A successful Superfuzz experiment

[antonis]

In my current understanding impedance imbalance mean zout <> zin
zout Q3 = 10K, so I try to make it same by (12*56)/(12+56)=9.88K

Now I've got you..

There is no need to make them equal to Q3 respective output impedances (unless you care about maximum power transfer..).
Just lower their value, togheter with Q2 Collector & Emitter resistor value lowering, for the rest of impedances to be scaled up (as seen from respective drivers view point) so their inequality to be scaled down..

I might confused you, so just think of A // B //C1 and A // B // C2 for A = B = 10k, C1 = 100k, C2 = 101k and then for A = B = 1k with C1 & C2 retained the same..

[zbt]

" ... or: (J)FETs"

I mean darlington has high hfe

for Ben

all my transistors are BC547, 8 or 9 of various makes/grades/ages. the first 3 are 190, 130, 118 hFE, the rectifiers are 472/474, output is 283. I trimmed the bias on the phase-splitter for best headrooms (the 9k1's are what I have). the octave lift shown is what I'm going with. the extra capacitor is meant to AC ground the unfed base without shifting any DC's or poppings. it will work in either leg of Q3. if you don't want that nasty distortion, lift the collector.



in the rectifier, I found it easier to see what it wants to do if you lift the diodes during the breadboard stage. that's why I've fiddled the base resistors, the collectors are now at 4V33 and provide the largest clean swing, small signal, which sometimes peeps-out from under the diodes. I think I observed higher hFE gaving greater rectifier output, but it seems the octave goodness varies with frequency. changing transistors changed the balance and the freq peak. most of this is obscured when the diodes are 'in'.

lifting the octave may increase the noise as well, in which case you can tweak on the balance trimmer, and it may disappear. I've got 1% bias resistors and can tune with a 500R trimmer, but 2k is on the bb now, still keeps the tune 'close' even when 'way off'.

diode lift is just a 1k resistor. the volume increase makes stock sound puny, so it can stay as is. stock volume is puny, by the way, even after messing the last stage for more gain and using a log balance pot.

I dunno if I've come to the same outcome as digi, but I have found some things. there's mayhems in all directions; saturations, disappearings, louder re-appearings in the tail. lotsa times it screams 'feedback', but doesn't actually. someone who can actually play a note might even get some toans from, but not me. I haven't got a fork, so I'm sticking my toothpick in it, no more changes.

You could always use a precision fullwave rectifier with the same effect as the push-pull amplifier used in the circuit. 

Thanks for offering me another crispy circuit, Sir Ron, I reserve for later.

There is no need to make them equal to Q3 respective output impedances (unless you care about maximum power transfer..).

a! ok now

I would clarify again my mind

"When we match transistors it can mean different things because they have more than one parameter.  "Matched" transistors often matches Vbe for a given current (as well as gain to a degree).   As it turns out Vbe only has a small effect for the Super Fuzz.  What is very important is to match the hFE."

https://dragonflyalley.com/synth/images/TransistorMatching/ianFritz-transmat0011_144.pdf
it's a relief, only hfe

"From the perspective of the phase-splitter we want the signals at the collector and emitter to be equal magnitudes."

-1V 0 1V,     ? Volt value 

"To simplify the argument we will ignore the effect of the transistor's output impedance at the emitter (re) and the fact the collector current doesn't quite equal the emitter current."

IE = IB + IC,  IB so small

"Under these assumptions:  The gain for signal at the emitter of the phase-splitter is 1 So for matching we need the signal at the collector to have gain 1."

A = 1 = RC / RE

"And for that we need the load on the collector to equal the load at the emitter."

RC = RE

"Now, the load on the emitter is:   the 10k emitter resistor, the 22k bias resistor, the 100k bias resistor, and (most importantly) the input impedance of the rectifier transistor all in parallel."

RE = 10k // [470R + (22k // 100k)] 

"The load on the collector is the 10k emitter resistor, the 22k bias resistor, the 100k bias resistor and the input impedance of the *other* rectifier transistor all in parallel.   We could throw in the effect of the 470R's we well."

RC = 10k // [470R + (22k // 100k)] 

"The key point here is the input impedance of the rectifier transistors need to match."   

maybe at this point I'm confuse, imbalance between rectifier transistors itself

"That impedance is actually quite low compared to all the other resistances making up the load impedances.     
The input impedance of a transistor stage is directly proportional to hFE (actually hfe+1)."

zin = (hFE + 1) * re

"So if the rectifier transistor gains don't match the input impedances don't match and that stuffs up the balancing of the signals on the collector and emitter."

diff
hfe  ohm
10   450
20   900
30   1350

different value of 30 already make 1K differences, what makes a bulk differences? 10?

