Never noticed before.... crybaby buffer

[ayayay!]

Okay so this is what many (sometimes myself) consider the "crappy buffer."  But really, it can't be that bad...

Anyway I noticed last night this doesn't use your typical Vref (1/2 voltage) via something like a 470k connecting to the base of the tranny.  Instead it uses full voltage, knocked down to 10 micro amps using a 1.8M resistor.  So no need for Vref.  I ran into a pedal once that only added Vref for the buffer purpose we use more commonly.  I could have saved myself some board space and a couple electros in the process.

Thought this may help in the future if you want to omit some extra parts.  Not a revelation and I'm probably the last guy to know, but in all the articles on buffers out there, I've never noticed this. 

[CynicalMan]

JFET buffers are even easier. 3 parts and high input impedance.

I don't understand why people dislike this buffer. The input impedance is high enough and it shouldn't change the tone of the circuit. It just seems to me like the "it's not how they made it in the 60s so it must be worse!" reasoning.

[ayayay!]

Yeah I much enjoy the JFET buffer I put at the end of my Dr. Boogey.  I don't know, I mean, I think maybe it's just that it's a buffer in front of the wah.  I've stopped removing it in my wahs now.  I bypass it w/ an appropriate switch nowadays, but I don't mind leaving it in.  Since I up the input cap to a .1 anyway, it seems fine and the results are more consistent when a buffer is there.

Anyway, like I said, it keeps you from having to build an extra voltage divider if that's the only need... just food for thought in future builds... kinda thinking out loud...

[Gus]

That is standard textbook biasing.  Why would one call it a crappy buffer?

There are few ways to bias transistors all have pros and cons.

http://www.fairchildsemi.com/ds/MP%2FMPSA13.pdf

Note the hfe goes down with Ic.  The circuit operates under 10ma hfe speced at two Ics 10ma and 100ma(10K and 9VDC has to be under .9ma you know that just looking at it)  Now the input resistance will be the 10K emitter R times the hfe at the Ic the circuit runs at in || with the 1.8meg.  Things to note the transistor is a darlington so it has high hfe and two BE drops.  How big is the input signal being an emitter follower the out will be just a little smaller than the input, gain has to be less than one.

Someone selected a resistor value to place the emitter operating point at a range that will work with the transistor type selected.  

It is not how I would do it but it works.  I might build it like the EF section of this
http://www.aronnelson.com/gallery/main.php/v/gus/wha1.gif.html?g2_imageViewsIndex=1

[ayayay!]

That is standard textbook biasing.  Why would one call it a crappy buffer?

Like I said, "But really, it can't be that bad..."  I'm just saying in the past I've stated I'm not wild about it, but I'm officially jumping off that bandwagon. 

That's good info Gus thank you.  Just the kind of discussion I was hoping to start.  There's still too much buffer misinformation out there. 

[zombiwoof]

People have reported that the buffer in the Dunlop Crybabys makes the pedal thin sounding.  It was added to the wah by Dunlop to reduce the tone-sucking in bypass, if you true-bypass the wah the buffer is no longer necessary for the original purpose, and many say removing it makes the wah sound better.  The original Voxes and Crybabys didn't have the buffer, but because of the bypass wiring used the highs and lows were affected when in bypass (the "tonesucking" issue), so if you put in a TB switch, and remove the buffer, it will sound more vintage and not have the main problem associated with the originals.  If you like the tone of the wah with the buffer after TB-ing it, leave it in, if you want a more vintage sound take it out.  Simple.

Al

[PRR]

> knocked down to 10 micro amps using a 1.8M resistor.

Where did you find 10uA??

> That is standard textbook biasing.

Textbooks will start with two base resistors. Any mature designer with a loose-spec problem sees that a single/normal transistor will often meet specs with a single resistor (maybe half the textbooks brush this idea).

Hand analysis is trivial. hFE is ratio of collector (or emitter, near-enuff) current to base current. A 2Meg resistor base to B+ has the effect of 2Meg/hFE resistance b+ to emitter. Say hFE is 200. 2Meg/200 is 10K. We have 10K emitter to ground. The two resistors drop equal voltages. The base-emitter area lands about halfway between B+ and ground, a good first target for good output. And this half-way is consistent as B+ varies.

hFE is NOT constant. It may be 100 or 400. Do the math. For hFE=100 the emitter lands near 1/3rd of B+; for hFE=400 the emitter lands near 2/3rd of B+. Is this acceptable? At extreme hFE we still have a third of 9V-0.6V=8.4V or nearly 3V of swing. Signal here won't be over 1V. It's fine.

> MPSA13

Ah, that's different. Page 2 of the Fairchild datasheet shows hFE near 1mA (as Gus says, it can't be over 0.9mA but probably near there) as 100K or 100,000!

