable as a guitar amplifier but is often used as an LFO because the low current drain reduces power supply glitches, so the noise goes up by a factor of the square root of the current but the gain goes up directly with current as the current through the JFET increases.  The resistive noise of a JFET is approximately 1/gm or the reciprocal of the transconductance.

The NE5532 was designed as an op amp for use in mixer boards.  The voltage noise is low but the current noise is high, so a 5532 works best with source impedances of 10K to 20K.  This is far too low for a guitar.  The current noise is multiplied by the impedance seen at the device input and added as a square root of the sum of the squares to the voltage noise to form a total noise.  FET input op amps usually have very little current noise - enough that it can be ignored for most designs at high impedance.

[tempus]

Thanks for your replies, and sorry for the lack of info in my OP.

The TL072 would be in a noninverting configuration with input resistors of 232K and a feedback resistor of 390K. It would also be powered by +/- 12v.

if you had the JFET used in the TL072 input stage operating at the same source current as in the TL072, a single amplifier stage would give you 1/(SQRT 2) of the noise of the differential input stage used in the TL072 because the input stage is differential and the noise adds as the square root of the sum of the squares for two FETs in a differential amplifier.

amptramp, I'm sorry but that I'm having a hard time understanding that. It looks like you're saying that a single discreet JFET circuit should have less noise than the 072 because there is only 1 JFET rather than 2. Am I correct?

[amptramp]

Thanks for your replies, and sorry for the lack of info in my OP.

The TL072 would be in a noninverting configuration with input resistors of 232K and a feedback resistor of 390K. It would also be powered by +/- 12v.

if you had the JFET used in the TL072 input stage operating at the same source current as in the TL072, a single amplifier stage would give you 1/(SQRT 2) of the noise of the differential input stage used in the TL072 because the input stage is differential and the noise adds as the square root of the sum of the squares for two FETs in a differential amplifier.

amptramp, I'm sorry but that I'm having a hard time understanding that. It looks like you're saying that a single discreet JFET circuit should have less noise than the 072 because there is only 1 JFET rather than 2. Am I correct?

Correct.  If you had a pair of JFETs with a noise level of 10 nV/(SQRT Hz) and you put them together in a differential stage, the stage would have a noise level of 14.14 nV/(SQRT Hz).  The noise levels of uncorrelated noise sources add as the square root of the sum of the squares.  The current noise and voltage noise are uncorrelated, so they also add as the square root of the sum of the squares.  I once designed an amplifier with a noise level of 0.64 nV/(SQRT Hz) - but it was optimized for a 50 ohm input source impedance, in this case, a HgCdTe optical sensor.

[antonis]

I might be the last person who could be entitled to speak on behalf on noisy/noiseless curcuits but I think there are some pretty thorough practical dissertations about "noise", like Douglas Self : https://epdf.pub/small-signal-audio-designb2a423050b68c4e826f4ec45b9a6637520099.html & https://epdf.pub/audio-power-amplifier-design-handbook58585b1e8538616db8ff33e83647d0aa26964.html..

P.S.
I only wish to have the time & willingness to entirely study some of them..

[jonny.reckless]

The NE5532 was designed as an op amp for use in mixer boards.  The voltage noise is low but the current noise is high, so a 5532 works best with source impedances of 10K to 20K.  This is far too low for a guitar.

I use NE5534 and NE5532 on the front end of guitar amps and pedals all the time (although I prefer the sound of a discrete JFET input stage usually as it adds some character). Most guitars pickups are around 10k DC resistance with approx 2H of series inductance. The noise attributable to NE5534 current noise is about the same as the input voltage noise for a typical guitar pickup impedance at 1kHz, and together they are about 14dB quieter than a TL072 which is almost entirely voltage noise. I've done side by side comparisons in the front end of high gain distortion pedals and the reduction in audible hiss when you replace a TL072 with an NE5532 or LM4562 is dramatic. You do have to be careful with resistor values and DC offsets due to input bias current but a cap after the first stage is all that is usually needed. The NE5532 is low noise, cheap, readily available in every package, has vanishingly low distortion, can drive low impedance loads easily, has a wide phase margin, clips gracefully, and is tolerant of imperfect decoupling. It's one of my favorite op amps for audio applications. The only thing you have to be wary of is input bias currrent, but designing around this is not hard. TL072s are great when you absolutely have to have picoamp input bias currents, but this is rare in most electric guitar and audio applications. Some people like TL072 in distortion pedals because they like the sound of the 072 recovery from saturation when it's clipped hard - that's a subjective thing, but objectively it's a noisy op amp for use in stomp boxes such as overdrive, distortion or compressors which are likely to go in front of a drive pedal.

+1 on Doug Self's small signal audio design book. Lots of useful information and practical ideas there.

[jonny.reckless]

I once designed an amplifier with a noise level of 0.64 nV/(SQRT Hz) - but it was optimized for a 50 ohm input source impedance, in this case, a HgCdTe optical sensor.

That's really impressive, like the self noise of a 100 ohm resistor at room temp. I've never managed to get anywhere near that performance. Any tricks you are able to share?

[amptramp]

I once designed an amplifier with a noise level of 0.64 nV/(SQRT Hz) - but it was optimized for a 50 ohm input source impedance, in this case, a HgCdTe optical sensor.

That's really impressive, like the self noise of a 100 ohm resistor at room temp. I've never managed to get anywhere near that performance. Any tricks you are able to share?

