So, I'm trying my hand at designing with feedback (just a simple CFB amp).
This is the amp (not a power amp, just an amp):
(Supply is 18V, as shown, through the wiper's action, it provides between 13dB and 36dB of gain).
This is the AC plot.
Taken from the output.
Now, apart from the fact that I don't really understand how the feedback actually works here (the output is in phase with the input, and I'm taking the signal from the output and I'm injecting it into Q1's emitter - shouldn't it actually be phase reversed? But if I try that everything falls apart).
I can see that (according to the simulation at least) there are no nasty peaks and, although not Butterworth smooth, at the lowest gain (and maximum feedback) there are no peaks in the response. That said, there is more than unity gain at well beyond phase reversal. I suspect that if I want to work out the phase margin, I should measure something else (and I know it's just a simulation, but I certainly haven't got a 100MHz oscilloscope anyway...), but the response seems not to have any oscillations going on. If I try and add a 30p cap between, say, base and emitter of Q2...
I expanded the range because there is that "interesting" gain increase at very high frequencies (I don't know if it's truthful at all, anyway it's still below unity gain).
Is this better? Is this CFB design inherently stable even without a cap? Am I even going in the right direction?
This is the amp (not a power amp, just an amp):
(Supply is 18V, as shown, through the wiper's action, it provides between 13dB and 36dB of gain).
This is the AC plot.
Taken from the output.
Now, apart from the fact that I don't really understand how the feedback actually works here (the output is in phase with the input, and I'm taking the signal from the output and I'm injecting it into Q1's emitter - shouldn't it actually be phase reversed? But if I try that everything falls apart).
I can see that (according to the simulation at least) there are no nasty peaks and, although not Butterworth smooth, at the lowest gain (and maximum feedback) there are no peaks in the response. That said, there is more than unity gain at well beyond phase reversal. I suspect that if I want to work out the phase margin, I should measure something else (and I know it's just a simulation, but I certainly haven't got a 100MHz oscilloscope anyway...), but the response seems not to have any oscillations going on. If I try and add a 30p cap between, say, base and emitter of Q2...
I expanded the range because there is that "interesting" gain increase at very high frequencies (I don't know if it's truthful at all, anyway it's still below unity gain).
Is this better? Is this CFB design inherently stable even without a cap? Am I even going in the right direction?
Feedback compares two versions of the signal of the same polarity. Their difference is the error signal.
Your circuit has two inverting stages, followed by a noninverting stage. So the feedback is of the correct polarity.
If you have much ringing at minimum gain, try a series RC network connected across R3.
As an initial guess, try 2k in series with 1nF. Also, scale down R4 and R8 considerably.
Your circuit has two inverting stages, followed by a noninverting stage. So the feedback is of the correct polarity.
If you have much ringing at minimum gain, try a series RC network connected across R3.
As an initial guess, try 2k in series with 1nF. Also, scale down R4 and R8 considerably.
I need a relatively high input impedance, what's the problem with those values? Too much noise? The noise analysis tells me that SNR is, like, 100dB as it is, and all concentrated in the lowest octaves.
Anyway, my (aborted) studies taught me that feedback is usually subtracted from the input signal. The idea is that you introduce an error that is opposite in phase to what the amp introduces itself, therefore balancing itself out (assuming the feedback is fast enough). And the "drawback" is that you lose gain, since part of the output is subtracted from the input. Is the emitter a "negative" port?
And why isn't the cap between base and collector of Q2 a good solution? As it is, bandwidth extends significantly beyond the audible range, I suppose slew rate might be lowish, but how low?
Anyway, my (aborted) studies taught me that feedback is usually subtracted from the input signal. The idea is that you introduce an error that is opposite in phase to what the amp introduces itself, therefore balancing itself out (assuming the feedback is fast enough). And the "drawback" is that you lose gain, since part of the output is subtracted from the input. Is the emitter a "negative" port?
And why isn't the cap between base and collector of Q2 a good solution? As it is, bandwidth extends significantly beyond the audible range, I suppose slew rate might be lowish, but how low?
I know this, but my idea is that if an amp oscillates in SPICE then it will almost certainly oscillate in real life.
Q1 needs DC base current. The R4/R8 divider must supply it while not being excessively loaded.
The simulation shows 2uA base current. I suppose what you mean is that bias stability is jeopardised by the high resistances (and therefore susceptibility to current variations).
Calculate the worst case DC base current (for the minimum beta) into Q1.
Make the unloaded DC divider current Vcc/(R4+R8) ten to thirty times that amount,
Make the unloaded DC divider current Vcc/(R4+R8) ten to thirty times that amount,
The stability of the CFB amplifier needs to be obtained using the open-loop (transimpedance) plots. However, I think the OP's curves are closed-loop ones.
Underlining #2, #6 & #8 about biasing the input bjt: way to high.
And in line with #4 & #9: closed loop plots from simulations.
These video-amplifiers are very stable and applied in numerous circuits. But don't step from the curbs, as it will become hazardous.
This modified video-amp circuit is at the edge.
With the given supply voltage, only a DC-analysis does show some difficulties.
With the given R4-R8, the base of Q1 is expected at 3.4V (ignore the base current as the collector current is very very low).
The emitter is around 2.8 or can be 2.9V or even 3.0V.
So the current through R9 (1k5) is near 2mA, mainly drawn from Q2 of course.
But 3.0V at the emitter plus 2mA through R11 (9k1) yields 3 +18.2V = 21.2, exceeding the supply voltage...
Q3 is then biased above the supply, rendering into nothing.
Another issue is R7. What is its purpose? A second feedback loop introduced. I'm not talking stray and substrate capacitances now.
If Q3 is not locked up by bias restrains, it's fortune telling what will become of this poor chap.
The C4 (time constant to consider) - U1 (pot) - R5 (notice delta{5k0} +82E !) as an AC-anchor?
In zillion applications of this circuit, you can be sure to omit this leg if the rest is cooked like high cuisine.
Build it. Proof & pudding.
And in line with #4 & #9: closed loop plots from simulations.
These video-amplifiers are very stable and applied in numerous circuits. But don't step from the curbs, as it will become hazardous.
This modified video-amp circuit is at the edge.
With the given supply voltage, only a DC-analysis does show some difficulties.
With the given R4-R8, the base of Q1 is expected at 3.4V (ignore the base current as the collector current is very very low).
The emitter is around 2.8 or can be 2.9V or even 3.0V.
So the current through R9 (1k5) is near 2mA, mainly drawn from Q2 of course.
But 3.0V at the emitter plus 2mA through R11 (9k1) yields 3 +18.2V = 21.2, exceeding the supply voltage...
Q3 is then biased above the supply, rendering into nothing.
Another issue is R7. What is its purpose? A second feedback loop introduced. I'm not talking stray and substrate capacitances now.
If Q3 is not locked up by bias restrains, it's fortune telling what will become of this poor chap.
The C4 (time constant to consider) - U1 (pot) - R5 (notice delta{5k0} +82E !) as an AC-anchor?
In zillion applications of this circuit, you can be sure to omit this leg if the rest is cooked like high cuisine.
Build it. Proof & pudding.
Spice won't emulate the physical layout good. Or , .... the quality or aging effects of the passives.
It does pretty good at emulating improper compensation. My Sonance , I was playing around with the 10pF lead cap . Went too small , -
and just as predicted - I had unsteady bias and a warm Zoble network. They used stupid old ceramics that were WAY off , a 15pF silver mica
made for a happy circuit !
OS