Isn't the zobel just a band aid, shorting the HF to ground and preventing the HF from returning to the negative input and so interrupting the HF loop. The approach used by Oliphant must be better, in that it prevents the HF from forming at all.
Depends on what you consider a band aid. I would argue that the Zobel is a solution that allows the amplifier to work with a wide range of capacitive loads.
~Tom
~Tom
I am not saying remove it, but these tests show that the amp is already unstable even with a near perfect load.
First get it stable without the zobel, then test different load types to see what needs to be done to keep it stable.
I think if the Loop phase is kept to 0 deg. The amp is more likely to also be stable under more difficult loads.
First get it stable without the zobel, then test different load types to see what needs to be done to keep it stable.
I think if the Loop phase is kept to 0 deg. The amp is more likely to also be stable under more difficult loads.
Last edited:
If you want loop gain >> 1 and stability, you will have to accept at least 90 º of phase at the unity-gain frequency as you will need at least one dominant pole in the system. Assuming we're taking real systems with finite bandwidth here anyway.
~Tom
~Tom
I think Oliphant's first plot shows 216 deg of phase shift at 1 Mhz, Now some of this could be caused by the small cap at the input.
My point is to first get the amp stable without the zobel with easy load and only use the zobel to remove instabilities caused by different loads, this will also stop the op-amp and zobel resistor from getting hot and so improve the amps efficiency. this will help to prevent activation of the op-amps build-in protection circuitry.
This was my original reason for seeking stability and with yours and others help I can now somewhat understand the mechanisms involved.
My point is to first get the amp stable without the zobel with easy load and only use the zobel to remove instabilities caused by different loads, this will also stop the op-amp and zobel resistor from getting hot and so improve the amps efficiency. this will help to prevent activation of the op-amps build-in protection circuitry.
This was my original reason for seeking stability and with yours and others help I can now somewhat understand the mechanisms involved.
The LM3886 is effectively a quasi output stage. The PNP transistor in the bottom half has a poor frequency response. To make the amp stable you slug the output with a zobel network. You need a low value resistor and a big capacitor. You cannot arbitrarily make the time constant the same ie small capacitor and big resistor and get the same effect.
Interesting discussion by the way.
Interesting discussion by the way.
As far as I can tell, Oliphant shows the closed loop response. For stability, however, we want the OPEN LOOP response. My new toy should be here tomorrow, so I will finally be able to take some of these measurements myself and show what I mean.
~Tom
I am starting to understand what you mean. I think you are looking for the poles in the open loop response to estimate the gain and phase margins.
http://www.diyaudio.com/forums/soli...as-lateral-output-perfect-38.html#post2675591
I never did, but another Member did post the link to the paper.
http://www.diyaudio.com/forums/soli...ility-techniques-zobel-et-al.html#post1653831
http://www.diyaudio.com/forums/soli...as-lateral-output-perfect-38.html#post2674088
http://www.diyaudio.com/forums/solid-state/147301-jlh-100w-l-mos-2nd-life-2.html#post2619567
http://www.diyaudio.com/forums/solid-state/94029-when-output-inductor-needed.html#post1105724
Last edited:
I see it as passive filter versus active filter. Like blocking DC by placing a large cap at the output or a smaller cap in the feedback loop. Place a filter in the feedback loop to block or prevent HF?
Andrew: Right on! Thanks. In particular, the link to "the math" is useful to me. I think I'll have to run some LM3886 sims with a speaker model for stability and explore the Zobel network some more.
~Tom
~Tom
You'll notice I have been harping on about this since soon after joining.
The HF route for currents has to be SHORT, else the HF will look for a lower impedance route.
Or, do you not accept that either, because I have not "proved it"?
The HF route for currents has to be SHORT, else the HF will look for a lower impedance route.
Or, do you not accept that either, because I have not "proved it"?
The "speaker Zobel" has nothing to do with the amplifier Output Zobel, nor the Output Thiele Network.
