Lumitron (MuTron III) low volume noise issue

[Quiddel]

Hello,

I recently built the MuTron III clone by Aion called Lumitron, and it does what should do, everything works as designed and it sounds amazing. Awesome project!

Unfortunately I have a noise/crackling/distortion issue with very low volume signals (e.g. when barely touching the strings). I uploaded a sample to soundcloud to illustrate what I mean. Just to be clear, there is no background noise, just when a very low signal is applied, it distorts. If the signal falls below the threshold, the distortion is gone. The issue is independent of the supply, I also tried a battery.

https://on.soundcloud.com/RWYTk

Here is the link to the demonstration: https://aionfx.com/app/files/docs/lumitron_documentation.pdf

And the voltages for ICs, Z1 and the vactrols:

Input voltage
9.0V

Vactrol left (LED side)
(+ square) 1.69V
(- round) 0.05V

Vactrol right (LED side)
(+ square) 1.69V
(- round) 0.05V


Z1 (BZX55 C12)
square 8.7V
round 0V

IC1 (RC4558P)
Pin    Voltage
1    0V
2    0.02V
3    0.02V
4    -8.3V
5    0V
6    -0V
7    0.03V
8    8.7V

IC2 (RC4558P)
Pin    Voltage
1    -0.02V
2    0V
3    0V
4    -8.3V
5    0V
6    0V
7    0.03V
8    8.7V

IC3 (RC4558P)
Pin    Voltage
1    0.45V
2    0.15V
3    0.08V
4    -8.3V
5    0.09V
6    0.09V
7    1.75V
8    8.7V

IC4 (LT1054 CP)
Pin    Voltage
1    8.7V
2    4.6V
3    0V
4    -3.87V
5    -8.3V
6    2.5V
7    1.4V
8    8.7V

I couldn't find a major issue in the voltages, did I miss anything? Anyway, I reflowed all solder joints just in case, but it did not improve the issue. I got the BZX55-C12 zener in the musikding kit, instead of the 1N4742A, I compared the data sheets and it looks good to me, but could that be an issue?

Thanks a lot for the help,

Greetings,

Matthias

[antonis]

Does it happen irrelevantly of GAIN pot setting..??

[Rob Strand]

There's so many possible causes.  You will need to dig a bit deeper.

- opamps crossover distortion:  maybe not at this point
- noise from DC/DC converter getting into audio path: possible
- something to do with the rectifier IC3A (possibly linked to DC/DC converter noise).

I would try temporarily lifting the IC1A side of C11 then connecting it to ground.
If the noise/effect is still present maybe it's noise in the audio path.
If not, connect C11 back to normal then maybe look at the rectifier.

I'd be tempted to add a small cap across R15.  Start with 100pF to see if it fixes it at all.
If it helps maybe try to back it off to 15pF to 22pF or so.
While this may appear to fix the problem the true problem could be DC/DC converter noise.
(perhaps even noise getting into the DRIVEA switch wiring).

Be interesting to know if things change with the up/down switch.

[Quiddel]

Thanks a lot for your replies! I finally got back home and could test what you suggested.

The Up/Down Switch indeed reverses the issue. With "Up", that's what I described before, there is crackling/distortion with low signals, e.g. barely touching the strings. When the drive is set to "Down", the crackling appears when the input signal is high, it sounds pretty much like a fuzzy distortion.

The gain pot does affect the issue, but it's a bit complicated. The volume of the noise goes down and up with increasing or decreasing the gain correspondingly, but way less than the actual signal. Also, when the Drive is in "Up", a lower gain also leads to more distortion (because the signal is more in the low volume range where it distorts I guess). With Drive at "Down" it gives me more noise at higher gain settings (as the distortion kicks in at higher signals in Down).

Desoldering C11 did remove the noise, so it's not in the audio path. I put it back in.

Unfortunately a 100pF cap across R15 did not change anything, the noise is unchanged.

Thanks again for your help, I really appreciate it!

Greetings,

Matthias

[Rob Strand]

Good progress.

The Up/Down Switch indeed reverses the issue. With "Up", that's what I described before, there is crackling/distortion with low signals, e.g. barely touching the strings. When the drive is set to "Down", the crackling appears when the input signal is high, it sounds pretty much like a fuzzy distortion.