So as far as bias relationships go, the 2Meg acts like 2Meg/100K= or 20 ohms. There is "no" voltage drop across it.  Taking 9.0V supply and 1.1V Vbe we have 7.9V across two resistors, a 10K and a 2Meg levered-down to 20 ohms. We get 7.88V across the 10K, the emitter sits within 1.11V of B+, the base sits 0.02V below B+.

And base current is really 0.02V/1M8 or 0.011uA.

From this point, we may swing -down- most of that 7.88V, perhaps 6V. Since we don't need over 1V swing, fine.

Darlingtons are tricky. We can pull the base up above B+ (thanks to cap-coupling). But at base 0.6V above the collector node (B+) the first transistor quits transistoring. So something happens for 0.62V positive swing. In fact the first transistor now works as a diode, "Darlington" action fails. The second transistor still works with hFE perhaps near 100. We can go another 0.6V up, with reduced input impedance, but still higher than the guitar impedance. So in fact the circuit will pass nearly 1.1V positive peaks, ample for guitar.

It does smell like Q1 could maybe be a simple transistor, not a Darlington. In fact I wonder if someone mis-read 13 for 18. It might make more sense to use the Darlington in the second stage, the filter. And since this is a high-production pedal and 13 costs a penny more than 18, you'd think that Dunlop's designer thought it out right. The 13/18 mixup may be in production or someone's re-drawing.

> makes the pedal thin sounding.

The 0.01u after the 68K is liable to shave a lot of bass. I do not know if that value is needed to make the filter resonance happen, desirable so heavy bass does not overload the high wah sounds, or leftover from some non-buffered plan. Faster to try a bigger value than to think about it.

[Gus]

TAOE(textbook people at this forum should know about) has a good section about the bias used in the first stage buffer and why it has problems.

I often don't leave answers but I do leave hints.

[ayayay!]

Where did you find 10uA??

Is my math wrong?  I=E/R.  So 9.6 (actually wouldn't that be 9.1 w/ the zener?) divided by 1.8M ohms = .00001 (rounded up from .00000555...  splitting hairs now.)  Did I do that wrong?  Or is the further explanation towards the bottom of your post?  Good info, thanks.

In fact I wonder if someone mis-read 13 for 18.

Yes it's a 13.  The other 2 are 18. 

The 0.01u after the 68K is liable to shave a lot of bass. I do not know if that value is needed to make the filter resonance happen, desirable so heavy bass does not overload the high wah sounds, or leftover from some non-buffered plan. Faster to try a bigger value than to think about it.

Yes, I've both tried and thought.     Like I said above, I choose .1uF. 

I'm not trying to push the discussion towards what sounds like what.  That's all subjective anyway.  But my choice of topic was that this method of building a buffer hasn't been discussed much and it's worth noting. 

[DiscoVlad]

Well, it's certainly not perfect...

simulating the buffer circuit in spice shows the limitations of this type of buffer pretty well.

Bearing in mind that this is a simulation, and hence dependent on how good the component models are, the output is attenuated slightly, and also isn't symmetrical, which isn't good, but probably a worthwhile trade off for the greatly increased input impedance.

Of added note, the simulation results were independent of the transistors used (started with a pair of 2n5089, then tried BC547, BC549, 2n3904, and 2n2222) illustrating how much more "ideal" a Darlington pair is over a standard bipolar transistor.

It's a simulation so should go with a heaping pile of salt, I'd suggest bread-boarding just the buffer up, and putting it on a scope to see for yourself what difference (if any) the buffer actually makes, rather than a "vintage is better, just because" appeal to tradition.

Personally I'd rather use a JFET or better yet opamp (a JEFT type e.g. TL072) buffer on the input and output because they alter the signal less (i.e. they're actually clean!), while still giving the pedal decent input and output impedance characteristics. IMHO most of the "vintage" sound comes from the capacitor values (not the type of capacitor) around the wah pot anyway.

[Johan]

Okay so this is what many (sometimes myself) consider the "crappy buffer."  But really, it can't be that bad...

..for whats its worth..many players, myself included prefer the GCB95 version of this ancient circuit to the other because of the buffer..it takes she tonesuck away and lets more treble through the wha. that in turn makes it a bit more aggressive and work very well with the neck pickup on a lespaul through a marshall...try it...
J

[PRR]

DiscoVlad , wrong connection:

[DiscoVlad]

Oops! No wonder it didn't look like I was expecting Fixed now.




These are (hopefully) correct, and show that with the Darlington buffer there is a very slight phase delay, but otherwise the signal Output is unchanged from the Input.

Once again, because the hfe of the transistor pair is so high compared to a single transistor it doesn't matter which type of device is used.

[PRR]

> with the Darlington buffer there is a very slight phase delay

The Darlington's delay is insignificant for guitar frequencies.

