Two band Parametric EQ

[Rob Strand]

Not having here those parts I tried with 5.1kΩ 1.6kΩ 100nF
Did you expect this? How would noise change with lower value capacitors? Or what values will diminish most noise?

Well it does work.  I can see and measure the lower noise.  I don't know what volume setting on the PC produces the same noise level that you hear in the room with your amp.   To me the two sample seem to have different parametric EQ settings.  It actually sounds bad!

It could be improved by changing the 1.6k to say 470 ohms.

One problem I see is you have LM348 opamps.  They are quite noisy.  There are eight of those which increases the noise quite a bit.  A TL074 would help.

To make any further improvement you would need to run at 15V-0V-15V supplies (I'm not 100% sure what you are running now).  You would use pre-emphasis *and gain* on the first stage then you would have de-emphasis and attenuation after the last stage.   The idea here is it keeps the signal as high as possible through all those opamps.   A gain of 5 might be a good start.

One thing you should try is connect only one band at a time.  Set to the same position on each band and make sure both EQ circuits are roughly producing the same amount of noise.  If not one band has a problem.

Beyond that those circuits can oscillate.  Sometimes oscillations can make a circuit go noisy.  If you haven't debugged oscillations before it can be hard and on a circuit like that it's even harder.

Yet another thing to try is to connect the de-emphasis network to the output of the first opamp.  Then see how much noise you get.  That will do two things:  it will tell you the noise level of one opamp.  If that is already high you know you are in trouble and you need to change the opamps.  The second thing it does is it  simplifies the circuit and removes the chances for oscillation to occur. 

You should add a 100uF across each power rail.

[antonis]

And a 100nF ceramic as close as possible to every op-amp power supply shouldn't do any harm..

[Guitar_Duderino]

Well it does work.  I can see and measure the lower noise.  I don't know what volume setting on the PC produces the same noise level that you hear in the room with your amp.   To me the two sample seem to have different parametric EQ settings.  It actually sounds bad!

See information added to my previous post. New sound file added with no change of potentiometers... They are set to that graphic. Just capacitors were changed.

One problem I see is you have LM348 opamps.  They are quite noisy.  There are eight of those which increases the noise quite a bit.  A TL074 would help.

I am using one 9V battery with Max1044 charge pump (not cause of noise, since same noise is heard with one other battery instead of Max1044) and opamps are TL074.

It could be improved by changing the 1.6k to say 470 ohms.

Noise recording with 5.1kΩ 10μF electrolytic and change from 1.6kΩ till 5s to 270Ω (had none of 470Ω) till end.
Again Guitar (turned off) > Equalizer > Amplifier (with strong distortion).

https://www.dropbox.com/s/6h7g8w8buxbu7dh/1.6k-270.m4a?dl=0

Greatly diminished noise!

And from 270Ω to 220Ω to 100Ω:

https://www.dropbox.com/s/9cl49z3tum6zdh8/270-220-100.m4a?dl=0

From these with 100Ω least noise is heard...

Returning to 270Ω here transition to addition of passive highpass filter and to addition of lowpass active filter (parallel capacitor added to retroactive resistor in input opamp):

https://www.dropbox.com/s/ryv90wz3egz93tm/non-highpass-lowpass.m4a?dl=0

Here change of 5.1kΩ to 8.2kΩ to 100kΩ:

https://www.dropbox.com/s/2yyct6qnz1i0w1c/5.1k-8.2k-100k.m4a?dl=0

100kΩ greatly diminishes noise, but by opamp saturation clean sound is all distorted.

Here 10μF 5.1kΩ and change from 270Ω to 100Ω:

https://www.dropbox.com/s/autyyv5d7rxh7cg/270-100.m4a?dl=0

100Ω diminishes noise, but clean sound begins being distorted.

So now I have this at input and output:



(input filters in green, output in blue with second band to its high extreme)

Emphasis was to add gain to treble, but adding gain not just to treble also diminishes noise...
And gain limit here seems to be opamp saturation, right? How do we know that value?

You should add a 100uF across each power rail.

I have electrolytic capacitors of 10μF across power supplies of Max1044 (see my first posts with BYOC circuit, which I follow). You mean electrolytic capacitors of 100μF here would be better?

