White Noise JFET Gain Stage

[lars-musik]

Hi folks,

I just built a "Dr. Roberts" clone. It sounds great, seems to be a new always-on candidate for me.

However, it has quite a high level of whitish noise. I am not really sure if it's the last gain stage or if this one just amplifies the
noise produced in the circuit before it. Swapping of the 2N2222a for lower noise 2n5088 doesn't change anything
The problem is really starting "audioprobe-audible" at the junction of C19 and C20.

I built it using SMT, so there's that.

Do you spot any parameters that would help to reduce noise? I am thinking along the Johnson-Nyquist resistor dependend noise stuff, but I am not really qualified to make such assumptions.

Last gain stage:


Complete scheamtic:






Thanks! 
 

[amptramp]

Get rid of R29 at the input of the µ-amp stage (formed by Q6 and Q7).  The only function it serves is to isolate the amplifier from the preceding tone control stage and to inject noise.

At the input to the circuit on the larger diagram, you have an attenuator formed by R1 and R2 but if you can tolerate the increased signal strength without unwanted overloading, it pays to have as high a signal level as you can stand at the input to increase the signal-to-noise ratio all the way through.  The 1 megohm resistor to ground is unnecessary since R2 serves that function and R2 can be increased to get more signal.

[PRR]

not really sure if it's the last gain stage or

Almost always the first stage. Or if designed by guesses, the stage with the lowest signal level (which means the most amplification after it).

I think you start to hear it at C29 because that's the sensitivity of your listening probe.

Ron is right: R29 is not helping any.

Do you really need this much signal gain?

[Rob Strand]

As a test, short the input side of R23 and R29 to ground and see if the noise is still present.

If the noise is largely unchanged then the noise is coming from the mu-amp stages (Q4 through Q7).

Ideally if you reduce R23 & R29 you should make R30 and R24 the same values then scale the cap values for C15 and C18 up by the whatever factor you scaled the resistor down.

Given you have a charge pump the layout and supply noise shouldn't be overlooked.   Supply noise could get in through R17 then it will be amplified by the mu amp stage.

[R.G.]

It's worth remembering that JFETs have a lot of gain up into RF as well. It's a long shot, but if it's singing at a few MHz, that can be reflected down into audio as an angry-sounding hiss. Might be worth whacking the RF gain with a small-pf cap.

[Rob Strand]

The gain of the first stage (Q1) will make noise at the output of Q1 from the 33k+27k input resistor roughly equal to the noise from the 1M resistors at the inputs of the mu amp.  So if there is an overall gain between the output of Q1 and the output of Q3, which I'm sure there is, then the noise from the 33k+27k resistors will dominate the noise from the 1M resistors at the input of the mu amps.

Things like C8, which boost the high frequencies, are only going to make the noise from the input stage worse.

[Rob Strand]

I had a look ar the circuit. 

The main cause of the noise is the gain pot resistance and resistors around the input of the second together with Q2.  You can't do much about Q2 - changing it doesn't help.  Since the circuit has so much gain in the second and third stages it amplifies that noise from the gain control.

Really the whole circuit needs a redesigned to get the gain pot value down to 100k (or less) instead of 250k.

In order to keep the number of mods down I've limited the changes.

The main idea is to increase R19 so there isn't as much gain from the second stage boosting the noise.  The idea being you can advance the gain control a bit to compensate.  Since C8 interacts with the gain pot I've had to juggle the values of R11, R12, C8 so the response is the same in the mid position.  There's no right values here since you can match only one gain control setting.  An in-between mod would be R11 and R12 both at say 27k and set C8 to 3n9 or 4n7 by ear.

I've also knocked out some noise from the first stage.  The original ckt has a divider followed by a gain stage.  That's not great for noise.  So I removed the divider and reduced the amplifier gain so the overall gain is like it was before.  Then I fiddled with the input values so the input impedance was the same as before and the frequency response was the same.

Here's the summary of the changes:



Really the circuit needs a complete overhaul but that's going to require a lot of mods and then some verification to make sure it hasn't changed the sound or at least hasn't degraded it (improvements are possible in the ears of some).

The main theme of these mods is minimal changes to the parts and minimal changes to the tone.

[lars-musik]

Wow! Thanks all for your numerous suggestions and your time. I bridged the 1M preceding the µAmps - as expected, the volume rises so the output volume can be reduced some. A slight overall improvement in noise floor I'dd say. The same with the inout voltage divider. However, I can't really tell what it does to the sound that is expected from this circuit. I think in future builts, a potentiometer or two trimmers might be a good idea.

For Robs modifications to implement, I`ll have to order some parts. So that might take a while, but I am really looking forward to doing an A/B comparison.

Fantastic forum - as always!

Here's an image of my first builds. At the moment I am fine with the noise as the basic sound of those is really adding some nice grit to my general tone in the rehearsal room.

Cheers!

[Rob Strand]

For Robs modifications to implement, I`ll have to order some parts. So that might take a while, but I am really looking forward to doing an A/B comparison.

As a very simple mod in the right direction you could just change the 220R resistor to 470Rr.   With a direct A/B comparison you might detect a small difference - it would be interesting if you can't!  Remember for an accurate comparison with the 470R resistor you need to advance the gain control a bit to get the same level - that was the whole idea.   You can do the other "2)" mods later.