Resistors in opamps and noise design

[Steben]

I stumbled upon this comparison between very good performing opamps, both jFET and BJT:



Now look at the total noise level and the resistor noise level.
Now wonder why we have huge noise and use so much ohms in too many opamp stages......
Ideally, IMHO, one shoudl design with low impedances and use a discrete input stage to bring the impedance down.
Historically, tube amps use large values.... yet.... they use huge signal levels compared to SS circuits....

[antonis]

Now wonder why we have huge noise and use so much ohms in too many opamp stages......

To avoid loading distortion, perhaps..??

[Steben]

Now wonder why we have huge noise and use so much ohms in too many opamp stages......

To avoid loading distortion, perhaps..??

Loading distortion because we use 10k instead of 1M? Don't think so.
Even loads to 2k are fine with most modern opamps.
NE5532 can drive to 600 ohms.

Look at the Rat and how small the resistors are in leg to ground. This results in huge gains. If we lower the feedback loop resistor we get less gain yes, but much less impedance. It works.

Now on to the SD-1... It uses a BJT buffer in front. This has a couple of ohms output impedance. Not 10k! Then over to the gain section:  The 100k bias resistor is quite large. Here too halving the impedances in all feedback and leg values or even less works perfectly. And then tone control .... oh wait! this has much lower impedances already? Why? Because it works just as good.

But 1 thing is for sure, if resistors are choosen like that, opamp quality will never be crucial in noise control.

[antonis]

 

So, you put the blame on resistors for any noise present in a circuit..??

[Steben]

So, you put the blame on resistors for any noise present in a circuit..??

Well, what I want to know is why the guitar business design(ed)(s) with so much impedance even when the input of a stage is no longer high impedance. High impedance is simply a burden.
In the HiFi world low impedance is "the normal". Because of a hole load of reasons. And the top of it is high impedance does not give "tube tone" at all on itself.
So why?
Two things come to my mind:
1. tube tech was/is high impedance. That's why we need transformers on the first place to bring tube power to a speaker. Low impedance tubes are well ... pretty non existent.
Noise was what it is. People lived with it.
2. Guitars have a "high output impedance"? Well, yes and no..... A single coil can get around 4k to 5k ohms at min if the volume pot is at max. Lower is rare. But ..... it has inductance as well.... this forms a filter with the amp. In a nutshell: not the guitars "ohms" is a problem, but the amp behind which needs high impedance to rather counter the inductance, not simply the resistance.

Tube amps have high input impedance. No one cared in the early days. On top of that, Tube amps are high voltage. the AC signal swing in a tube amp goes well beyond some volts even after the first stage. Noise has more signal to be masked in. In SS design many volts already are unpleasant if clean tone is wanted. Result: signal/noise ratio is lower. This makes resistors AND devices noise important.
To an extend, sag in tube amps help in getting saturation and compression at lower gain settings than a SS amp.

[antonis]

Are we talking about stompboxes..?? 

[Steben]

Are we talking about stompboxes..?? 

Of course! A SS amp is a stompbox with a power amp after.
A Tubescreamer is a preamp channel.
Pop some boxes after each other and you have a cascaded stage design.
etc
etc

ANd in stompboxes the supply is even limited in most cases to a mere 9V.

[antonis]

If so, noise to cost ratio should be more of concern..

[R.G.]

There are some very pointed textbooks on low noise electronic design. Just sayin ...

The quick and dirty of low noise design is to
> use a low-excess-noise first stage
> use a first stage with an impedance appropriate to the source impedance
> use a first stage with enough gain to get the signal up out of the noise contributed by following stages
> use the lowest possible and reasonable resistances for signal and biasing purposes (that is, design for low thermal noise)
> use low-excess-noise resistors (effectively, don't use carbon composition, and prefer metal film or wirewound)
> use low-noise transistor or IC devices
> don't pollute this with poor grounding and wiring techniques

The raw guitar pickup presents a difficult challenge. Just taking a signal from a previous stompbox is not difficult.

The guitar pickup has a resistance of about 4K - 18K as far as raw thermal noise in the source is involved. But the inductance of a guitar pickup - one to four henries!! - makes designing a low noise first stage difficult. You have to run the input impedance high to avoid rolled-off highs (i.e. "tone sucking") and that runs up the thermal noise of your first amplifier stage.

On the other hand, guitar pickups are not particularly low signal levels compared to other signal sources, so there is a leg up there.

[Steben]

Cost?

I don't think this cost more than a normal SD-1 with mojo 4558 (  ):
(in fact it cost about the same)

[Steben]

There are some very pointed textbooks on low noise electronic design. Just sayin ...

The quick and dirty of low noise design is to
> use a low-excess-noise first stage
> use a first stage with an impedance appropriate to the source impedance
> use a first stage with enough gain to get the signal up out of the noise contributed by following stages
> use the lowest possible and reasonable resistances for signal and biasing purposes (that is, design for low thermal noise)
> use low-excess-noise resistors (effectively, don't use carbon composition, and prefer metal film or wirewound)
> use low-noise transistor or IC devices
> don't pollute this with poor grounding and wiring techniques

Well, exactly! Why haven't they read those books?
Given the fact so many pedals have BJT buffers (high impedance) why did they neglect steps 4+ so many times?
A BJT buffer followed by huge impedances?
A BJT buffer equals to about 1 to 2 nV/rtHz.... good opamps have about 5 to 10. The absolute best slightly less.... What they used about as  double....
And boutique pedals just copy bias and basic design ...

