Noiseless MOSFET boosters: ZAZEN and QIGONG

[Fancy Lime]

Hi everyone,

every so often questions pop up on this forum that boil down to: "My SHO (clone) is noisy/hissy, what can I do?" My instinctive reply would normally be: "Build a better booster" but that is not particularly helpful unless I can say how to do that, now is it? So I finally got around to scribbling down the schematics for the stage that I like to use as an input booster for overdrive pedals. It is essentially a negative ground P-channel version of the SHO with some mods and improvements to improve gain control characteristic and get rid of the annoying crackle. The main advantage of using a PMOS instead of NMOS stage is the much lower inherent noise.* This thing is essentially dead silent to my ears even at the maximum boost factor of 15, unlike any BS170/2N7000 booster stage I have ever tried. I call it the ZAZEN BOOST:



As it is, the frequency response is flat. Film capacitor C4 makes sure that the increasing impedance of the electrolytic C3 toward higher frequencies does not cut highs, while the large C3 ensures not to cut lows. If you want a treble booster, you can replace these two with a single 100nF to 1µF film cap. You can also add a Bright switch that connects a small cap (4.7-47nF) across the Gain pot.

If you need more boost and don't mind getting your tone dirty, try the ZAZEN's two-stage brother, the QIGONG DIRTY BOOST:



This is more of a proof of concept to demonstrate how to a multi-stage implementation of this thing could look. As it is it provides very loud boost with some mild overdrive in the upper quarter of the gain range. Lots to be optimized to personal taste. Double the second stage and add a RAT-style tone control and output buffer for a nice "tube"-sound (in marketing speak terms) overdrive. A Bright cap across the gain pot may again be a nice plus. Or increase the supply voltage to 18V for more headroom (untested but I see no reason why that should not work) and make it a really loud clean booster.

The transistors I use are marked "BS250" and come in an E-Line package, which is TO-92 compatible but flatter. I bought them without a brand name but according to the interwebs the only manufacturer using this package for BS250's seems to be Zetex and the full part name is "BS250P". The ones I use look exactly like the uppermost image here:
https://uk.rs-online.com/web/p/mosfet-transistors/2156688/

Happy soldering,
Andy


EDIT:
* Clarification: This is not a general law, I should have been more careful stating this. It would be more precise to state that my limited number of tests with a limited number of devices suggest that the BS250 suffers significantly less from noise than the BS170 and 2N7000. That is why I included, which BS250 I was using. I have not come across any materials that allow me to deduce if this is universally the case for all BS250's and BS170/2N7000's, let alone all PMOS and NMOS transistors. Some people on internet forums seem to think it is, others disagree, but that doesn't help much, does it?

[Plexi]

Niiice!
I'll see about that BS250's

[Gus]

A schematic from the past. Does anyone remember when the Obsidian was first posted?

http://www.diystompboxes.com/analogalchemy/sch/obsidian.html

[marcelomd]

Hmmm,

I'm playing with a modified Box of Rock and going nuts with the noise.

I'll try this PMOS approach if I can find the damn parts around here.

Thanks!

[Fancy Lime]

A schematic from the past. Does anyone remember when the Obsidian was first posted?

http://www.diystompboxes.com/analogalchemy/sch/obsidian.html

Not exactly the same schematic, I would say But you are of course correct if you assume that the Obsidian is where I first got the idea to try and see if PMOS stages are indeed quieter than NMOS stages, as Joe claims. No surprise there, they are.

BTW, a while back I converted the Obsidian to negative ground and added a few things to further improve the noise level as well as an output buffer (optional). I thought I had published that here somewhere but can't find it. maybe I forgot. Well, here it is with output buffer:



And here without:



If I were to turn these into a small production series, I would change the biasing method to feedback bias like I did for the boosters. Seems to work "out of the box" with no adjustments for all BS250's I tried so far, whereas the voltage divider bias is a bit fussy with the transistor characteristics. But that may or may not be different for different batches of BS250's.

Cheers,
Andy

[Fancy Lime]

Hmmm,

I'm playing with a modified Box of Rock and going nuts with the noise.

I'll try this PMOS approach if I can find the damn parts around here.

Thanks!

A PMOS BoR sounds interesting. Would that be a Pox of Rock then? I built a BS170 BoR few years back and was also very dismayed with the racket. Keep us posted!

Andy

[rankot]

It's very interesting, have you ever tried it with IRF/D series PMOS HEXFET?

[PRR]

> advantage of using a PMOS instead of NMOS stage is the much lower inherent noise.

