Reduce hiss - improve s/n ratio

[Wonderdog]

I just rehoused a Dano Rocky Road.  I did the unity gain mod.
It has too much hiss.  Otherwise sounds very nice.
It seems to me there must be some very simple way to filter out the hiss. 
Maybe undo the unity gain mod to restor the boost then wire in a low pass filter.

Any ideas as to how I would wire in a simple low pass filter that would only cut out the hiss?

[R.G.]

It seems to me there must be some very simple way to filter out the hiss. 

It does seem that way, doesn't it? Unfortunately, if there was a simple, easy way to filter out the hiss without also affecting the signal, all circuits would already incorporate that.

[Wonderdog]

   True.  

But this circuit has the benefit of a built in volume boost over unity gain when you kick it in.  So there is plenty of signal to work with.  In fact the effect is almost unusable unless you remove a resistor and reduce it back to unity gain.

So I was thinking you could install a passive low pass filter that took out only very high frequencies to remove only the hiss, and any signal loss from the passive circuit might itself reduce the output back to unity gain. Perfect.
How would I wire that in? What cap value should I use?

[R.G.]

So I was thinking you could install a passive low pass filter that took out only very high frequencies to remove only the hiss,

Yeah. That's a problem. white noise, like you seem to be describing, has all frequencies in it. It's guaranteed to overlap your signal. Removing hiss is always a compromise against removing enough hiss to make a difference and not losing too much signal.

How would I wire that in? What cap value should I use?

Well the simple thing to do for a simple low pass filter is to use a series resistor followed by a capacitor to ground. Take the signal off the connection of the resistor and cap.

You're going to have to figure out what resistor and cap values. The frequency where the capacitor is shunting half the signal+noise volume to ground is
F = 1/(2*pi*R*C) where pi is 3.14159..., R is in ohms and C is in farads. Strictly speaking the output impedance of the source which drives this is part of the series resistor, but in most cases it can be ignored; especially in this case where you are going to have to make a personal test of deciding when you've removed enough without it being too much. Start with R=10K,  C = 2200pF and adjust to taste.

[Wonderdog]

Cool!  Thanks   

[head_spaz]

Without a schematic, I can't help much.
But if it uses opamps, you gotta remember that some of them are NOT unity gain stable.
You could be fighting high frequency oscillation. In the MHz possibly. It's not at all uncommon with today's ultra fast chips.
Try adding 0.1uF decoupling caps from the voltage rail to ground-ref on each opamp. (each rail)
You could also add a very small cap in the negative feedback loop of the stage in question... starting with say 50pF or so.
And, you can put it back as a gain stage, but then use a voltage divider at it's output to attenuate the signal down to level you think it belongs. Attenuation will attenuate the noise too.

[Wonderdog]

I have a schematic in a  .pdf file, how do I post it?

[Wonderdog]

Deleted

[Processaurus]

And, you can put it back as a gain stage, but then use a voltage divider at it's output to attenuate the signal down to level you think it belongs. Attenuation will attenuate the noise too.

Exactly, that's the simplest, transparent way to reduce the impact of a noisy process, get the maximum gain going in (before it clips), and attenuate the output.  If you can double the input gain, then you can attenuate the noise by half.