The gain pot does affect the issue, but it's a bit complicated. The volume of the noise goes down and up with increasing or decreasing the gain correspondingly, but way less than the actual signal. Also, when the Drive is in "Up", a lower gain also leads to more distortion (because the signal is more in the low volume range where it distorts I guess). With Drive at "Down" it gives me more noise at higher gain settings (as the distortion kicks in at higher signals in Down).

Check your values for R16, C12, R17.   Maybe even replace C12 or tack solder a larger value like 22uF across it.

What it seems like is the audio signal is passing through the rectifier but not getting filtered enough.   Maybe your optos are fast ones and all that together means the audio is modulating itself.   That will cause distortion *for sure*.

No guarantees but at worst a step closer.

[Quiddel]

I finally found the time to check your suggestions!

Unfortunately it did not resolve the issue. R16, C12 and R17 had the correct values.

I tack soldered a 22uF cap across C12, and also another 47uF cap, so 73uF in total but the noise is still there, although the noise did have less high frequency content ("got muffled") with increasing capacitance. So I took the big hammer and put 220uF across. The noise was barely audible, with a hefty high cut, but the envelope filter itself obviously did not work properly anymore.

The cutoff of the RC lowpass with R16 at 330 Ohm and C12 at 73uF would have been at 6Hz and the noise was still well audible in the range of 1kHz. I know it doesn't correspond to a direct lowpass in the audio path, so I am not sure what to make out of it anyways.

Thanks again for the help! I am a bit at loss right now how to proceed. I tried swapping the RC4558 ICs, but no change. Could it be a faulty LT1054?

Edit: Found another LT1054, no change.

Best greetings,

Matthias

[Rob Strand]

tack soldered a 22uF cap across C12, and also another 47uF cap, so 73uF in total but the noise is still there, although the noise did have less high frequency content ("got muffled") with increasing capacitance. So I took the big hammer and put 220uF across. The noise was barely audible, with a hefty high cut, but the envelope filter itself obviously did not work properly anymore.

The cutoff of the RC lowpass with R16 at 330 Ohm and C12 at 73uF would have been at 6Hz and the noise was still well audible in the range of 1kHz. I know it doesn't correspond to a direct lowpass in the audio path, so I am not sure what to make out of it anyways.

Thanks again for the help! I am a bit at loss right now how to proceed. I tried swapping the RC4558 ICs, but no change. Could it be a faulty LT1054?

You have done well with the debugging so far.   At this point it seems like there is a problem between the envelope circuit and the opto drive.   Something is modulating the opto's, possibly at a fast rate.    It's definitely turning into a more difficult problem.

At this point I'd be trying more invasive debugging.  Also I would pull out the signal generator and oscilloscope and start looking at the waveforms around the output of the opamps, and especially around IC3B.

I would also lift the IC3A side of R16 (330 ohm).  Connect a 22k pot + 27k resistor in series then wire those to the floating end of R16 and +VA.   Vary the pot to sweep the filter manually to see if there is any weird behaviour throughout the range.

Some simpler ideas:

Something to try would be to add a 100pF cap from the output of IC3B (pin 7) to the inverting input of IC3B (pin 6).   If IC3B is oscillating that should stop it.

Check the values of R9 and R10.   From your pic one looks OK but the colors of the other looks a bit weird in your pics.

I'm not overly fond of connecting the opto LEDs in parallel, as shown on the schematic.   A series connection would be far less problematic.  (Something weird could be happening here.)

There is still a chance power supply ripple is the cause.  That's where an oscilloscope would help.

[m4268588]

• Does -VA (-9V) fluctuate when a signal is input?
• Is it improving by changing the envelope follower's Op-Amp to xx1458? (or something similar)

[Quiddel]

Is it improving by changing the envelope follower's Op-Amp to xx1458? (or something similar)

Thanks for both your suggestions! I actually found a LM1458 in my stash, no change. However, I just out of curiosity I swapped IC3 for a TL072 and voilà, the noise is completely gone.

As far as I understood, the main difference between the RC4558 and the TL072 are input impedance and slew rate (RC4558: 1V/µs; TL072: 13V/µs). Someone has a guess why the TL072 now fixed the issue?