As I said, it was optimized for a low-noise optical sensor and since it was a HgCdTe photoresistive sensor for thermal imaging, it had a source of current to bias it as well as an amplifier.  The gain stage was based on differential 2N4405 PNP transistors with emitters fed by a self-biased JFET current source feeding an LM318 op amp and compensated for the combined passband.  I derived from the noise calculations that the LM318 had an input noise of 11 nV/(SQRT Hz) before National published any noise figures and when they did publish it, it was the same number.  If I could have found NPN transistors that had as little noise, it would have been a lot simpler - tie both LM318 inputs to the negative supply, shutting off the input transistors in the LM318 and tie the collectors of the low-noise input NPN transistors to the bal/comp inputs at the collectors of the input transistors and you have a composite amplifier that uses the same compensation as the LM318 with no changes.  National did this themselves with the LH0061CH hybrid where they put n-channel JFET's in the LM318 first stage and got a high-speed JFET-input op amp (and I was the first customer in Canada for them).

[diffeq]

tie both LM318 inputs to the negative supply, shutting off the input transistors in the LM318 and tie the collectors of the low-noise input NPN transistors to the bal/comp inputs at the collectors of the input transistors and you have a composite amplifier that uses the same compensation as the LM318 with no changes.  National did this themselves with the LH0061CH hybrid where they put n-channel JFET's in the LM318 first stage and got a high-speed JFET-input op amp (and I was the first customer in Canada for them).

You can do the same with NE5534, with its balance and comp/balance pins.

[tempus]

Thanks again for all the replies.

I suggest a J113 with a 1k source resistor and 2k2 drain resistor assuming a supply of 9 - 18VDC. You'll get a voltage gain around 2. This will be about 8dB quieter than a comparable gain preamp using a TL072, assuming a 10k source impedance. You can get lower noise with a low noise op amp such as the NE5534 or LM4562 of course. I happen to like the character a JFET front end imparts to a guitar signal chain - a bit of second harmonic warms up the tone a little. Keep the gate stopper resistor small to maintain low noise - I find 1k is about right for the first stage.

I don't have a J113 lying around, but I do have J111, J201, and J176. I've tried a few different things with the J111, with no success (I'm assuming this is at least in part due to the J111's I_DSS of 20 vs. the J113's 2 mA). I'm still trying to figure some of this stuff out and will do more experimenting, but using the J111 as a buffer and increasing the gain at the interface yielded a somewhat quieter signal vs an 072 just in a listening test. Not an enormous difference though, which makes me wonder if it might just be easier to just use the 072. I do want try it out properly though (like with all the required gain coming from the JFET) to find out for sure.

Is it possible to design something within my required parameters using any of the JFETs I already have? I'm still learning the JFET stuff, but I'm sure that a big consideration is going to be my desired output Z, which needs to be 1K or less. I'm kind of assuming that the output Z will be set primarily by the source resistor, which will also be dependent on the I_DSS of the device. Is this correct? Also, the J176 has an I_DSS more in line with the J113, but it is a P channel device. I have  -12v available as a supply option. Would using a P channel JFET simply be a matter of connecting everything the same way but using a negative voltage supply?

[jonny.reckless]

The J111 has a relatively large VDSoff, so you need a bigger negative bias voltage to make it work. This makes it impractical for use in 9V circuits but if you've got 30V to play with they are great. You can use a J201 just fine. You'll probably want something like 4k7 in the source and 10k in the drain for that, depending on the specific part you use. Ideally you want the drain voltage to sit around 2/3 of VDD.

Here is a useful app note on JFET biasing which you might find helpful:
https://www.vishay.com/docs/70595/70595.pdf

[tempus]

Thanks again jonny for your informative reply. Thanks to you and others posting here I'm learning more and more (hopefully I'll be self sufficient soon!).

I was re-reading this thread, and am also interested in the 5532 post you made. I had considered this as an option, but I thought that the input Z would be too low for a guitar pickup. Is this not the case? What opamp configuration do you use (like just a basic non-inverting setup?), and what value of resistor for R_in? Looking at the datasheet, I see that the 5532 has an input R of 300K, so a 1M resistor would give and input R of 230K, which I'm pretty sure is what Steve Vai uses in the Legacy amps.

I have to confess to never actually trying one of these opamps because I seem to remember posts by (possibly R.G. and/or PRR?) saying that the bigger problem with pickups and their inherent high output Z was with current noise which the TL072 handled better. What are your thoughts on that?

I'm not necessarily abandoning my learning more about JFETS, but if the 5532 is going to be the quietest option, I'll probably just go with it. Definitely gonna try one on the breadboard though...

[jonny.reckless]

A TL072 will definitely be noticeably noisier than a NE5532 in a guitar amplifier circuit, all other things being equal. The reason is that the 1 megohm input resistor appears in parallel with the guitar pickups from a noise point of view. An NE5532 with a 1M resistor on its input and the input socket left open will make a lot of noise. This noise will mostly be gone when you plug the guitar in. The actual impedance the op amp input sees looks like around 10k DC resistance in series with 1 or 2 Henries inductance, as the pickups shunt the 1 megohm.

If you are dealing with piezo pickups or other very high impedance sources then current noise is a major consideration and you need a FET input, but for regular strat type pickups it's not.

I typically use 1M input resistor with a 1k series "blocker" to prevent parasitic oscillations due to cable impedance and stray cap from the guitar lead. If you are using a series input capacitor to block DC, make it big enough that so that the noise is shunted by the pickups at all audio frequencies. Bear in mind that input bias current is also a consideration, so you may need a DC blocking cap after the first stage to remove offsets caused by input bias.

The NE5532 does have quite a low input resistance, in practice I find typically around 400k in non-inverting mode as a unity gain buffer, but you can get around this by bootstrapping if you really want high impedance input for AC. I don't usually bother. The effect is just to tame some of the shrill high end off the pickup, which also helps with hiss a bit.

Having said all this, my favorite sounding (non tube) guitar input stage is still a self biased JFET source follower. For your application I would still recommend a common source discrete JFET using a J113 or similar.