Common collector ( buffer) and measures to increase stability?

[Max999]

I am trying to make a very simple single transistor bjt buffer circuit. Now I have read that this design can oscillate. I have also found two possible measures to combat this instability.

Are these measures against instability/ oscillation even neccesary in our application?

If so, the solutions I have found on the net are these:

1 One option is to decouple the collector to ground with a capacitor ( C3), and add a resistor between the collector and the supply rail ( R4). R4 is proposed as 100r or less, a value for C3 is not proposed. Schematic is below. What would be a good value for C3?


2 Another option is to put in "base stopper", a series resistor connected at the base. Proposed value is between several hundred Ohms and 2K7. What would be a good value?

Would option 1 or two be a better solution to the problem?

[R.G.]

Have you ever actually made one of these oscillate, or only read about it?

[Transmogrifox]

I am trying to make a very simple single transistor bjt buffer circuit.

You will notice there is almost never anything done in common commercial stompboxes to address possible instability in emitter follower buffers.

Unintended instability is more of a hypothetical possibility rather than something that happens in practice. The single BJT emitter-follower buffer is hard to destabilize, even on purpose.

One example of a single BJT used in an emitter follower configuration that is unstable is one variant form of the Colpitts oscillator.  It oscillates when there is either an inductor or crystal connected at the base, and some kind of capacitive feedback.  Getting the combination of impedances onto the base as well as the correct parasitic capacitances to form the feedback from emitter to base is like shaking a bag of loose nuts and bolts and ending up with one of the bolts treaded into a nut.  It's possible, but you'll probably shake it until the bag breaks before you get that to happen.

I have to design a Colpitts oscillator on purpose to even get it to oscillate. 

That said, confirm this one first before you spend your time solving a problem that millions of stompboxes around the world don't have:

Have you ever actually made one of these oscillate, or only read about it?

If you do have one that actually oscillates, I can't give you a rule of thumb about what resistor value to put in the base because this all depends on the impedances connected at base and emitter.  For this you would experimentally find the resistor value that just barely makes it stop oscillating, then multiply by a factor of 5 or 10 to ensure good margin.

[Max999]

Thank you R.G. and Transmorgrifox.

R.G., I have only read about it. The problem I have is that I do not have a scope so it is hard for me to check if the ones I have breadboarded oscillated. I even do not know how the oscillation would manifest itself. Would the oscillation be a high pitched unhearable sound or would it cause anomalies on the ac wave?


Transmorgrifox, your anology is very reassuring. I am happy it is not easy to make a single bjt emitter follower unintentionally unstable.
Do you know if the commercial stompbox designs are all checked under scopes? I sometimes think the designer does not think it is neccesary because they are dealing with sound design and not sound reproduction, like in hi-fi.

[thermionix]

Ultrasonic oscillation would (I think) cut the signal out or at least seriously disrupt it.  If everything sounds normal, and you don't hear any squeals or motorboating, you're good to go.

[R.G.]

The reason I asked that was because I have never had an emitter follower oscillate by accident in over four decades of using them. I did once get an emitter follower to oscillate when I was trying to. It took about a day to get it to do so, and then it was barely able to sustain oscillation.

Followers of all sorts are prone to this, including opamp followers. The high open loop gain of an opamp follower makes them more prone to it than simple transistor followers. In opamps, the thing that kicks it off is usually capacitive loading on the output. Many opamps specify a maximum capacitive load for this reason.

My best advice is to simply use the follower the way you want and ignore the possibility of followers oscillating until you have some funny situation that cannot be explained any other way. My doctor tells me that medical school students are told that when you hear hoofbeats in the hallway, don't go looking for zebras; it's usually a horse, maybe sometimes a donkey or mule.

[Max999]

Thank you Thermionix and R.G., I will put the fear of having an oscillating emitter follower to rest.

I do have another issue with biasing the emitter follower ( for a volume pedal) that I have simulated in LtSpice.
Why is there 2.81V at the base here?



I want to bias it at 9.6V, is this even possible?

[thermionix]

Weird.  What happens if you halve R2?

[R.G.]

I do have another issue with biasing the emitter follower ( for a volume pedal) that I have simulated in LtSpice.
Why is there 2.81V at the base here?

Well, let's think. If we simply accept the results of the simulation as valid (and one should always question the results of simulation until proven to be rational) then the simulator is telling us that there is 9V on one end of R2 and 2.81V at the other end.  Ohm's Law tells us that there is then 6.19uA flowing through R2 into the base. We believe that the emitter cannot be vastly different from 0.6V below the base until otherwise proven impossible, so the emitter is sitting at about 2.21V. That implies what the emitter resistor is conducting 2.21V/3300 = 670uA. So the transistor is producing a DC current gain of 670/6.19 = 108.