So summarised: perhaps there is some prospect in better pedals regarding noise quality rather than inventing the wheel?

[antonis]

Cost?

(Re)design cost..

But from what I've understood, you wish to prove all pedal effects designers unworthy so plz be my guest..

[Steben]

Cost?

(Re)design cost..

But from what I've understood, you wish to prove all pedal effects designers unworthy so plz be my guest..

What I want to prove is just like every taboo: many pedal designers are in a design cocoon, just next to 4558 and germanium adoration.
Reminds of that chap that proved quite correctly that a correct silicon with a correct resistor in series mimics exactly any germanium in curve.
Such guys are known to be neglected soon.

(Re)design cost?
I'll start charging then! 

Guitarists are used to noise. And those are the market.
So what is needed is pure hardcore advertisement and making people belief they need this holy grail, aka less noise (mostly in high gain). Something like "look this is a high impedance, this is a low impedance, hear the difference". That is the only solution. Facts don't matter.

So... I need another SD-1 and swap the resistors.

[swamphorn]

...[Running] the input impedance high...runs up the thermal noise of your first amplifier stage.

While circuits with high input impedance can be more sensitive to external noise, large bias resistors add an insignificant amount of noise to circuits. The thermal noise of a bias resistor is shunted by the smaller source impedance.

[POTL]

Good
The graph shows 2 indicators, with a capacitor in negative feedback in parallel with a resistor and without it.
A properly sized capacitor does a great job of removing some of the noise and not removing excess high frequencies.
Regarding the input impedance, remember an important point, in addition to the output impedance of the guitar, there is also its inductance.
According to various sources, this is 1-5 henry, depending on the pickups.
Guitar inductance + circuit input impedance = low pass filter.
I agree that it is possible to reduce the resistor values ​​within the circuit, sometimes not even adapting it to these changes, but the input impedance reduced by a factor of several can spoil the whole sound.

[antonis]

large bias resistors add an insignificant amount of noise to circuits.

Which noise amount can be further reduced by a small value shunt capacitor..
(e.g. noise induced by a 47k bias resistor is almost the same with the one by 470k//100pF.. - and you have a handy cap for input LPF, EMC or any other purpose..)

[Steben]

I agree that it is possible to reduce the resistor values ​​within the circuit, sometimes not even adapting it to these changes, but the input impedance reduced by a factor of several can spoil the whole sound.

Yes. in case of a Jfet it is minimal though. A jFET is voltage driven, which means there is little current (commonly spoken "none") flowing through a resistor to ground at the gate in a simple input stage. The gate is at ground. Any resistor will make a Ground to ground connection over there. Noise comes from impedance guitar // resistor. jFETs are perfect guitar input devices.
Any resistor bias which invokes a bias current (BJT stages, Opamps, ....) gives most noise.
Which leads to my synthesis: jFET stage followed by opamp with as low resistor values as possible.
The input stage does not need to have really huge gain. The jFET will have some gain smitten on a 0.05 - 0.3V signal and pedals work at 9V typically. Heavy humbuckers will easily overload it which is bad for a universal input stage design and (to lesser degree) high gain will need pentode character of the stage, somewhat throwing away triode sound which some prefer. The job is done: getting rid of the guitar loading at a low noise cost.
The low noise opamp can give the "huge gain" needed for drive purposes, but can achieve this with lower values, since the jFET output is not the inductive load a guitar is. Anything beyond this opmap will be of less influence. But outputs of opamps are low impedance anyway, so why use really unnecessary high resistor values which force the following stage again to have high input impedance etc etc?

When working on the MG10 amp is noticed Marshall at least knew this when they chose 10k pots for the level pots for once instead of +100k.

[antonis]

jFETs are perfect guitar input devices.

Interesting enough..
(I presume terms/parameters like lack of transcondactance, Gate-Source capacitance and increased leakage current for elevated temperature are taken into account for your dictum..)

P.S.
Shall we add "for sources of higher than 50k or so ouput impedance" to your dictum and shake hands..??

[Steben]

jFETs are perfect guitar input devices.

Interesting enough..
(I presume terms/parameters like lack of transcondactance, Gate-Source capacitance and increased leakage current for elevated temperature are taken into account for your dictum..)

P.S.
Shall we add "for sources of higher than 50k or so ouput impedance" to your dictum and shake hands..??

Unless non linearity (adding small amounts of rather even order harmonics) is an issue. (In guitar circuits certainly not!  )
But as mentioned before, the beauty of a jFET is more about the need in building a wall against treble loss than matching a "resistance" while adding tone colouring. Pure resistance does not need to matched, it has to be taken in account in some amplitude loss. Guitars have an inductance and some resistance at full volume, while having mostly resistance at low volume knob settings. A medium level strat with some tens of kohms (volume pot), which leads in a fuzz face to bright clean up! Which on the other hand is .... yes... noisy! fuzzy is not only a name, it is a logic as well  An almost dead low strat is 200kohms.

In all other stages of a guitar to speaker walkdown I actually see less value in a jFET as building block (unless mu amps). Of course in what I see as an ideal line of circuits there are no high output impedances as 100k, 500k and 1M pots.... which are plentiful as output pots in pedals. But those are already problems to be solved, not bright design.

[merlinb]

A discrete JFET might get you an extra 12dB of SNR compared with a humble TL072, but there are not many situations where you need that extra 12dB. Everything else in the guitar signal chain is usually noisier and will swamp your efforts, including the hum picked up by most guitars. 

As PRR might put it, you're tyring to build a better glass-bottomed boat for fishing in your mud-filled pond...