Low Noise Op Amp - Discrete vs. Integrated

[void_error]

First of all, this is no audiophool crap.

A friend of mine asked me if I could build him an improved clone of one of those bog standard guitar overdrive pedals which basically consists of an opamp in a standard non-inverting configuration with two anti-parallel diodes between the output and inverting input with a gain of up to 100 and another opamp stage with a tone control (just a simple adjustable filter).

This thing will feed into a guitar amp also running at high gain. This means the noise will be amplified.

What's the best option here? The original pedal used a JRC4558 opamp but it's too noisy to be run before a high gain guitar amp. Will a TL072 or TL082 do? Or is it better to go with a discrete circuit? The whole thing will be powered with 9V from a wall adapter.

[Zero999]

The NE5532 or its under-compensated cousin the NE5534 which is faster and lower noise but is only single rather than dual and needs a compensation capacitor to behave itself at low gains.

http://pdf.datasheetcatalog.com/datasheet/on_semiconductor2/NE5532-D.PDF
http://www.ti.com/lit/ds/symlink/sa5534.pdf

[void_error]

Forgot to mention: it also needs to have high (>1M) input resistance.

[dannyf]

Will a TL072 or TL082 do? Or is it better to go with a discrete circuit?

Nothing you described seemed to be out of the ordinary and I would think any of those el cheapo opamp will work.

I wouldn't worry about the chip's noise performance at your signal levels.

[dom0]

Forgot to mention: it also needs to have high (>1M) input resistance.

$random COTS J-FET audio op. E.g. TL072

[T3sl4co1l]

Basic principle of noise versus gain:
Put a gain of 10 in front and use 6V instead of 0.6V (forward drop) diodes.  Now the noise (due to the distortion stage) is 10 times lower.  Attenuate output by 10x to fix output level.

Of course, you don't have the dynamic range on a 9V to do that (+/4.5V swing, and that's with an R2R amp), but that's the right idea.  You could use 3V zeners, though they're rather soft (probably soft enough to be okay, considering Vf is even softer).  Or stacks of regular diodes.  Or blue LEDs (~3V Vf -- bonus points for putting them on the top cover as a loudness indicator?).

I have to wonder if it really makes so much difference having 100 gain at the low level; why not 10 or 30?  Or if it's the clipping that matters, ramp up the gain so it's got a hysteresis band instead; now gain around zero input is completely zero, and slightly larger signals are heavily clipped as usual.

Not really much you can screw with, since, as far as I know, guitarists are extremely picky about their "sound".  You can scale things linearly up and down and it won't make much difference, but if you start playing with the range and break points of the nonlinearities, you'll "change the sound".

As for just... low noise op-amps, maybe an LM833 would have better results?  Not much difference in the basic noise level of most general parts, and I expect it will be easily swamped by ambient noise, EMI and stuff like that.

Tim

[Marco]

Put a gain of 10 in front and use 6V instead of 0.6V (forward drop) diodes.  Now the noise (due to the distortion stage) is 10 times lower.  Attenuate output by 10x to fix output level.

Are the diodes relevant for the noise? When they are forward biased the circuit has no gain and when they aren't forward biased they shouldn't contribute much noise should they?

[dom0]

Schematic, please.

(Can we have a meme image for that?)

[Zero999]

As for just... low noise op-amps, maybe an LM833 would have better results?

He said he wanted an input impedance of more than 1M. The input resistance of the LM833 isn't specified on the data sheet but looking at the internal schematic, I suspect it's lower than 1M.

Forgot to mention: it also needs to have high (>1M) input resistance.

$random COTS J-FET audio op. E.g. TL072

That's easy with a FET input op-amp or a source follower before a conventional op-amp but why is high impedance so important? Generally higher impedance op-amps seem to have higher noise too.

[void_error]

Basic principle of noise versus gain:
Put a gain of 10 in front and use 6V instead of 0.6V (forward drop) diodes.  Now the noise (due to the distortion stage) is 10 times lower.  Attenuate output by 10x to fix output level.

Of course, you don't have the dynamic range on a 9V to do that (+/4.5V swing, and that's with an R2R amp), but that's the right idea.  You could use 3V zeners, though they're rather soft (probably soft enough to be okay, considering Vf is even softer).  Or stacks of regular diodes.  Or blue LEDs (~3V Vf -- bonus points for putting them on the top cover as a loudness indicator?).