Why?

MOSFETs are notorious for 1/f noise. I do not know why P-type would be better.

In"serious" low-hiss work, BJTs dominate below a few K source and JFETs above that. Either can give hiss far below a guitar's self-hiss (internal resistance).

[R.G.]

You might want to think about:
- extrapolating any particular characteristic from a limited sample size, especially something like "noise-LESS"
- historical issues of stability for high gain devices in the reversed-ground connection.

Just sayin' 

[PRR]

He may have a point about MOSFET P-type being less hiss than N-type. I am trying to digest this:
https://www.silvaco.com/content/kbase/noise_modeling.pdf

However it is very clear that the 1/f noise corner for the MOSFETs tested is far above the audio band, leading to high audio and sub-audio hiss (rumble). Above the 1/f corner we can estimate hiss from conductance, which is similar in small MOSFETs or JFETs (and better in BJT). But 1/f noise corner in JFET or BJT is typically below 1KHz, often near 100Hz, so MUCH less noise in the audio band.

MOSFETs are troubled because the active area is either "surface" (full of defects) or sub-surface ion-implanted (ions damage the crystal). Carriers fall into "traps" and come out again at random instants. Being crystal-level instead of atom-level, the relaxation time is slow, leading to rising sub-sonic (even low sonic) random noise.

"A variety of measurements of the excess noise of metaloxide-semiconductor field-effect transistors (MOSFETs) suggests that their noise is associated with capture and emission of charge carriers by localized traps at, or near, the Si/SiO2 interface. In very small devices, discrete fluctuations associated with the capture and/or emission of single charge carriers may be observed, especially at cryogenic temperatures. Superposition of the noise of many fluctuators leads to the ubiquitous inverse frequency dependence commonly observed in larger area devices. A 1/f noise spectrum has been shown to result from superposition of thermally activated processes with a uniform distribution of activation energies or from tunneling processes with a uniform distribution of tunneling distances. Uniform distributions are not necessary, however, as 1/f-like noise can result from a log-normal distribution of switching times."
http://www2.oberlin.edu/physics/Scofield/pdf_files/ted-94.pdf

[PRR]

Comparable N and P types. At 1KHz, the SiON oxide types both give just over 10^-22 A^2/Hz output current hiss. P type a hair lower, but way within the overall randomness of these plots and certainly real devices.

[Fancy Lime]

Hi guys,

thanks for the discussion!

@ Ranko
No, I only tried the BS250 vs the ubiquitous BS170 and 2N7000. Mostly because these are so common and seemed a good place to start.

@ R.G.

- extrapolating any particular characteristic from a limited sample size, especially something like "noise-LESS"

True. I clarified that in the original post a bit (see EDIT). The whole "PMOS is less noisy than NMOS" claim is based on very limited sample size (aka anecdotal evidence) and "what I read on the internets". I therefore declare it the whole truth and nothing but the truth and anyone who dares question me is part of the conspiracy that is out to get me or brainwashed by Them. This is how things work nowadays, right? Kidding aside, I really should be more careful with my claims, I'll explain in more detail below. Also "noiseless" should have been "less noise" and was not supposed to mean "noise free" but emphasis gets lost in typing I'm afraid.

- historical issues of stability for high gain devices in the reversed-ground connection.

Also true. Since our last discussion on this topic I have run into a few instances where I could not get negative ground P-channel/PNP stages to work for no obvious reason. Dark magic, probably. It really bugs me that I don't understand why this sometimes works and sometimes not. Anyway, I'll keep trying and see where it leads me. Sometimes patterns emerge only from large datasets.

@ Paul
Well, my initial reaction to first reading this claim in Joe Davissons Obsidian description was similar to yours: Why the hell would that be the case? And I did not really find a satisfying answer. One thing I have found over and over again in forum and PowerPoint presentations on University sites is that Surface Channel PMOS (typical for complex IC's with small transistors) devices seem to have about the same noise figures as NMOS, whereas Buried Channel PMOS (typical for "large" devices like TO-92 transistors) seems less noise-prone. This is supposedly related to, as you quoted:

A variety of measurements of the excess noise of metaloxide-semiconductor field-effect transistors (MOSFETs) suggests that their noise is associated with capture and emission of charge carriers by localized traps at, or near, the Si/SiO2 interface.

