Jim Williams' Test Your Analog Design IQ #17

Another Jim Williams analog IQ test! Can we identify and stop the source of oscillation in a simple voltage reference circuit?

[@ksbs2036]

I just watched this for a second time after watching it a few years ago. Really enjoyable video. Thanks for making it!

[@Infinitesap]

I really wish you would do more of this. It is some of the best material on YT. Very informative and useful for me. Thanks for the stuff you already did :)

[@nplanel]

Excellent ! That's how all electronics channels should do, by showing use case and explain how that works internally. Thanks to the Master Jim Williams and You !

[@Analogzoo]

np Thanks, really glad you enjoyed it!

[@InssiAjaton]

Way, way back when, before the uA709, we had our own op-amps. Then there were molded block op-amps from Philbrick and others. We tried one and found it needed an emitter follower to boost the output current. Just like this story tells, we met an unexpected Colpitts oscillator. And my choice was the same, some resistance to the base of the added transistor. Well, it worked, but we did not keep using those modules for too long, as the uA709 (and later uA741) were as good, much smaller and also kept dropping in price.

[@typedef_]

Here's how I would have tried to solve it. I generally don't distinguish between ordinary ringing and continuous oscillations so I would begin by viewing the circuit in it's oscillatory state as a second order system with a damping ratio of 0, oscillating at a frequency F1. Based on that, there is clearly a capacitance between the npn emitter and ground. Call that C1. I would put a very small capacitor C2 in parallel with C1 and observe the new frequency of oscillation which would be smaller. Call that F2. The value of C1 can now be found easily and it is equal to (F2^2 * C2)/(F1^2 - F2^2). Now that you know C1 and F1, you can calculate a value of a virtual inductor L from F1 = 1/(2*pi*sqrt(L*C1)). This is your undamped LC system. You can now calculate a resistor value R in order to make the damping ratio equal to 1 for critical damping. The formula for that is R = 2*L/sqrt(L*C1). I would of course use a slightly bigger value to make it a bit overdamped (but not too much hehe). Put that in the base of the transistor and it should work I think...

[@DevangsinghSankhala]

That was a nice explanation! I got something similar using control system transfers functions as well. Of course I used the 2nd order equations directly and selected R for a large damping. Got it!

[@christophschuermann6512]

Very good video! Jim was a great guru in the analog domain and not only there. You could also go for the 10 ohm resistor, if you add some small changes to the circuit.

[@TecKonstantin]

Sorry for the late comment, you can also think in terms of impedance. The follower has a negativ input impedance this negative impedance presented to a resonant circuit will force an Oszillation, by adding the 100ohm the impedance becomes positive so no Oszillation is possible.

[@HaykDingchyan]

Yes you are right ! We should put Rbase=|-Rinput| to compensate .

[@dancollins1012]

Great vids, am going yo watch your whole series, thanks

[@ismagine]

Thanks for the videos, your channel deserves more viewers! Any plans to make new videos??

[@electronicsprojects4708]

Great series, are you going to go through more of these exercises? I downloaded the pdf, thanks for sharing these.

[@DiaconescuAlexandru2024]

I initially thought it would be due to the trace parasitic inductance from the 15V supply to the collector of the transistor which formed an LC oscillator with the collector-emitter parasitic capacitance. If all else would have failed I'd have added a ferrite bead around the pin of the base.

[@w2aew]

Nice!

[@Analogzoo]

+w2aew Thanks! Tracking down unwanted oscillations is always fun; there's so many different ways that unwanted feedback & phase shifts can creep into even simple circuits. :)

[@Analogzoo]

+w2aew Just now noticed this got flagged as spam! That should be fixed now. Excellent advice and video as always! The 350MHz bandwidth on that 2467 is nice. I've been looking at picking up an older 485; the first scope I ever used was a 454A and since then I've always liked the 400 series scopes.

[@w2aew]

I love my 485 too. Extremely sharp focus on that CRT.

[@catalin3407]

Never thought that the parasitic inductance and low capacitors of the transistor can make oscillate a circuit

[@cyberlightbeing]

Awesome!

[@ragavanparamasivam]

hello sir,can u suggest any books for circuit designing with concepts aswell the problems?

[@Ritam_404]

can you make more videos from this book. Please

[@sorova]

Really need a Bode plot though, show phase margin etc

[@ottoomen5076]

what happened to this guy?

[@kishoresaldanha815]

how do you know it is the Cbe and not the Ccb (Miller Capacitance ) of the common emitter stage that is causing the oscillations. the resistor added to the base could form a first order low pass filter with input miller capacitance (Cmi) and damp down the oscillations .

