Changes to further reduce noise?

[stonerbox]

Simple input stage. What would you do to clampdown the noise floor even more? Already tried to keep resistor smallish and filter highs. Space limitations forces me to keep parts to a minimum but I could fit a few more if needed.

[antonis]

What's the purpose of R23..?

P.S.
I should make R20 47k, at least..

edit: For lower (shot) noise, incease Collector current..
(or change another BJT with lower r_bb)

[stonerbox]

The point is to prevent influence over the low pass (C15).

What's the purpose of R23..?

P.S.
I should make R20 47k, at least..

edit: For lower (shot) noise, incease Collector current..
(or change another BJT with lower r_bb)

[antonis]

The point is to prevent influence over the low pass (C15).

Let previous effect output impedance to take care of it..

[stonerbox]

The point is to prevent influence over the low pass (C15).

Let previous effect output impedance to take care of it..

?

[antonis]

?

How do you calculate that LPF and what items are involved in..??

[anotherjim]

As Q7 emitter is in phase with Q5 collector, can C15 go from Q7 emitter to Q5 base? Would that improve overall hiss reduction?
Don't forget filtering of the 9v supply.

[antonis]

Miller effect turns 820pF to 13nF which, IMHO, is a big value for C-B capacitance..
(at 10kHz it appears as a 1k2 C-B feedback resistor..)

[stonerbox]

As Q7 emitter is in phase with Q5 collector, can C15 go from Q7 emitter to Q5 base? Would that improve overall hiss reduction?
Don't forget filtering of the 9v supply.

Miller effect turns 820pF to 13nF which, IMHO, is a big value for C-B capacitance..
(at 10kHz it appears as a 1k2 C-B feedback resistor..)

Took the advice from antonis and implemented his changes + moved and upped the cap to retain same cut off.
Noise floor is definitely improved but still not fully ideal.

[Rob Strand]

Unfortunately you can get into conflicting requirements with this stuff.

The original preamp has about 12dB gain and 100k input impedance.

The 68k + 39k form a divider (about x1/4) at the input of the circuit which is never good for keeping noise down.
The 68k + 39k also add noise to the input.
In order for the overall gain to remain constant the 1/4 division means the transistor must amplify 4 times more.
That means the noise will go up by a factor of four and to stick the knife in a bit more the 68k + 39k adds more noise.

Suppose we aim to remove the input divider and still keep the input impedance 100k and overall gain the same:



So we have reduce the noise by a factor of 5 (12.5dB),  and enormous improvement.

What have we lost?
- If the input is driven hard or the preamp clips, the input impedance of the original circuit won't go below 68k.
  If the low noise circuit clips the input impedance could drop to 1.5k.
- If we use low gain transistors the input impedance of the original circuit won't got below 68k and is fairly
  insensitive to the gain.     The input impedance of the low noise circuit is more dependent on transistor gain.
  It's still fairly good though, even gains of 100 will have a reasonable input impedance (because the 1.5k
  emitter resistor is reasonably high.

So the bottom line is they could sound different when clipping.

You can add an input resistor like 47k to the low noise circuit to keep the input resistance high.  The effect on noise
is much less severe than the original circuit since the 47k only forms a divider with the 100k input impedance.
(The resistor values will be a bit different on the first transistor in order to match the gain and input impedance.)

[stonerbox]

Rob you are a life saver! Hm... I always had the impression big (1M) voltage divider resistors would cause more noise? Obviously not the case with Q3+Q4 in your design.
To push the noise -12dB further down is amazing compared to my initial mess.

[Rob Strand]

Rob you are a life saver! Hm... I always had the impression big (1M) voltage divider resistors would cause more noise? Obviously not the case with Q3+Q4 in your design.
To push the noise -12dB further down is amazing compared to my initial mess.

For guitar stuff the two circuit might not sound the same when they clip.   In which case you might need to add some series resistance.

Something I didn't include was the effect of the Cbc cap.   With the original circuit the impedance feeding the input is  68k//39k//470k  = 23.5k.   The guitar impedance doesn't have much effect.   In the low noise circuit the filtering effect of the cap is not dependent on the guitar impedance.    So if you wanted to make the filtering effect more predictable you would need to add some more series resistance (and maybe tune the cap).   That will make the noise worse.   So the trick is to add as little as possible, maybe 10k.   The low noise circuit has higher bias resistors and a higher emitter resistor so the series resistance will have less of an effect than the original circuit.

Adding series resistance is always undesirable but there's always considerations other than noise so you generally have to do some delicate balancing in the design.

As for large resistors causing noise, it all depends.   It depends where the resistor is connected.   For example 1M bias resistors on opamps add very little noise because the noise is shorted out low impedance of the thing driving the input.

You can see here how the same three resistor arranged in different ways has a different affect on the noise,


The key point is keep to the resistance in series with the source as low as possible!