Because in a Buried Channel device, carriers cannot jump between channel and oxide. Therefore Flicker-Noise is supposedly suppressed in Buried Channel PMOS devices, leaving only random carrier motion and bringing the noise down to "typical BJT-levels", whatever that means. Other people claimed that the lower noise of PMOS was due to the lower carrier mobility of holes compared to electron in an NMOS but that seemed even more speculative to me. Not like any of this was backed up by actual data that were linked, so who knows. I sincerely doubt that any of it is completely correct and from everything that I found in current academic research on noise models (which isn't much) it seems like we don't really understand all factors that go into noise all that well but no-one seems to bother too much since we have semi-empirical noise models that fit the actual characteristics well enough so we don't need to bother with all minor factors of the underlying causes.

Long story short: lots of speculation, few well-sourced facts accessible. So I decided to just check if the PMOS transisors I had kicking about were less noisy than the NMOS that I had. Turns out they were. So my revised advice to the DIY MOSFET aficionado would be: If the noise of your BS170 bugs you, you might want to try if you can use an equivalent PMOS design using the BS250 instead and see if it gets better. It may but YMMV, as always. To me it seems well worth a try for something as simple as a single-stage booster. Some people have suggested that getting the "good" BS170's from Zetex also solves the problem, if you can get them and are willing to pay the mojo surcharge. I prefer brewing my own mojo, so I use BS250's instead.

As for going BJT/JFET instead of MOSFET: For clean sounds, sure. For high-gain overdrive, no problem. There I can hardly make out any difference between JFET and MOSFET overdrive, if tuned accordingly. If anything, the JFET sounds better. However, there is this window of gain at the onset of overdrive, where MOSFETS sound best to my ears because they transition more smoothly into very dynamic clipping, which I have never been able to replicate with JFET's or BJT's. JFET SPRR stages are the closes thing I know, but no cigar. So for guitar purposes, the small noise penalty of a MOSFET stage is worth the slightly better sound to me. But that's more feeling than anything else. I should try and see if overdriving a BS250 with a preceding JFET SRPP stage would not make a nice overdrive...

Hope that clears a few things up,
Andy

P.s.: I really love this forum for discussions like this. How come we can have a nice civilized conversation about sound and such (which, let's be honest, has quite a bit to do with unsubstantiated beliefs) but as soon as people talk politics or religion or whatever on the internet it's all "§~[& you, stupid $#!+head" and the like?

[BetterOffShred]

I have some 250's.. I may give this a whirl thanks for sharing it with us.  I love a good dirty boost.  I just built a couple Saturn 5 clones and those are pretty awesome, Im also digging the Red Rooster booster.. That's a pretty fun one.   

[anotherjim]

So... mess with CD4007 P/N pairs? A couple can easily be used discretely.
I've made common source and source follower stages with P or N and not noticed a huge difference, but I must stress that I wasn't sniffing for noise levels - I just don't remember either being noticeably noisy.

[DDD]

It is possible that CD4007 hisses more than discrete MOSFETs due to the well-known reasons such as insulation leakage, e.t.c.

[karis12]

Nice work! I bought some BS250s just to try out the lower noise MOSFET boost thing (my version still followed Mr. Orman's original design) and indeed it was dead quiet compared to my old 2N7000 build (which was still reasonably quiet through a very clean amp mind you). I solved the problem of positive ground by using a charge pump to invert the voltage.

[swever]

Thank you for another great contribution, Andy! You are shooting out new designs faster than I can build old ones that are ready and tested.

Gotta stop doing debuggin all of my one thousand and one half-baked pcb's and building new ones and finally pull out my breadobard and get to the real deal with all the stuff you posted here.

[Fancy Lime]

Hi there,

I just happened upon Marc's very cool build of the Zazen (https://www.diystompboxes.com/smfforum/index.php?topic=36392.msg1143578#msg1143578) and thought I'd re-link the broken pictures.

https://postimg.cc/kB7snSLt

https://postimg.cc/148nD36Y

Cheers,
Andy

[bluebunny]

Yeah, cool designs!    Thanks Andy.  Got a couple more of yours in the pipeline...  Somewhere...  Eventually...        

[BetterOffShred]

I had printed these out when Andy first posted this, and never got around to building either one, and then the thread got bumped and o decided to take a crack at it.    Made a layout for the Zazen and built it , works great sounds awesome.  Used the same layout and attached a smaller layout for the second mosfet to turn it into the Qigong, with a strip between and a couple jumpers to make it neat and all on one board.   Really like both, will be boxing them both.  I'll post my layouts if anyone wants a stripboard layout.  They are hand drawn too .. 

Thanks for the schematics Andy