I was fighting coupling-cap phase shifts... I bumped the output cap to emphasize the transistor action, the tone-shaping action is not our main concern.

At 0.7V peak input the positive peak gets a bit soft. This is not a big deal for guitar effects.



HARMONIC _ FREQUENCY _ _FOURIER _ _NORMALIZED
_ _NO _ _ _ _ (HZ) _ _ COMPONENT _ _COMPONENT
_ _ 1 _ _ _ 1,000 _ _ _0.6799 _ _ _ 1.000E+00
_ _ 2 _ _ _ 2,000 _ _ _0.0016 _ _ _ 2.461E-03
_ _ 3 _ _ _ 3,000 _ _ _0.0015 _ _ _ 2.323E-03
_ _ 4 _ _ _ 4,000 _ _ _0.0013 _ _ _ 1.965E-03
_ _ 5 _ _ _ 5,000 _ _ _0.0010 _ _ _ 1.597E-03
TOTAL HARMONIC DISTORTION =  0.42 PERCENT

This sim shows 94nA base current where I calculated 11nA. The 3904-3904 pair is not optimized for Darlington like the MPSA13 is, and my SPICE model may be unrealistically low-hFE for the very low current in Q1.

[PRR]

OTOH, with a simple non-Darlington, we can pass 2 Volts (three times more) with lower THD:



1 - 1,000Hz _ 1.898V    1.000
2 - 2,000Hz _ 0.0038    0.002
3 - 3,000Hz _ 0.0009    0.0004
TOTAL HARMONIC DISTORTION =  0.207 PERCENT

[ayayay!]

Very nice PRR, thank you.  Not a bad trade off using the 2n3904, I'd say.  Again, I would/do bump that output cap to .1 anyway so, for me at least, this would work.  Question though, why do you have the 68k pointing to ground?  nevermind.

I'll probably try this out for kicks.   This has been a fun one guys, thanks for this!  

[PRR]

Question though, why do you have the 68k pointing to ground?  nevermind.

As you figured: I see the 68K going out to the right. The NFB through the resonator "might" force a very low impedance after the 68K. Q2 base impedance too. But I know the external load is at least 68K. It may be near 100K as in Vlad's sim, whatever.

Since the buffer has 10K DC load and the external AC load is "at least" 68K which is significantly higher, I can assume that if it drives 68K fine it will drive higher loads fine. (If it did not drive 68K, I might have to look into what comes after the 68K, which is a resonant mess.... so I am glad to find that it drives >=68K fine.)

[DiscoVlad]

I made it 100k because I was treating it as a discrete buffer, and ignoring the Wah circuit (68k resistor onwards).
From a DC biasing standpoint the parts which matter are the 1.8Meg, and 10k resistor

Just did a simulation of Darlington vs. single transistor (using the BC547C model)

Darlington: Ib(Q1): 3.94814 nA
Single: Ib(Q1): 1.28266 uA

So the input impedance of the Darlington pair is considerably higher than a single transistor. This is probably why it's used.

[Gus]

If you are simming stuff.  Sim the bootstrapped EF in the link I posted before in this thread.  One could use this as an input buffer,  High input resistance and easy to find no adjusting needed use of BJTs no need for jfets

[PRR]

> Darlington: Ib(Q1): 3.94814 nA
> Single: Ib(Q1): 1.28266 uA
> So the input impedance of the Darlington pair is considerably higher than a single transistor.

Static DC current is a clue, but NOT a good measure of input impedance.

Stick some significant resistance in front, measure the voltage loss. You see I used 30K, an approximation of guitar impedance.

Darlington - 1,795K (**)
Single - 714K (*)

Darlington IS higher. But nothing wrong with 700K.

OTOH, at higher levels both break-over to near-dead-short.
(**) to +0.65V peak
(*) to -3V peak

0.65V peak is a big guitar signal, but not impossible. I'd be a bit miffed to track a slight tizz to an input stage which went short so near normal levels. It's also prone to RF pickup of strong/close radio transmitters.


I didn't notice that this is NOT the stock Dunlop but an addition. I was puzzled, because it does not seem production-quality. It surely works; IMHO it could be done differently.

I suspect that the modder used resistor values intended for a single transistor, thought "input impedance of the Darlington is higher!", and used the -13 without a full re-think.

The 10pFd input cap is pointless (there's a couple pFd in the transistor, 200pFd in your cable). And if you want to stop RF, you want both shunt and SERIES elements. Also some series resistance keeps an overloaded transistor from shorting the signal, and protects the transistor from stupid mis-pluggings (speaker outputs).



Input impedance is 2Meg up to 2V peak. When over-driven there's still 33K, not a diode-short (<1K). Freq response is -1dB at 20KHz but down 20dB in the AM broadcast band, 40dB at 5MHz (older police radio), nominally 50+dB at CB frequency. "Ooops!" inputs to 80V do no damage.