For transparency, I did replace C12 permanently with a 22uF cap, as I mistakenly thought before that putting it across C12 solved the issue. I don't know if the TL072 also would have solved the issue with the original 4.7uF at C12. However, I decided to leave the 22uF in for good, as my enclosure starts to take damage from taking it apart too often. Soundwise it doesn't change much imho.

However, thanks a lot for your help! I am happy to finally put it on my board, the Mutron octaver felt lonely without its evelope filter cousin already.

All the best,

Matthias

P.S. Here is how it looks btw, happily dwelling next to the amazing L4 Bass Preamp by aion.

[Rob Strand]

As far as I understood, the main difference between the RC4558 and the TL072 are input impedance and slew rate (RC4558: 1V/µs; TL072: 13V/µs). Someone has a guess why the TL072 now fixed the issue?

I think the opamp is oscillating.

This suggestion I made earlier might make the other opamps work.

Something to try would be to add a 100pF cap from the output of IC3B (pin 7) to the inverting input of IC3B (pin 6).   If IC3B is oscillating that should stop it.

The TL072 could have fixed the problem because of some other reason.   Upto to you what you want to do.   The TL072 might fix it because it has a lower input capacitance and that helps avoid oscillation.   The 100pF cap tries to make the circuit work without any risk of input capacitance (or other effect) causing oscillation.

[Quiddel]

Thanks, I think it's a classic "if I had more time issue", I would definitely try it. However, my focus is more on that it works atm and I can use it.

I hope to find time soon to revisit this and I'll for sure put the 100pF across pin 7 and 6 with the 4558. :-) Can't promise though, because there is quite some things coming up in my day job.

Thanks again!

Matthias

[duck_arse]

I'm not overly fond of connecting the opto LEDs in parallel, as shown on the schematic.   A series connection would be far less problematic.  (Something weird could be happening here.)

that looks to be a translation/tracing error, and probably should be corrected. interesting that the opamp driving the leds wrongly also seems to be the fix.

[aion]

I'm not overly fond of connecting the opto LEDs in parallel, as shown on the schematic.  A series connection would be far less problematic.  (Something weird could be happening here.)

that looks to be a translation/tracing error, and probably should be corrected. interesting that the opamp driving the leds wrongly also seems to be the fix.

I first saw it done in parallel in Madbean's Nautilus/Naughty Fish going back to 2012 or so. Looking at RG's original Neutron, the dual LED method (called option 3) looks to have the LEDs in series, so the parallel configuration may have originated with Bean.

What's the potential downside to putting them in parallel - is there more of a potential for unequal driving of the filter if the LEDs are slightly different in specs? I've gotten a ton of great feedback on this one, and of course Bean's is legendary, so it definitely works consistently as shown. But if it can be improved then I'll revise it.

[Rob Strand]

What's the potential downside to putting them in parallel - is there more of a potential for unequal driving of the filter if the LEDs are slightly different in specs? I've gotten a ton of great feedback on this one, and of course Bean's is legendary, so it definitely works consistently as shown. But if it can be improved then I'll revise it.

You would get gross mistracking of the two integrators.  It could change the sound.  People tend to get away with it because they buy their optos at the same time so the optos are from the same batch and the LED characteristics are close.  If you fried one and had to replace it then some more of the evil would show up.   Also, some smaller vendors might not have stock from the same batch so there's a chance you could be unlucky and get different batches from the start.

[duck_arse]

I thought this would be an easy and valid fix, shifting/adding one more 330R to proper current share. but I'm not sure about that feedback connection on R22A - the original 330R stopped the feedback going as far as 0V. Rob?

[Rob Strand]

I thought this would be an easy and valid fix, shifting/adding one more 330R to proper current share. but I'm not sure about that feedback connection on R22A - the original 330R stopped the feedback going as far as 0V. Rob?

Should help quite a bit.  R22A and R22B should be 2x330=660 ohm, so 680 ohm standard E12 value.  In the original both LED currents pass current through the 330R and are sensed.  Now R22A is only sensed and it current from one LED passes through it.   As a bonus 680R helps tracking even more.

It's a shame the PCB has 330R to ground, if the Sweep pot went to ground it would be very easy to graft your mod onto the existing PCB.