"With this in mind the, mod of adding the resistor in series with the emitter output of the phase splitter obviously doesn't help matching the collector and emitter *voltages*; which is exactly what I saw."

ouch, I thought the 470 was to avoid distortion, and that value for 10 differences, now become accessories, why not just replace it with diode
might be Kay Fuzz dont use it.

"Another issue with unmatched hFE on the rectifier transistors is it makes each rectifier transistor operate at a different current and that makes gain for each arm of the *rectifier* unmatched.   This is much less of an influence than the input impedance issue.  Typical the side with the high gain transistor will have a higher rectifier gain and higher voltage gain so it make both mismatching mechanisms on that arm worse."

find low match hfe

find my luck at digikey 2SC4207-Y, but not at my local store 
not know would be match, may be others can test it.

this also interesting
"- Possibly use separate emitter networks for each rectifier to stop hogging of the bias current."

[Eb7+9]

These were a popular sale for me in the 90's ... you only need to match hFE and use 1% resistors in the rectifier section to get consistent results // matching signal caps doesn't hurt either...

http://www.lynx.net/~jc/pedalsRoyal.html

[puretube]

I mean darlington has high hfe

I mean (J)FET has high input-impedance.

[zbt]

That would be interesting too, since the JFet only requires a 1M resistor bias
does it have to be matched too?

once again I would clarify my reference interpretation

Base on Solidhex

    C        B        E                 hfe
Q1  6.01     0.66     0.117     C539    100 
Q2  8.88     6.01     5.41      C828    200
Q3  6.23     3.27     2.72      C828    200
Q4  3.1      1.7      1.11      C539    100
Q5  3.1      1.67     1.11      C539    100
Q6  5.75     0.98     0.361     C539    100


suppose it use C539 TO18 hfe MIN 250 for low noise version
and C828 TO92 hfe MIN 65 as plain vanilla
I try to search and I could not find with C539, it is easy to spot because,
it would be different package, I found only all with same C828-Q
so by batch it would be all near low or high hfe

Using VCC set to 8.88V I try to check for another transistor with 100, 150, 250, hFE
and all voltage are close to Solidhex ref for hFE all around 100

So I replace Q2 and Q3 to test if it using different transistor with 250 hFE
interesting enough Q2 dont make different voltage, but Q3 go near 3V and 6V
so I am sure enough it all using same transistor.

But for Q4/Q5 I am curious why VC can get so low, cause lower hfe tend to be higher VC
would it be using high hfe, just for this two, my measurement for VE are bit lower and VC are bit higher
then I check I am using 22K transistor, so it probably it because 10K trimpot
I change 22K to 27K, and it come closer,
Test using 250 hFE would be below 3V, so Q4 and Q5 also around 100 hFE

**The good reference using low hFE and using trimpot to set it right**

now it even lower, how about 200 is easy to find

These were a popular sale for me in the 90's ... you only need to match hFE and use 1% resistors in the rectifier section to get consistent results // matching signal caps doesn't hurt either...

http://www.lynx.net/~jc/pedalsRoyal.html

agree

mentioning fuzz face, kind dejavu it took my pride away
meh, is just easy circuit, took BC550 on it, and the result ...

Q1 and Q2 look like that, I try set Q1 lower as already test Q2 can be 200
also Q3 and Q6 200, for Q6 headroom (4.5) I just change RB 15K to 22K and RE 1K to 2K2

**Q4/Q5 should be match**

but still curios for Q4/Q5 

[Ben N]

I could not say about match pair transistor

but on solution
"- Make the rectifier darlingtons, or"

Why? Does overall hfe matter? The originals were under 200. IIRC.

[Mark Hammer]

I was anxiously awaiting delivery of a transistor tester I had ordered, so I could evaluate transistors for a variety of purposes - stompbox and aynth.  It arrived yesterday, and I was disappointed to find that it was the plexiglass enclosure for the tester, rather than the tester I *thought* I was ordering.  Make sure you look and click carefully before hitting "Buy now".