That actually seems pretty reasonable, unless the transistor NPN model says it's to be vastly different. So it may  be working as designed.

I want to bias it at 9.6V, is this even possible?

Sure.  Assuming that the transistor's current gain doesn't get vastly different at different currents, then if the output at the emitter is 9Vdc, and you keep the emitter resistor at 3.3K, then the emitter resistor is conducting 9V/3300 = 2.7ma. If the transistor's gain is still about 108, you need a base current of 25uA. So you need to get 25uA into the base to make your wishes on the emitter come true.

If you make R2 = 1 ohm, the voltage drop across it will be 25uV, and so the base will sit pretty much exactly at the same voltage as R5 and R6 make it, and all will be well. I'm presuming that R5 and R6 will have their values changed a bit to make 9.6V instead of 9.0V.

But your input impedance will be only about 5K, the parallel combination of R5 and R6. Presumably that throws away the reason for having an emitter follower. You can make R2 be 100K, and so the 25ua through it will make a drop of 0.1M * 25uA = 2.5V. You could make this work by changing the values or R5 and R6 to make the voltage at their junction with R2 be 2.5V higher, at 12.1V, not 9.6V, and now you get 9.6 at the transistor base and 9V at the emitter, as well as an input impedance of 100K.

Still not what you wanted? Use a transistor with a higher gain. 2N5088 has a typical gain of 400, so the base current drops to about 6uA at Ie = 670uA, and now you can use maybe a 390K or 470K for R2. MPSA18 has an even higher typical DC HFE of about 600. You could also use a darlington device like MPSA13 or MPSA14 for the NPN, and get HFEs in the 3000 to 10,000 range. That gets you back to a 1M.

[Rob Strand]

Why is there 2.81V at the base here?

If the transistor gain is low you might see that.

Weird.  What happens if you halve R2?

Or increasing R1.
Or both.

If R1 and R2 are fixed in value then normally you would decrease R5 or increase R6.   It would be classed as a poorer solution over changing R1 and R2 if they are not fixed values.

[thermionix]

Derp.  I misread R2 as 100K (not used to seeing 1000K, except one old pot I have that's marked like that).  I didn't do the math, but thought it seemed like more than a reasonable amount of base current to get such a drop.

[antonis]

@MAX999: You've used a typical bias arrangement for MosFets but "destructive" for BJTs..
(MosFets don't mind about their Gate current but BJTs DO mind a lot..) 

If you insist of using BJT buffer and also want high input impedance you have to either use a second (cascaded) emitter follower or bootstrap R2 (significally lowering it's value..)

[R.G.]

OK, I should have done more recommentation along with the analysis.

@Max999: you're running into some classical problems of doing bipolar buffers the simple way. Bipolars have limited input impedance, and limited current gain.

You can do what you're trying if you use a high gain NPN, like the 2N5088, 2N5089, or MPSA18, or a darlington like the MPSA13 or MPSA14 (these are examples, there are many others) or decrease the current flowing in the transistor by increasing the emitter resistor so there is less base current needed to maintain the DC conditions. Or increase the bias voltage on the other end of the bias resistor. Or both.

Or you can go sneaky. You can use a capacitor from the emitter to the junction or R2, R5, and R6 to drive the bias voltage with the output AC voltage, a technique known as bootstrapping. This has the effect of making the >>AC<< input impedance be much higher than the DC impedance. This lets you decrease the value of R2, down to maybe 100K or so, have your biasing be as you intended, and still get over the nominal 1M input impedance you're probably trying to get.

[PRR]

> Well, let's think.

+1.

> Why is there 2.81V at the base here?

Ask Ohms Law. ALL electric problems work-out if you compute all voltages currents resistors.

In this case you also need to know the Base Current. BJT transistors have leaky inputs. A useful fiction is to assume a current gain, hFE. This can be 20 to 1,000 in different types, 3:1 different within the same type, and vary more than that over all the currents you might want to work a type at. I was taught "assume >50" (because the hFE~~20 parts were obsolete and hFE>100 still cost more). Today I assume I can pick an NPN off the junkpile and get hFE at-least 100. Indeed as R.G. calculated, you got (for your sim-pile part) 108. Frost on my sliderule gives 111.

Doing the math around leads to a drop in R2 of "30V", down from a nominal 9V source. Which is impossible. Really the current in R1 falls off, Vb and Ve fall-off, until everything balances. R.G.'s calcs show this happens very near your observation with a very reasonable hFE.



Input impedance may be (IS!) a concern. Obviously the 300K sucks. But the 3.3K reflects over as 3.3K*hFE or about another 300K, making 150K. If this is for guitar, that is low. Moreover it is similar to the nominal 100K pot you want the buffer to drive, so the buffer isn't buffering (only about 1.5X).

