That many/most amps ring into capacitive loads is well known. They tend to ring more as feedback is increased and the interaction of the feedback network and the op cap is to blame. The mechanism is as described by all the literature out there (google ‘driving capactive loads’ ).
If this is true, then absent loop feedback that includes the OPS, the amp will not ring ? Is this true in practice ? (assuming no op inductors of course)
Does anyone have experience with testing sq wave ringing on amps without gnfb ?
If this is true, then absent loop feedback that includes the OPS, the amp will not ring ? Is this true in practice ? (assuming no op inductors of course)
Does anyone have experience with testing sq wave ringing on amps without gnfb ?
you can have oscillation with local feedback - even the emitter follower - the rules really are not at all different
if you want low audio output Z some form of feedback is the only practical choice - if you don't want odd behavior with arbitrary cable or load capacitance then you have to take the feedback loop stability consequences in to consideration - with local or global negative feedback
ringing at 100s kHz to MHz with a test square wave can be an indication of stability - or not
if you use a series L of a few uH the ringing on the Cload+inductor end is not a diagnostic of stability margin
if you want low audio output Z some form of feedback is the only practical choice - if you don't want odd behavior with arbitrary cable or load capacitance then you have to take the feedback loop stability consequences in to consideration - with local or global negative feedback
ringing at 100s kHz to MHz with a test square wave can be an indication of stability - or not
if you use a series L of a few uH the ringing on the Cload+inductor end is not a diagnostic of stability margin
Amps with global negative feedback do not necessarily ring with capacitive load.
The capacitive load together with the amp output impedance forms an R-C lowpass.
This R-C lowpass is an additional phase shift in the GNFB loop.
When the phase margin of the Amp gets too small the Amp rings. But if you include this
additional phase margin in the calculation of the GNFB loop the Amp will not ring.
You can even use the R-C lowpass phase shift to stabilize the Amp.
Generally an Amp without GNFB does not ring with capacitive load. But be aware of inductive loads (L-C) 🙂
The capacitive load together with the amp output impedance forms an R-C lowpass.
This R-C lowpass is an additional phase shift in the GNFB loop.
When the phase margin of the Amp gets too small the Amp rings. But if you include this
additional phase margin in the calculation of the GNFB loop the Amp will not ring.
You can even use the R-C lowpass phase shift to stabilize the Amp.
Generally an Amp without GNFB does not ring with capacitive load. But be aware of inductive loads (L-C) 🙂
Be aware that emitter/source/cathode followers are all very prone to oscillation with capacitive loads, so it's likely, I suppose, that an example which is nearly oscillating will exhibit ringing.
mooly thanks - so your posts show that ringing will occur regardless of the fact that the OPS is outside the nfb loop. Keith and jcx posts indicate the same.
udok - any chance you have examples of amps that have been tested and do not show the characteristic ringing per the last line of your post ?
udok - any chance you have examples of amps that have been tested and do not show the characteristic ringing per the last line of your post ?
Yes, lossy elements can do wonders, especially near output/feedback node. I like your solution/amp.
The output RL netwok will cause an amplifier to ring into a capacitive load, period. Even if the wavefrom on the input side of the RL is perfectly square. It's normal and it won't hurt anything. And you are far better off with this controlled, damped ring than an uncontrolled one that may even break into full oscillation if the RL is omitted.
Yes, the ringing that any series inductive component causes is easy to show even in LTspice. Omitting the inductor can be disastrous as you say (just think Naim) but I wouldn't take stability issues because of omission of the inductor as a generalisation.
Yup most of us get why ringing occurs when you have an op inductor.
The interesting question is why it seems to occur without the inductor (unless damped or compensated somehow)
The literature tells us that the extra pole caused by the cap load reduces the phase margin and therefore reduces stability. Ok fine but why then doesn't the amp oscillate into oblivion with any cap load instead of just having three or four damped cycles ?
Last edited:
If you were to hang a 2uF cap directly at the GNFB take-off point (with no other load), it would be much more likely to oscillate into oblivion. Even 6 feet of speaker wire or pretty much any additional loading is enough to prevent this in most cases, even without the damped inductor. But nobody really wants to tempt fate, or someone hell bent to testing stability. And amps with lots of global feedback are more likely to have phase margin problems than ones with less/none. Not a hard and fast rule, but the more poles you're trying to juggle the harder it is to walk the line.
The emitter follower acting as an output stage is characterized by a current gain that drops with increasing frequency. As a result, its output impedance rises with frequency.
An impedance increasing over frequency is equivalent to a series inductor inserted between the "inner" emitter and its terminal accessible by the real circuitry.
Having said this it is evident that this series inductance together with any capacitive load connected to the output creates a resonant tank - giving rise to more or less oscillations. And these oscillations may be observed even with simple emitter followers without any feedback.
An impedance increasing over frequency is equivalent to a series inductor inserted between the "inner" emitter and its terminal accessible by the real circuitry.
Having said this it is evident that this series inductance together with any capacitive load connected to the output creates a resonant tank - giving rise to more or less oscillations. And these oscillations may be observed even with simple emitter followers without any feedback.
And the more gain (and the more poles) the worse the situation gets. Anyone who builds an output stage triple without base stoppers deserves his fate.
Hello kasey,
There are really two different causes of ringing:
1.) Destabilisation of Amps due to capactive loading.
2.) Ringing of line inductance with capacitive loudspeaker loads
The two causes can be solved by an output inductance with a parallel R for damping.
The following simulation asume a perfectly stable Amp (simulated by a square wave generator) and shows what happens in case of L-C loading.
The left case has no ringing because the damping Resistor is right. The next case has ringing because of a wrong Resistor.
Be aware that the inductor is always present as 1 cm of cable has approximately 10 nH of inductance.
At high frequencies the impedance of the L is higher than the R, and therefore the parallel combinations is essential equal to R.
This R isolates the Amp from the capacitive load. Therefore most Amps have an output L-C.
Greetings,
Udo
Attachments
Hello wg_ski,
You are right - output triples are difficult to design.
An Amp with global negative feedback is always a second order system in practice (at least before the open loop gain goes down to one).
A first order system is an R-C low-pass. This has 90 degree phase margin, which is always stable and has no ringing.
A second order system has ringing when the phase margin is less than 45 degree.
It became unstable when phase margin is zero and open loop gain is still greater than one.
More open loop gain means that phase margin is getting smaller and if phase margin is less than 45 degrees ringing occurs.
Greetings,
Udo
- Status
- Not open for further replies.