I have to wonder if it really makes so much difference having 100 gain at the low level; why not 10 or 30?  Or if it's the clipping that matters, ramp up the gain so it's got a hysteresis band instead; now gain around zero input is completely zero, and slightly larger signals are heavily clipped as usual.

Yup, it's the clipping that matters. The so-called high-output guitar pickups can go above of 2V_P-P output at certain frequencies, depending on how hard you pluck the strings.

Not really much you can screw with, since, as far as I know, guitarists are extremely picky about their "sound".  You can scale things linearly up and down and it won't make much difference, but if you start playing with the range and break points of the nonlinearities, you'll "change the sound".

Picky just about describes them. I've been playing around with this kind of stuff a few years back when I used to practice on my guitar on a daily basis and tried pretty much every possible way to clip an opamp's output signal and I'll have to say that "destroying" the signal coming from the guitar and while not torturing your ears is pretty much black magic. And lots of filters.

As for just... low noise op-amps, maybe an LM833 would have better results?  Not much difference in the basic noise level of most general parts, and I expect it will be easily swamped by ambient noise, EMI and stuff like that.

Tim

The LM833 works from 10V... Might as well slap in a charge pump to double (minus 2 diode drops) the 9V input. I've got some TL082s, LM358s and NE5532s lying around so I'll start testing after I get some sleep. Whichever has the lowest audible noise with the input shorted and gain set to maximum (around 118) will be the winner. The NE5532 will probably need an emitter follower to bring input resistance up so the input signal coming from the pick-ups doesn't get attenuated.

Schematic, please.
(Can we have a meme image for that?)

Here's a link. Ignore the bottom part with the jfets, it won't be used.

That's easy with a FET input op-amp or a source follower before a conventional op-amp but why is high impedance so important? Generally higher impedance op-amps seem to have higher noise too.

Guitar pick-ups have high DC resistance roughly between 5k - 20k and high inductance (1H - 10H) so I have to make sure the opamp's input resistance doesn't affect the amplitude of the input signal at frequencies between 50Hz-ish and a few kHz.

[Marco]

Guitar pick-ups have high DC resistance roughly between 5k - 20k and high inductance (1H - 10H) so I have to make sure the opamp's input resistance doesn't affect the amplitude of the input signal at frequencies between 50Hz-ish and a few kHz.

The input impedance of the circuit you linked is only around 300K ... also I'm pretty sure the NE5532 datasheet input resistance is differential, that gets bootstrapped by the feedback so it's not really what the source sees.

[Lightages]

Off the top of my head, the NE553X is the one to go for, as a jelly bean cheapo part for a guitar pickup

[void_error]

Tried it and there's no noticeable difference between the NE5532 and TL072 or TL082, the hissing noise is still there. The LM358 is noisier though but not by much. Will an opamp made out of discrete parts have lower noise? I'm thinking less transistors - less noise.

[dom0]

Will an opamp made out of discrete parts have lower noise? I'm thinking less transistors - less noise.

You really don't want to go that road down for a guitar amp. Especially not if you have no prior experience with transistor amplifiers...

[void_error]

Will an opamp made out of discrete parts have lower noise? I'm thinking less transistors - less noise.

You really don't want to go that road down for a guitar amp. Especially not if you have no prior experience with transistor amplifiers...

I'm not building a whole guitar amp, just a part of the circuit I linked. The whole thing will be connected to the input of a guitar amp which is going to compress the input signal therefore the noise of the circuit will have to be as low as possible. I'll breadboard a discrete opamp for the first stage with the diodes (which also acts as a compressor/limiter) and see if it makes any difference.

On another note, this thing sounds pretty good although it spews a lot of noise out.

[Zero999]

Guitar pick-ups have high DC resistance roughly between 5k - 20k and high inductance (1H - 10H) so I have to make sure the opamp's input resistance doesn't affect the amplitude of the input signal at frequencies between 50Hz-ish and a few kHz.

With a 300k load and a 10H pick-up, the cut-off frequency be 4.77kHz.

Tried it and there's no noticeable difference between the NE5532 and TL072 or TL082, the hissing noise is still there. The LM358 is noisier though but not by much. Will an opamp made out of discrete parts have lower noise? I'm thinking less transistors - less noise.

It's not that simple.

Did you take into account the noise of the power amplifier?

And what about the thermal noise generated by the guitar pick-up? The calculations suggest the thermal noise generated by the guitar pick-up will be higher than that of a TL072.
https://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise

[dom0]

Well you get about 14 uVrms (over 20 kHz bw) noise from the feedback of the first stage and that is amplified by everything that comes after it in the signal chain. Now I have no idea what your signal level is at the input of that stage, so these 14 uV might be the dominant noise source, or not.