The 30:1 ratio between R1 and Rload is striking. Nice if you can do it. Usually leads to an extravagant design. Resistor-loaded amplifiers can work fine with DC-feed to AC-load ratios of 5:1 down to 2:1. For 100K load this suggests R2 could be 20K to 50K.

Note the large 0.006mA-0.030mA base current. This is not steady. Random fluctuations in this current will induce hiss, bypassed by whatever is connected as the source. Here the source is dead-zero (plus 10uFd) so no hiss. Real sources in g-world may be 5K to 500K. The random hiss in a large base current will be a large hiss voltage. This opens up a giant can of bees. However a BJT facing a guitar will probably want an emitter current closer to 10uA (0.010mA), and sub-uA base current, for optimum hiss parameters.

PLAGIARIZE!! This plan (Q6) and similar has worked on millions of much-used products. Figure out what the voltages and currents are for self-information. Or just steal the darn thing.
https://www.hobby-hour.com/electronics/s/schematics/od1-overdrive-schematic.gif

BOSS liked to use 2SC3378GR which is hFE=200-400 @ 2mA, and seems to be a low-current part. I would check BOSS's design assuming hFE of 200, then again with 400. This is in-line with R.G.'s suggestion of a high-hFE part, but not all that much higher. If you work it out, the higher resistor values improved input impedance more than the higher hFE.

Today the practical answer is to toss in half a TL072. Input current is near zero. Output signal current can be larger than <<3mA. Supply current can be less. THD of the EF may be small, but a TL072 will be smaller. I realize you are learning something. Sometimes what you learn is that factories can make amplifiers much better and cheaper than you can hand-build.



Simple standalone emitter followers hardly-ever oscillate.

I dunno where you found the idea to decouple the collector.

We usually de-couple there to reduce crap in the rail, from AC-derived power or sneak-around from other stages. A first try value for the cap would be greater-than the coupling cap we would use to pass signal along. Given 100r, I would pencil 100uFd for coupling, so >100uFd for decoupling. That's a huge cap for a small stage, and raises the question- why 100r? Is that optimum?

Oscillations: A capacitor load on the emitter causes an inductance looking into the base. There's always stray capacitance around the base, so there is a tuned circuit. There's always smaller capacitances here and there. With extreme bad luck it oscillates.

Series base resistor very close to base may help but is not the best way.

What is usually done, when fear of trouble is high, is a series resistor at the output (emitter). If this is significantly larger than hOE, any external load (capacitor) is swamped and the emitter does not look into a near-short at high frequency.

Maybe more of a problem than self-oscillation: we are now surrounded by "oscillators" which we call cellphones etc. If signals get into a low-loss cable with zero or infinite impedance at both ends, it bounces forever. For a very long cable the Characteristic Impedance is 50 to 200 Ohms, and that will kill the first bounce. But for audio lines less than miles long, a poor match will damp-out radio frequencies more than quick enough. If you want a "low" impedance, force it to be over 30 Ohms and mostly resistive (not just a B-E junction). If you don't expect <50K loads, 5K is an upper value for audio loading but <1K will damp radio better. I've run long lines in harsh places with 430r single termination and little trouble.

hOE of a transistor (Shockley's Law) at 1mA is 30 Ohms. 100r is a fair first-draft build-out resistor for most audio transistor chores. In the BOSS plan above, Q7 drives the Great Unknown, and works below 1mA (maybe 0.3mA, 100r hOE), should not face low-Z loads, so 1K was picked.

This has the added advantage of limiting the peak current when audio transients drive very long cables. I once burned a transistor testing 10KHz square waves into a simulated 300 foot cable. For that non-guitar case, 22 Ohms was plenty to limit the peak current. (And a real 300' cable would have many Ohms before the C built up.)

However if you are just going to a volume pot in the same box (a known and benign load), a series resistor is almost never necessary with jellybean transistors.

[Max999]

Thank you Thermionix, R.G., Rob Strand, Antonis and PRR. The wealth of knowledge you are all sharing is invaluable for me.
It seems that for a discrete input buffer of atleast 1M I can either bootstrap a BJT or use a Jfet. I have looked at the Darlington option in a Crybaby, and the input impedance is about 0.7M. The Jfet has a gain of 0.9 as a common collector and this is a bit low when comparing it to a bjt.

Combining R.G. and PRR's answers I came to the conclusion that plagiarizing a bootstrapped design is probably the way to go. So I started looking at the Pete Cornish buffer.

This is the P.Cornish buffer

When I do an AC simulation feeding it an AC source of 1V I get this frequency response:

There is a 4.8dB hump at 2.78Hz, which looks odd to me.
I can correct it by increasing C2 to a large value, say 220uF. The frequency response is then:

The input impedance curve also becomes flatter with using the 220uF, which is 1.65M in the human hearing frequency range.