[@Analogzoo]

+Kishore Saldanha Because the transistor stage is a non-inverting common collector configuration, not a common emitter. The Miller effect increases the capacitance between the input and output of an amplifying stage only for inverting configurations: en.wikipedia.org/wiki/Miller_effect

[@kishoresaldanha815]

+devttys0 thank you,sir

[@Infinitesap]

Awsome :-)

[@antigen4]

just take out your 30pf cap to make it oscilate - or use a different op amp

[@Jarrettmonty99]

But then that's a different source of oscillation and not the desired parasitic oscillation

[@t1d100]

Woot!

[@deepcompile529]

great

[@steverobbins4872]

I have to disagree with your Colpitts explanation for the oscillations. The resistor would have a SERIES parasitic inductance, not PARALLEL. (Sorry, for the caps. No screaming, just trying to be clear.) I think a better explanation has to do with the bandwidth of the NPN. The 2N2222A is not a high freq transistor. At just a few MHz the collector current can lag the base current by 90 degrees, or more! (This fact is important for understanding many BJT oscillator circuits.) Try a quick simulation with LTSPICE, and see for yourself. The LM301 (with 30pF compensation cap) used as a voltage follower will have a bit less than 90 degrees phase margin. That's why it worked when you connected the feedback network directly to the output of the opamp. But the additional phase shift introduced by the NPN can easily reduce the phase margin to zero. BTW, I noticed that some have suggested that Miller capacitance was somehow involved, but this is clearly not the case because the collector is tied directly to the +15V rail with no resistor, so you can't get a multiplying effect on the input capacitance.

[@Analogzoo]

Thanks Steve! I agree with you that the Miller effect is not the culprit here. But you’re saying that the stray parasitics are not what cause the oscillation? That’s interesting, as all the explanations that I’ve seen regarding BJT amplifier oscillation - particularly emitter followers - list stray parasitics as the fault. I think I understand your premise though: it is not the BJT that is oscillating, but rather the LM301; the BJT simply adds sufficient phase shift such that the total phase shift around the feedback loop is 180 degrees, causing the LM301 to break into oscillation. If that is true, then I have some additional questions: 1) Setting up a simple 2N2222 emitter follower circuit in LTSpice, I do see the phase difference in the collector current vs base current. What I don’t see is a corresponding phase shift in the base voltage vs emitter voltage. Given the voltage feedback nature of the op amp, I would expect that there would need to be a phase shift in the voltage being fed back from the transistor’s emitter to the op amp’s inverting input in order to degrade the LM301’s phase margin. 2) Assuming that the surrounding circuitry is causing a sufficient phase shift to support oscillation, how is the externally compensated LM301, with a typical GBW of 1MHz, oscillating at over 20MHz? I would not expect it to be able to provide sufficient gain at a frequency more than 20 times its GBW in order to support oscillation. 3) How does this explain how emitter follower circuits which are *not* part of a feedback network (e.g., just a simple buffer amplifier) break into oscillation? Oscillation in discrete amplifiers is a common phenomenon, and emitter followers are particularly notorious for it.

[@steverobbins4872]

Ooops!! You are right. There is no voltage phase shift from the base to the emitter. I just assumed this was the same issue I faced on a different circuit long ago, but it's not. And you are also right that parasitics must be involved because the frequency of oscillation is much higher than the loop crossover freq; specifically, there must be some sort of resonance that provides enough gain to overcome the -40dB/decade roll-off (the LM301 seems to have a second pole just above 1MHz). Perhaps it is something that isn't modeled in LTSPICE (because the Bode plot looks good) such as the lead inductance of the NPN. I think that would produce a series resonance looking into the base because of Miller effect. That would explain why the base resistor helped; it reduces the Q of the resonance. Using a surface mount version of the 2N2222A might prove this hypothesis. BTW, it's a pleasure watching your vids. You're a very sharp guy. And yes, I did subscribe to your channel.

[@steverobbins4872]

Oh, I just got it to oscillate in LTSPICE. Just 10nH in series with the base, and 10pF in parallel with the load makes is oscillate at very high freq in the transient analysis. And I didn't even add the second pole to the opamp model, I'm just using the ".lib opamp.sub" macromodel. But I'm still not seeing exactly where the problem is on the Bode plot. Perhaps you could figure it out because I have to get back to work. But this is fun.

[@Analogzoo]

The BJT's lead inductance plays a role I'm sure; in the video the solderless breadboard and the jumper wires connecting everything together are certainly contributing plenty of parasitics too!

[@paulperry7091]

@@steverobbins4872 1 cm of wire has an inductance of about 10nH. Probably not a coincidence.As for those breadboards.. Bob Pease called them "slabs of trouble", and he was right!

[@twobob]