[zbt]

Why?
I just dont have 2SC1583, LM3086, etc

darlington has higher hfe so it probably work also, but I dont know about AS394H  the internal also not so simple

other result

"I think I observed higher hFE gaving greater rectifier output, but it seems the octave goodness varies with frequency. changing transistors changed the balance and the freq peak. most of this is obscured when the diodes are 'in'." duck_arse

"Typical the side with the high gain transistor will have a higher rectifier gain and higher voltage gain so it make both mismatching mechanisms on that arm worse." Rob Strand

Does overall hfe matter? The originals were under 200. IIRC.

depending on who's looking

If design for perfect could be yes

If I ask uncle Matthews, why Sir is just one transistor different?
This transistor is called John Whatever (with a cigar)

If I have to put C539 and C828 may be two batch, have to measure each,
would not be so costly? Make simple buy bulk 

For under 200 could be ok

Then if I throw all for >200hFE, could be a problem for some topology like Q1 & Q3
is like fuzz face which in my experience Q1 better be low,
for others Q3 dan Q6 probably not


Stay safe Sir Mark,

May we ask you how much hfe 2SC1583 of yours, is it lower than 200?

Thanks in advance

[Eb7+9]

Why? Does overall hfe matter? The originals were under 200. IIRC.

absolute hFE levels don't matter in the rectifier section ... as long as they're not stupidly low

[zbt]

"(In case of a tie, the professor wins.)" Hayt Neudeck 

[antonis]

I could not say about match pair transistorbut on solution
"- Make the rectifier darlingtons, or"

Why? Does overall hfe matter?

It surely DOES matter for Q4/Q5 stages impedances inequality settlement..
(the higher the impedances values seen from Q3 outputs the lower their practical difference..)

[zbt]

hfe     Rpi     zin //18K (100K//22K)
100     4500    3600 
200     9000    6000  change RE/RC Q3 to 5K6
300    13500    7714
400    18000    9000
500    22500   10000  close to RE/RC Q3
600    27000   10800
700    31500   11455
800    36000   12000
900    40500   12462
1000    45000   12857


(unless you care about maximum power transfer..).
use 500 hFE or change RE/RC Q3 as close zin
the higher hfe zin more depend on RA//RB (100K//22K)

for Q3 phase splitter
IE=IB+IC less IB more hFE, VC === VE should be improve octave, but how much compare to 10K 5% vs 1%
VE>VC around 6mV wouldn't be 470 ohm at collector side better be shorted

for VC=7V and VE=2V only add 0.75V headroom, the amount is controlled by expander,
so guess is ok, mod is simple change 150K to 100K

of course JFET is better, but again is expensive, hard to find today in my case

the problem with vintage, I feel tied to the past.
if I could go back in time, and build it with an hFE 200, the question whether it would be preferable too

looking https://wiki.analog.com/university/courses/alm1k/alm-lab-phase-split
change 100K//22K, make a single reference using 10K as 4.5ref and use diode
dunno this would work...

[zbt]

BOSS FZ-2 also use 100K, and weird 4K7 with 10K and 1K and no value of RE 
but of course RE=RC so RE=4K7 



world full of imbalance, using all 5% for mojo, in the end trimmer is enough, trim at emitter to reduce noise 

[Eb7+9]

world full of imbalance, using all 5% for mojo, in the end trimmer is enough, trim at emitter to reduce noise 

trimmer is for "bettering" during (mass) manufacturing using random devices ... likely done somewhat crudely by ear before boxing
bettering is not the same as best ... trimmer is there to compensate somewhat for device mismatch, nothing to do with noise ...

for doing DIY onesies and twosies use a DMM and get quite likely better performance than BOSS did
no need for ultra-matched packages in single-ended rectifier stages

it's about dealing with statistical factors, blown away by the mighty DMM

[zbt]

maybe because I tried it on a breadboard, yes the trimpot might change more.
so i tried this circuit, sing like celine dion and a little raspy like mariah carey 
using only one reference, so there are only four resistors that are match.

[zbt]

Finally ... low hfe is more pleasing for me
here is another board



Enjoy!

[amptramp]

When considering balance, the current through the emitter in the phase splitting transistor is higher by the amount of the base current than the current through the collector.  The emitter output is also lower impedance, so even with matched resistors, you have an output that is imbalanced by the ratio of 1/hfe.  A FET phase splitter would not have this problem since no gate current means the source and drain currents are equal.  A Darlington transistor would reduce this imbalance while retaining the predictability of biasing voltage, which can be all over the place with a FET.

The effect of imbalance is to allow some of the fundamental frequency to get through.  How much of the fundamental can we tolerate?  Do we really need to eliminate the fundamental?  There might be a fuller sound with some of the fundamental still in there.