A side question: Bad (high ESR, depletion of capacitance, leaky) filter electrolytics may also cause motorboating, no?
Best regards!
Best regards!
Not really a “side” question. If the caps do not present a low enough impedance, they cannot provide enough attenuation of low frequency feedback through the supply between the plates of the various stages. If the caps are bad they can’t provide the attenuation that they were supposed to.
As most of you will probably already know, the ESR of a capacitor is it's Equivalent Series Resistance, that is, resistance that is in series, or appears to be in series, with a capacitor, normally you want the ESR of a capacitor to be as low as possible because that series resistance causes a voltage-drop across it, high ESR reduces a capacitor's effectiveness when the capacitor is used to bypass any hum, and, or, noise on the +V supply rail to ground, the two main functions of a capacitor in an electronic circuit is to block DC (Direct Current), and to pass AC (Alternating Current), so, a capacitor should appear as an open-circuit to DC and a low resistance to AC, when a capacitor goes bad for some reason or another, it's ESR goes through the roof so to speak, a bad, leaky capacitor will appear as a partial, or, high-resistance short to DC, one place where you definitely do not want a leaky capacitor is right between the plates, or anodes of the Phase Inverter tube, and the Grid 1, or control Grids of your power tubes, otherwise the DC leakage will upset the power tube biasing which will make the power tube/s go into red-plating and eventual self-destruction, it could cause your expensive output transformer to burn-out too, a capacitor that has failed and leaked it's electrolyte, and dried-out will show a capacitance value significantly much lower than the marked value, so, it will be far less effective at doing its job in an electronic circuit.
Thanks!
My question's background was that I'm not able to guesstimate the impact of a filter electrolytic's ESR on the phase issue.
Best regards!
My question's background was that I'm not able to guesstimate the impact of a filter electrolytic's ESR on the phase issue.
Best regards!
Also note that the stability limit of a feedback amplifier is always 180 deg. around the loop. Sometimes it might look like 360 deg. , but that is polarity and not phase. Folk tend to use these terms as if they were sometimes the same, or at least sometimes related, but they are unrelated.
Phase is time and polarity is just the arbitrary way we connect the test probes, or the number of inverting stages. An odd number of inverting stages changes polarity but doesn't necessarily change phase.
It's a nitpick, but something that causes confusion if we let it.
All good fortune,
Chris
Phase is time and polarity is just the arbitrary way we connect the test probes, or the number of inverting stages. An odd number of inverting stages changes polarity but doesn't necessarily change phase.
It's a nitpick, but something that causes confusion if we let it.
All good fortune,
Chris
When an amplifier goes into self-oscillation for some reason at some particular frequency, at that particular frequency the positive feedback loop from output to input contributes a 180-degree phase-shift, with the amplifier circuit itself contributing a further 180-degree phase shift, the total combined phase-shift then equals 360-degrees, effectively the amplifier turns into, or becomes a phase-shift oscillator.
This misconception is so pervasive that plenty of knowledgable people accept it as true. An amplifier contributes part of the phase shift around the loop (amplifier + feedback network) and also contributes a polarity inversion, or not. The phase shift is frequency dependent but the polarity inversion, or not, is independent of frequency. One is a time shift, and the other is not - they're really unrelated.
For example, when testing for stability, we first make the feedback polarity correct (for positive or negative feedback), then we make the phase shift, through the amplifier and the feedback network, acceptably less than 180 deg. before unity gain.
All good fortune,
Chris
For example, when testing for stability, we first make the feedback polarity correct (for positive or negative feedback), then we make the phase shift, through the amplifier and the feedback network, acceptably less than 180 deg. before unity gain.
All good fortune,
Chris
I disagree with the assumption that the "lower the ESR the better" as a general concept. Some ESR can be quite beneficial at preventing Hi-Q resonances in LC filters.. Anything with RC filters, Yes.. the lower the ESR the better.
Chris, it is the criterion of stability in NFB: gain less than 1 at the frequency where the phase shifts 180 degrees. Alphabet, like Ohm's law.
You need huge DCR for motorboating. And of course, voltage drop due to current drawn by output tube if applied in phase to anode of a tube 2 stages before may cause positive feedback. When I was a teenager, my first guitar amp with Gu-50 outputs oscillated at infla-low frequencies when I dialed in bass in Baxandall tone control at full throttle. If I knew, I would add one more RC filter for B+ of my first 6J32P input stage. 🙂 But I did not know then.
