Project Big Marf

[pinkjimiphoton]

i think its just clipping in a different way. more compression with germanium, so if you raise the signal with it the noise is less audible.. ?

i still like the phase shift idea.. yeah... that's the ticket ....   

[pinkjimiphoton]

Thanks for those videos Ben.  Cool Muff variant.

Like anotherjim and others I've been puzzling over the noise clamping properties of the germanium diode.  My theory is that this effect is due to the leakage properties of the germanium diode when reversed biased.  The leakage clamps low amplitude noise while having a minimal effect on the large amplitude guitar signal.  This probably introduces crossover distortion too, but it's inconsequential given the massive clipping occurring in the gain stage.

I'd try using two germanium diodes in series.  It will provide for the same clipping threshhold as one silicon diode, but may also do a good job clamping the the white noise.

when i first came up with the mod, i tried two ge's in series, and the noise gate effect pretty much stopped, the hiss went back up noticeably.

since it only works on that one diode in the one orientation and position of the circuit, i am thinking more and more it is clipping the treble frequencies enough where the signal is overpowering the hiss. i hope someone can figure this out, cuz it definitely works (and i bet ya can recover the gain by using a higher gain transistor in one of the next stages.) and it would be cool to understand it.

yes, there is a tradeoff with the sustain, but ya can crank the volume a bit more and likely never notice it's missing.

[Ben Lyman]

Rob- I tried some bypass caps off of Q4e, definitely does something, big boost, big distortion. Kinda brings back the noise that the Ge diode is removing.

Jim- I tried some resistors from 1M up to 3M3, with and without any and/or all 3 diodes. Not any change to speak of.

Bool- thanks, that's a great idea. I tried 100n, then 47n, and now 10n seems to be the best at removing the flub. 10n seems pretty small, I wonder if the Marshall tone stack actually needs some adjustment so it can work better with different kinds of guitars.

Frank- thanks, I'll try that too. I did try two Ge back to back without the Si and it clamped the noise but the distortion flavor was a little different. A single Ge also clamps noise with no Si at all but of course without the two back to back 914 it doesn't sound anything like a BMP

Larry- I tried a few resistors with the diodes in like Jim suggested, then tried again without the diodes as you suggested. I didn't notice anything

[pinkjimiphoton]

resistors don't add compression to the signal, but diodes do. everything has to build up to the conduction point, and that clipping aspect can be exploited i think to literally comp the signal enought to clip the noise.

nice work trying all the ideas out ben
looking forward to where ya end up at. that last clip was sounding great

[Ben Lyman]

Jimi- I'm not sure it's ready yet. It does sound good with my strat but not with my HB guitar unless I switch to a 10nF input cap.
Maybe I need an input cap toggle switch more than a presence knob

[pinkjimiphoton]

a good compromise on the input cap would be around 47n. then its still sweeter, will work well with humbuckers and strat both. try 22n, 39n as well. but i'm betting the bigger cap will give it a little more balls for both the strat and the buckers. no matter what ya do, everything's a compromise. then it's a balancing act... ok, lost some signal here, but i can make up for it there kinda thing.
to me, 10n on a big muff is just way too thin, like a stock jcm800. i like it a bit warmer. but with the eq you have worked up, you should be able to compensate for different guitars.

maybe (thought here may sound contrary) but put a high pass filter at the beginning of the circuit instead of the input cap or feedback. then you can literally dial in and out the distortion level and mudd of the pedal, and make it overdrivey to the usual sludge.

just an idea

but i have found by adding a bass cut on my guitars, it gives me a huge variation in available tone, and the ability to dial in the distortion level from my guitar... more bass= more distortion. then you could sweep what sounds best for any guiitar with a simple knob. you're up to 5 at that point, but... you may find it useful.

[Ben Lyman]

Okay, I'll try 47n input to start with since it seems easiest.
My latest experiment was to replace C13 (22n) with a 10n and I think it might have helped but I'm not sure, I need to give my brain a rest and come back to it later.

[Quackzed]

on that diode noise thing, i'd imagine that if its causing gating then that gating action is probably the reason of the noise reduction. the noise, being samllish random squigles below the gating threshold get blocked till the relatively large 'actual' guitar signal comes along and easily makes it over the gating diode threshold... it'd be a super fast noise gate...

[pinkjimiphoton]

thanks for the explanation gil, that makes sense. cool beans! 

[Frank_NH]

Well, I have a big muff circuit model for LTSpice (and an LTSpice model for the 1N34A diode), so I ran this mod through Spice's noise analysis.  Using the germanium diode did lower the calculated noise level by 50%.  The orientation didn't seem to matter, and using two in series had a higher noise level than just one, but not as much as a silicon 1N4148.

So I still believe that the germanium diode leakage is responsible for the noise gating properties, but I suppose I'll have to breadboard some version of this circuit and check it out for myself.  I have a bunch of different germanium diodes I can test to see if they respond in a similar way.

[pinkjimiphoton]

awesome. when ya breadboard it tho, i think you'll find the noise reduction effect is lost if you use the reversed orientation.

i'm imagining that lt spice is looking at noise overall? maybe it doesn't account for the frequency? just a guess...

in a diode clipper, one half clips the top, one the bottom of the waveform. i'm thinking that in the shown orientation it's only working on
the higher frequencies, resulting in an audible reduction...

but if it were on the lower frequencies (reversed) you likely wouldn't hear the difference in noise. we hear it mostly in the upper mid and treble range.

so if it's reducing it 50%, but only on one half of the wave, i'm betting it's on the highs and not the lows.

does that make sense? sorry, kinda limited descriptions.

i think the leakage has a lot to do with it too. if the ge conducts at .3v or so, and the si around 6,
one half of the waveform being clipped is gonna do it 50% sooner than the other.. and put the signals out of phase when the higher conduction diodes open their gates.

i'm really wondering if by doing so, its using feedback out of phase with the high end of the effect to noise cancel. the phasing effect may even account for the slight drop in balls... again, most noticeable in the high end.

i'd love to see this on an oscilloscope.

makin' popcorn... 

[Ben Lyman]

I've reversed the diode orientation several times and the noise reduction is just as effective either way I put it 

[pinkjimiphoton]

then i'm wrong again like usual. i seem to remember it only working in one of the clippers, and only in one orientation, but man,
fuzzy... that was years ago

[Ben Lyman]

right about the one location, only at Q2.
But ya, it doesn't seem to matter which direction

[Frank_NH]

Jimi - thanks for your comments.  Ben - I recall your second video showed the two orientations and it seemed to be effective either way. 

Some things I've been pondering: 

* Assuming no guitar signal, the noise level, even when amplified by the first stage, is going to be too small for the silicon diodes to care.  They will never reach the 0.7V clipping threshhold.  So you can even remove the silicon diodes entirely and that won't affect the noise reduction (again, for NO signal - just noise).
* The output from the second transistor gain stage is inverted.  The noise signal will also be inverted.  Does this affect anything since it's being fed back to the input?

[pinkjimiphoton]

frank, i know for a fact if ya take a noisy track in a DAW and sample just the noise, then invert the phase of it and mix it with the original signal, bam, noise gone.

in my cybertwin, they fake a champ by inverting one of the 12's. it will definitely work.

i wanna see this thru an o-scope. i think that would be the definitive answer to why.

[anotherjim]

Well, invert and add back is what negative feedback is.

BTW, Ben, is "Marf" a play on your British (Marshall) angle and they way those Brits down in the south say "Mouth"? Out here on the perimeter, we say "Gob".

Anyway, had a mess with my built BMP. Ge diode (0A97, big glass point contact thing) most certainly do cut noise.
Tried resistor instead. 1M slight noise reduction. 470k good noise reduction. Well, duh, that's doubled the NFB so of course it did.
What else? Capacitance maybe. Needed 470p at least for a good noise cut (slight with 220p). Of course, a cap is "some resistor" depending on frequency and as it's between base and collector, will be a low pass filter. BTW, capacitance of the diode is tiny around 1pF, so it ain't that.
That's as far as I got for now.

Looked up 1n34 specs. Leakage at 10v suggest equivalent resistor about 330k. I don't know what it would be worth with the millivolts of noise acting on it, but I think reverse leakage increases fairly linear starting from 0v up to a point. So the equivalent resistor is possibly still in the about 500k park.

[Gus]

Is the collector voltage of the stage Ge different. Can someone measure the collector voltage with the GE then change it for a Si and measure it?

[pinkjimiphoton]

that is awesome jim

gus.. good to see ya chime in. any thoughts on why it works?

[Ben Lyman]

Jim, I was just thinking Marshall+Muff=Marf

Gus, I'm not sure but that might be what I did in the last video. Unless you mean measure collector with nothing but the Ge?