Is there any problem with increasing C2 to 220uF? What does C3 do? When I remove C3 I have not encountered any change.

[PRR]

> There is a 4.8dB hump at 2.78Hz

So??

Which string do you pluck to make a 3Hz tone?

When I was a boy, measurements below 20Hz hardly ever happened. The signal generators didn't go lower. (And SPICE was for food.) This *did* lead to some unfortunate sub-sonic trouble, but rarely.

C3 probably cuts radio reception. You don't need it until the day you DO, say playing a truck-stop with a parking lot full of boosted CB transmitters. Or these days, a cellphone laying on the amp.

[antonis]

When I was a boy, measurements below 20Hz hardly ever happened.

Those were Galenium catwisker happy days... 

@Max999: Don't exaggerate C2 value 'cause there might occur "side-effect" problems..
(e.g. bootstrap circuit has a gain significally greater than unity at a frequency f = 1 / (2π*square root(C1*C2*R3*(R4//R5)) which "converts" the bootstrapped component into a higher impedance of opposite sign. A cap appears to be inductive and a resistance becomes a negative one. It doesn't necessarily imply that the circuit WILL oscillate, since other parallel resistances, including source resistance, may produce overall positive resistance but better not to mess with complex calulations of gain at this specific point..)   

To more practical issues, R4 & R5 values can be significally lowered to form a "stiffer" bias voltage divider (taking in mind not to significally load R6) and also lower R3 value for "excess" base current availiability..
(after all, an enormus bootstrapped resistance is useless when shunted by a lower one, like h_FE times R6 - actually h_FE*(R6//R7//R9//whatever comes next..)

P.S.
For educational purpose, I'd propose to simulate a direct coupled dual emitter follower (complementary or not) with multiple bootstraps..
(e.g. from second BJT emitter to first BJT base resistor & also to splitted Emitter resistor & to Collector for leakage current reduction (overkill..!!)
but it seems to me we've puzzle you enough for a humble emitter follower.. 

[Max999]

AH the good old days PRR. About the frequency hump, I was just not expecting it, I thought someone would always try to make a buffers response flat because everybody always wants the "guitar straight to amp tone". It also puzzles my why a design that should have less gain then 1 has gain to produce the hump. But it is my lack of knowledge ofcourse, nature's laws are not bent specially for me.

@Antonis. Sorry I did not respond earlier when you mentioned using a dual emitter follower. Do you mean something like this design?



I will surely LT SPice it because it interests me a lot. Can I PM you if I have further questions about it?

Back to the Cornish buffer, is it the consencus that the hump in  frequency respons is not an issue? I mean, there is positive feedback in a bootstrapped design, and when it is below 1 there is no oscillation. Does the hump, which is produced by having gain in the circuit ( right?), a non issue?

[antonis]

@Antonis. Sorry I did not respond earlier when you mentioned using a dual emitter follower. Do you mean something like this design?

I will surely LT SPice it because it interests me a lot. Can I PM you if I have further questions about it?

Of course you can P.M. anytime but, IMHO, an "open" discussion should be interesting for other people too..

Good catch above but why not take advandage of the second Emitter availiable and bootstrap also first BJT Emitter resistor..??



Some rules of thumb:

-When biasing a BJT with a resistive voltage divider, try to set its resistor values the lower the possible, for "stiffer" bias source..
(at least, their parallel equivalent resistance lower than 10 times h_FE*R_E - the lower the better, provided PS current availiability..)

-(Un)boostrapped resistor value is strongly depended on divider's resistor values 'cause it's consider in series with their Thevenin equivalent resistance..
Its bootstrapped value is simply calculated by R/(1-A) where A is follower's "open" gain and it's calculated by A = R_E/(R_E+r_e) where r_e = 0.025/I_C
Of course, its bootstrapped value should be considered in parallel with any other impedance "seen" from base (like h_FE*R_E) hence dominated by a lower one.. 

-Bootstrapping capacitor value can be precisely calculated but it should be a waste of time/brain cells due to component tolerance & complex formulae so its value is mainly set by experience..

P.S.
A further proposal should be to replace first Collector resistor with an output load equivalent and also bootstrap it via second BJT Emitter but I already see some guys comming to slap my hands..

[PRR]

> puzzles my why a design that should have less gain then 1 has gain to produce the hump.

"unity gain Sallen-Key"

If Q is 1 or more, the response has a hump, even though the amplifier is unity-gain. Over 40dB bump is possible with some very extreme values.

I'm not able to explain it even face-to-face waving my arms like phasors; some Googling may find clues.

The boot-strap affair is not identical, but has the same R-C-R-C part-count and can have similar result. (It may be more alike than it looks at first.)