I'm not asking for proof. What I'm after is the understanding and the "why"s. A response along the lines of "you must do X" is not of any use to me unless it's followed by "because of Y".
A speaker Zobel and amp Zobel perform the same function - they flatten the impedance imposed on the amp across frequency. This improves amp stability.
"The Math" is useful to me because it includes a model of a speaker that I can punch into my circuit simulator. This will allow me to model the effects of various Zobel networks.
~Tom
Tom,
The "Inductive" Loop area for a speaker Zobel (If fitted) makes it useless to serve as an Amps Zobel network whose function is to insure a "Low" impedance load to the amplifier output stage at High frequency's (and hence why it must have a low impedance path at RF as seen by the amps OPS) - without this Zobel the output stage will have a resonance peak (or multiple peaks depending upon topology) somewhere typically between 2MHz to 20MHz depending upon output devices etc.
Without a Zobel the stability margin will be compromised - it must be noted that its a Local stability problem - local to the output stage which will effect the overall loop feedback stability. You might have plenty of PM for the overall feedback loop, while without the Zobel the output stage will tend to suffer a "local" stability issue.
The "Inductive" Loop area for a speaker Zobel (If fitted) makes it useless to serve as an Amps Zobel network whose function is to insure a "Low" impedance load to the amplifier output stage at High frequency's (and hence why it must have a low impedance path at RF as seen by the amps OPS) - without this Zobel the output stage will have a resonance peak (or multiple peaks depending upon topology) somewhere typically between 2MHz to 20MHz depending upon output devices etc.
Without a Zobel the stability margin will be compromised - it must be noted that its a Local stability problem - local to the output stage which will effect the overall loop feedback stability. You might have plenty of PM for the overall feedback loop, while without the Zobel the output stage will tend to suffer a "local" stability issue.
Last edited:
John, that's all true.
I would, however, argue that the Zobel at the speaker serves to smooth out the impedance curve, hence, make the amp see a relatively constant load versus frequency. Same can be said for the Zobel by the amp output. It serves as a load for the amp at HF, hence, makes the amp see a relatively constant load versus frequency. The only difference I see is that the speaker Zobel may be active at audio frequencies whereas the amp Zobel is active above the audible frequency range. One can't replace the other, but they serve the same purpose. Am I seeing this the wrong way?
~Tom
I would, however, argue that the Zobel at the speaker serves to smooth out the impedance curve, hence, make the amp see a relatively constant load versus frequency. Same can be said for the Zobel by the amp output. It serves as a load for the amp at HF, hence, makes the amp see a relatively constant load versus frequency. The only difference I see is that the speaker Zobel may be active at audio frequencies whereas the amp Zobel is active above the audible frequency range. One can't replace the other, but they serve the same purpose. Am I seeing this the wrong way?
~Tom
Tom,
The problem is that the Speakers Zobel is effectively isolated from the amplifier output by the inductance of the speaker cable (and that's ignoring the almost certain possibility that the Amplifier will have a series inductor on its output to isolate the output stage from capacitative loads) - thus it offers anything but a constant load at HF - and pretty much useless by anything over 100KHz. The OPS Zobel needs to be effective from say 2MHz to 20MHz and hence the requirement for it to be situated as close as possible to Output devices (Output pin) as possible and PSU Ground.
Basically the Speaker Zobel (if indeed its fitted) offers little to NO termination to the output stage at HF/RF.
The problem is that the Speakers Zobel is effectively isolated from the amplifier output by the inductance of the speaker cable (and that's ignoring the almost certain possibility that the Amplifier will have a series inductor on its output to isolate the output stage from capacitative loads) - thus it offers anything but a constant load at HF - and pretty much useless by anything over 100KHz. The OPS Zobel needs to be effective from say 2MHz to 20MHz and hence the requirement for it to be situated as close as possible to Output devices (Output pin) as possible and PSU Ground.
Basically the Speaker Zobel (if indeed its fitted) offers little to NO termination to the output stage at HF/RF.