[Rob Strand]

Emphasis was to add gain to treble, but adding gain not just to treble also diminishes noise...
And gain limit here seems to be opamp saturation, right? How do we know that value?

There isn't a single correct set of values.  However there are definitely values which cause trouble without giving much improvement.

This diagram shows what the circuit is doing.   The de-emphasis circuit at the output simple undoes what the pre-emphasis circuit does so you end-up with an overall flat response.




The key points to remember are
- You cannot remove the noise of the first opamp.   This sets a base-line for the minimum amount of noise.   
- The whole idea is the first stage boosts the signal (and it's own noise) by some gain.  If you make that signal strong enough the noise added by the following circuit is small in comparison.

If you have a large cap the circuit is essentially boosting all frequencies.   That always gives the least noise *BUT* it's gives the highest chance of  overload, which is what you have found in your tests. 

You have to find a compromise between noise and overload.  You want to boost the noise frequencies a lot so they get reduced and you want to boost the signal frequencies the least to prevent overload.

So the first thing to choose is how much gain.   Normally you don't want more than about 20dB gain (x10).  That should be enough to make the noise of the first stage to be stronger than about 20 opamps in the following circuit.  That means you shouldn't need R1 less than 470 ohms when R2=5.1k. 

Once you set the gain the ratio fH/fL is fixed.  For a gain of 20dB that ends up with  fH/fL = 10.

You don't gain much by putting fH above 20kHz it just creates problems without much impact on noise.  If you reduce the gain and drop fH to say 5kHz the noise isn't that much different to the 20kHz case.   Keeping the gain the same and dropping fH to 5kHz isn't the same thing that does reduce the noise *but* at the risk of overload.

The frequency fL is what determines overload.   For guitar you would normally put it around 200Hz to 1kHz.  If you don't have to problems with overload you can set it to a low value.  In that case the amount of gain is the thing that determines overload.   

Your 5.1k + 270R + 10u puts fL = 3Hz and fH = 60Hz so it's pretty much boosting 60Hz and above at the full x20 gain.  That's the whole audio band.  You can only do that if your opamps have a high power supply voltage so they don't clip.

So if you look at what is normally OK some of your experiments are going outside the above recommendations.

One thing that makes me suspicious is the gains greater than 20dB are giving an improvement in noise.   That means the EQ circuit is adding a lot more noise than we expect from 8 opamps.   Now some EQ circuits make the noise worse than what it needs to be.  I know that circuit *does* do that but I'm still surprised how much.

Another possibility is your TL074's are noisier than normal.   At this point I'm going to assume they are OK.  If you do a test to see how much noise comes from one opamp that would be good.  At least then you know what level of noise is the minimum and if it is OK (or even noticeable).

If the EQ circuit does make the noise worse than the opamps it would require modification and that's going to be a lot of work.  The filter structures would need to be revised and there's only so much you can do.

At this point my only suggestion would be to play around the with gain and the cap to reduce the noise as much as possible with overload.   Try to find the least gain (A) and the highest fL point which works.  Remember also when you boost bands the risk of overload increases, especially if boosting low frequencies say below 400Hz.

[Rob Strand]

While I don't understand why high gains are reducing the noise,  I think I understand why making fL and fH low frequencies helps.   The distortion effect probably has a lot of high cut so it's boosting the mids and low frequencies more than the highs.   Any noise in those bands get amplified greatly by the distortion effect.

I had a quick look at the noise.   It turns out that using the 3.9k to control the amount of boost is bad for noise.
I did some quick tweaks to the circuit to remove it.   I don't claim these are optimal changes
however the changes decrease the base-line noise by 6dB (ie half).  The noise at the peak
of the boost is decreased by 3dB.

I've tried to keep the amount of boost and the Q's about the same.  Also I made the Q pot
produce the same Q when it is centered compared to the original design.



What I haven't checked is if this circuit is more or less prone to oscillations.

Even with the mods, one band of that EQ produces noise which is four times higher than a single opamp.  That's the equivalent of 16 opamps in cascade.    So two of those stages is like the noise of 32 opamps.