Btw. does anyone know what the second stage is for? The very slight peaking in the frequency domain shouldn't be something hearable (except if I messed up decimal places, again)

[void_error]

It's not that simple.

Did you take into account the noise of the power amplifier?

And what about the thermal noise generated by the guitar pick-up? The calculations suggest the thermal noise generated by the guitar pick-up will be higher than that of a TL072.
https://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise

Guitar pick-up noise and guitar amp noise are fixed, can't do anything about them, although vacuum tube guitar amps tend to be quieter than their solid state counterparts.
The whole idea is to add as little extra noise as possible as the circuit I'm going to build will sit between the guitar and amp.

[Rerouter]

ground the input to your circuit, if the noise remains, its the op amp / circuit, if it disappears its from the guitar / cable / connections,

next up, how clean is the supply to your op amp, PSRR is generally pretty good on most op amps, but a wallwart will be spilling out all kinds of crud from what is already on your mains, and could be the noise you are hearing, i'm not trying to push you back to batteries, just sometimes its better to have an LC or RC filter on the input off a plug-pack for a small signal circuit.

[void_error]

ground the input to your circuit, if the noise remains, its the op amp / circuit, if it disappears its from the guitar / cable / connections,

next up, how clean is the supply to your op amp, PSRR is generally pretty good on most op amps, but a wallwart will be spilling out all kinds of crud from what is already on your mains, and could be the noise you are hearing, i'm not trying to push you back to batteries, just sometimes its better to have an LC or RC filter on the input off a plug-pack for a small signal circuit.

Testing was done with the input grounded and the noise is from the opamp. Might slap on a simple linear regulator if I have enough headroom. No idea about the ripple of the wallwart but if it's one "made especially to power guitar effects" it shouldn't be too bad although I would like to get away with a simple RC filter, total current drawn will be under 10mA.

Called the dude I'm making this for and he's running the pedal at minimum gain, which is roughly 12, I tested it at 120... and the volume pot sits at 50-60% so the noise will be much lower. TL072 it is, then. Input and output input/output emitter followers removed and volume pot moved elsewhere, not sure where but anywhere it doesn't change the frequency response of the second stage. I'm trying to get away with the bare minimum here.

[dom0]

Simulation shows expected results ; most noise stems from the first stage. Output referred noise is about 40 µVrms over 20 kHz bw, reducing the feedback values of the first stage to e.g. 470 ? / 56 k? gives 10 µVrms over the same bandwidth (assuming ein of 60 nV/rHz). This quartering of noise voltages equals to 1/16th of noise power and perceived noise loudness.

Note: I removed both the input and output emitter followers (snip at both "1µF BP" caps).

[dannyf]

Output referred noise is about 40 µVrms over 20 kHz bw

What would that be vs. the amplified signal?

[dom0]

Dunno. No one posted what kind of input level this circuit gets, or what the expected output level is. But it doesn't look very high too me, assuming a 0 dBV output 40 µVrms noise floor works out to about 90 dB SNR. So I guess the nominal output level is much lower than 0 dBV...

[Zero999]

The lowest noise op-amps I can think of are the LT1028/LT1128 but they're expensive and overkill considering the pick-up is far more noisy.

As this is for a guitar amplifier, how about cutting the treble to reduce the hiss? You can probably add a filter with a steep cut-off at 10kHz or even 5kHz and it will reduce the hiss without affecting the sound too much.

[Marco]

Tried it and there's no noticeable difference between the NE5532 and TL072 or TL082, the hissing noise is still there. The LM358 is noisier though but not by much. Will an opamp made out of discrete parts have lower noise? I'm thinking less transistors - less noise.

Get an AD8671, it will get you a couple dB less and it's pretty much as good as you're going to get ... the problem is that the source resistance limits how far you can get. You're already far into the realm of diminishing returns. You can shave a couple dB off by reducing the resistance in the feedback network (at the moment it has about the same magnitude as input noise, if you change the lower leg to 470 Ohm and the pot to 50K it will be out of the equation). In the end though all that's is going to maybe get you 6 dB.

Unless you made a mistake somewhere the noise is simply a by product of the 100x gain, even if you could take the amplifier entirely out of the equation it's not going to matter more than a couple dB with that source resistance. What you need to do is back off on the gain at low input amplitudes.

PS. to realistically test the amplifier noise with no guitar input hang a 10K resistor across inputs, not a short.