[duck_arse]

so more like this.

[Rob Strand]

so more like this.

That's it.

[aion]

I thought this would be an easy and valid fix, shifting/adding one more 330R to proper current share. but I'm not sure about that feedback connection on R22A - the original 330R stopped the feedback going as far as 0V. Rob?

Should help quite a bit.  R22A and R22B should be 2x330=660 ohm, so 680 ohm standard E12 value.  In the original both LED currents pass current through the 330R and are sensed.  Now R22A is only sensed and it current from one LED passes through it.   As a bonus 680R helps tracking even more.

Can you explain more about the advantages of this method vs. putting the LEDs in series? Here's my wild swing at what's happening -

• Straight parallel: senses both LEDs, but poor tracking if the LEDs are mismatched
• Series: senses both LEDs more accurately, and reacts accordingly with better tracking than straight parallel
• Modified parallel (Duck's diagram): uses only one LED as the reference point and applies the same current to both

Taking it as a given that all three methods would perform well if the LEDs were perfectly matched, what's the difference between series and modified-parallel if they are not matched?

I'm going to do a PCB revision, but it'll be tricky to fit in an extra resistor so I'm wondering if it's just a nice-to-have, or if it's a better-in-all-cases sort of thing and it should be the way everyone builds the Mu-tron circuit when using two optos.

[Rob Strand]

Can you explain more about the advantages of this method vs. putting the LEDs in series? Here's my wild swing at what's happening -

• Straight parallel: senses both LEDs, but poor tracking if the LEDs are mismatched
• Series: senses both LEDs more accurately, and reacts accordingly with better tracking than straight parallel
• Modified parallel (Duck's diagram): uses only one LED as the reference point and applies the same current to both

[/list]

Taking it as a given that all three methods would perform well if the LEDs were perfectly matched, what's the difference between series and modified-parallel if they are not matched?

I'm going to do a PCB revision, but it'll be tricky to fit in an extra resistor so I'm wondering if it's just a nice-to-have, or if it's a better-in-all-cases sort of thing and it should be the way everyone builds the Mu-tron circuit when using two optos.

My heads a bit wonky at the moment but I can at least answer these.

As far as tracking goes the series method wins hands down.   Doesn't care about mismatched LEDs at all.  The biggest advantage is both LEDs are carrying current so one LED cannot get cut off at low currents and cause gross mistracking.   It's a very easy change but ...
- For series you would use a single 680 ohm resistor not a 330 ohm, for similar reasons to Duck's mod.
  Other than that the LED driver is running in current mode so the change to series is largely transparent.
- Where things need to be checked is for reverse mode.  With the LEDs in series the opamp needs to swing a little more and you want the same LED current as the old ckt with no signal and reverse sweep.   While you can tweak the sweep trimpot that will change the forward sweep.  IIRC, as is the opamp actually runs pretty close to saturation so you may have to juggle some resistors to get it to be exactly the same.  Having said that the change to lamps to LEDs has the same problem!   (I could have a look at the details but I'm likely to screw it up if I do it over next fews days.)
- With the series mod the opamp is only supplying 1/2 to 2/3 the current so it might swing a little higher than the parallel case perhaps counteracting that issue a tiny tiny bit.
- In your original schematic the indicator LED has been tacked onto the original circuit and doesn't really depend on the single LED drops in the feedback loop matching the indicator LED.  I think you could leave the R23 and R24 as is without use impact on the users.  At worse maybe increase R23 to 1k2.

Duck's mod gives good tracking over most of the range.   You can still get a bit of mistracking at low currents - because the voltage drops across the 680 ohm resistors are not longer large compared to the LED voltage drop mismatch. However, because the optos have the 220k's across them it hides any bad behaviour from extreme mistracking.   I haven't checked the fine details because it's already a major improvement over the parallel case.   Just how much improvement series LEDs are over Duck's mods would need a bit more effort to check.  In reverse mode Duck's mod will work like your current circuit without mods but we could argue that the change from lamps to LEDs might have already messed with that a little already!

I got a feeling Duck's mod would slot in and behave like what you have now, or better, without having to reverify anything.   The series mode is likely to be better but there's things to reverify and *possibly* tweak.   So I guess that's your call.