"Polarity inversion" is a phase shift of pi radians. Ex:
cos(5)= 283.66291E-3
cos(5 + pi)= -283.66291E-3 = -cos(5)
(This follows from Euler's Identity: e^(jx)= cos(x) + jsin(x))
When you wrap a gNFB loop around an amp, the input is always the noninvert input, and the NFB is applied to the invert input. It's quite obvious when dealing with op-amps or OTL solid state topologies. Not so obvious when dealing with hollow state since the connection to the OPT determines whether you have a noninverting input or an inverting input. Murphy has a tendency to wire hollow state amps for positive feedback. When that happens, you need to reverse the connections, either at the primary side the finals drive (preferred) or the secondary (not preferred since the designer of that OPT selected the ground side of the secondary for a reason, and reversing it may not be optimal any more).
Polarity inversion also isn't guaranteed. This is seen when employing wide band amps. The hollow state wide band amp will likely require inductance to compensate for attenuation from device and stray capacitance. Not needed when using BJTs since you can always reduce the load resistor and increase the collector current to get back the gain lost. Still, you have device capacitance, and in both cases, you may not see a complete inversion of 180deg. At the high frequencies, you might have something like 120deg of phase difference in a common cathode/emitter stage.
You get phase misbehaviour at the extremes: high frequency phase shifts from device and stray capacitance, and the self capacitance and leakage inductance of the OPT. You may need RC networks to improve phase margin to avoid high frequency instability. Low frequency phase shift is due to bypass and coupling capacitors, so you need to stagger the time constants to prevent piling up phase shifts that can lead to low frequency instability.
If we reverse the connections to the secondary of a transformer, have we added or subtracted 180 deg. of phase? No, because phase is the time difference expressed as a rotating vector, and we haven't changed the timing at all. Polarity is just the choice of which orientation we connect our measuring device, and is independent of time. They are totally unrelated things that happen to look alike in special cases.
Put another way, an amplifier with three inverting stages doesn't have 540 deg. of phase shift before any time delays are considered - how would feedback work?
As I said before, many smart people have accepted this misunderstanding forever. And it confuses newcomers because it's incorrect.
All good fortune,
Chris
Put another way, an amplifier with three inverting stages doesn't have 540 deg. of phase shift before any time delays are considered - how would feedback work?
As I said before, many smart people have accepted this misunderstanding forever. And it confuses newcomers because it's incorrect.
All good fortune,
Chris
A bit off topic but it looks like the experts are here. Trying to troubleshoot hum in a tube amp build. It might have a bit to do with oscillation though. Was looking at the power supply ripple and saw about 100mv of 120hz ripple riding on about a few volts sine at 10-11hz?? Am I correct in thinking that the only way it could have that freq in PS is if it is on the edge of motorboating? Hooked up filter that consist of transformers breaking the ground connection at input, to try to fix hum. It didnt but then it did break into full motorboating. I'm assuming bc of additional phase of transformers. Any way adjusted feedback erlier in year to take care of motorboating and wondering if I need to reduce feedback a little more to be clearly stable? Also would like to know more about troubleshooting process for hum. Maybe a good SBS process or procedure. I'm trying to determine if it is from ground loop or just ripple in PS??
You definitely need to read this thread from the beginning, in order to figure out the root cause of your oscillations.
It can be as simple as output wire goes near input wire, or power stage and input stage ground wires are not wired properly to the power supply.
It can be as simple as output wire goes near input wire, or power stage and input stage ground wires are not wired properly to the power supply.
All I have is shorted and connected through a source with RCA jack cable. With shorted the noise goes away. But technically just way way down in level. I would say at least 20-30 db down, a very acceptable level. You can't hear it six inches away in dead quiet room, but can if ear is on the speaker cone. When source is connected you can hear it 3-5 ft away when quiet. Oh yeah the signal measured at the speaker is 50mV and is @180hz?? It is P-P design and I guess the 180hz is just the odd harmonic. I will retry with open and shorted and reply
Filter cap is very easy to rule it out with a LCR meter that can measure caps in the circuit (when power off of course).
You can also measure the AC voltage of Vb(B+) when it is oscillating. The ripple should be less than 1Vac on the Vb if the cap holds up.
Last edited: