It provides a HF load for the amplifier, hopefully designed to compensate for the rising impedance of the speaker/crossover load. Much in the same way that the load (assuming no Boucherot cell) is attached to the amplifier, the node that the output/zobel and feedback interconnect (this is the node that feedback is trying to null with).
Magma
They are selling albums faster than ever, partly thanks to endless touring. I've seen them four times here in the US over the last 30 years, most recently this summer in Atlanta, and Christian Vander was drumming and singing as good as ever. The young contingent in the crowd was even more enthusiastic than us old guys... Fantastic band, in my top five bands live ever in 45 years of concertgoing. for those uninitiated into the world of Zeuhl music:
https://www.youtube.com/watch?v=FlJKKtgreqw
If you can't understand the words, it is because they sing in Kobiaian...
Howie
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill, NC
WXYC Chapel Hill, North Carolina - 89.3 FM
They are selling albums faster than ever, partly thanks to endless touring. I've seen them four times here in the US over the last 30 years, most recently this summer in Atlanta, and Christian Vander was drumming and singing as good as ever. The young contingent in the crowd was even more enthusiastic than us old guys... Fantastic band, in my top five bands live ever in 45 years of concertgoing. for those uninitiated into the world of Zeuhl music:
https://www.youtube.com/watch?v=FlJKKtgreqw
If you can't understand the words, it is because they sing in Kobiaian...
Howie
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill, NC
WXYC Chapel Hill, North Carolina - 89.3 FM
This is a good question and deserves a proper response. I'm guessing you mean why is the output L/R network not inside the feedback loop.
A plain output stage has an output impedance below 1R generally, up to several MHz or more. This impedance is low enough to drive say cable capacitance without causing a lot of phase shift in the frequency range critical for stability.
After you add the L/R network, the output impedance becomes inductive and much higher at RF. This impedance has a lot more trouble IE phase shift when driving a capacitive load, and will even resonate since it is inductive.
In order for the feedback loop to be stable, the phase shift has to be kept down. The more phase shift you have to avoid, the less feedback can be allowed before it becomes unstable. The extra phase shift caused by the L/R does not interfere with the feedback since it is not inside the feedback loop. But what if it were?
Without any load, the amplifier would be stable, but you can't predict what load will be connected and how far that would push the phase. So now you have an amp that malfunctions depending on what's connected.
There are some benefits to putting the L/R in the feedback loop such as reduced output impedance, and the possibility of reducing distortion caused by the inductor if there is any. However the conductance you gain is insignificant compared to the resistance of your cable and of the voicecoil itself.
There was a Russian design posted some time ago that had a special network to put the L/R in the feedback loop without reducing stability, but for the reasons I just gave I don't think it is really useful, unless your amp requires a huge 2uH inductor to stabilize it, in which case the problem is arguably elsewhere.
Additionally, the reason the inductor/resistor is there in the first place, is because the amp is already failing to reduce it's output impedance at RF (causing as a result instability into capacitive loads). The L/R corrects for this. But when you put it inside the feedback loop, it is effectively in series with the output stage and worsens the impedance the feedback loop is already failing to correct. By adding the L/R you take one step forward. By putting it inside the feedback loop you worsen the problem it is correcting, and take two steps back.
Last edited:
Interesting point, I didn't really think about what a zobel would look like as part of the feedback network itself.
Daniel,
I do understand that we use the word Zobel for multiple positions and we can have a Zobel network hanging across a loudspeaker network flattening the impedance rise and another one in series with an amplifiers output to stabilize the amplifier. I just wondered why we couldn't include that output inductor/resistor network in the amplifier feedback circuit. There is actually a data sheet note now from TI on having a feedback network around the output inductor on their class-D amplifiers that helps to lower the THD numbers of those devices so wondered why you couldn't do the same thing with a class-AB amplifier. I take it that there are very real differences between the two classes of amplifiers.
I do understand that we use the word Zobel for multiple positions and we can have a Zobel network hanging across a loudspeaker network flattening the impedance rise and another one in series with an amplifiers output to stabilize the amplifier. I just wondered why we couldn't include that output inductor/resistor network in the amplifier feedback circuit. There is actually a data sheet note now from TI on having a feedback network around the output inductor on their class-D amplifiers that helps to lower the THD numbers of those devices so wondered why you couldn't do the same thing with a class-AB amplifier. I take it that there are very real differences between the two classes of amplifiers.
I think for class D it can help - ferrite cored power inductors are certainly not linear and feedback can help improve things up to a point.
However, I just don't see the value in doing it with a conventional linear amplifier. Some practitioners (JC, NP, CH) are even ditching the inductor. If you are running your amp without feedback i.e. open loop, then of course you don't need it anyway.
If you do put it inside the feedback network in a conventional linear amplifier, you have to deal with the load capacitance + output inductance ringing which can be quite severe up at 100 kHz +. Then there's the issue of what load capacitance range to design for since you have no control over the kind of speakers the customer will use.
Naah.
However, I just don't see the value in doing it with a conventional linear amplifier. Some practitioners (JC, NP, CH) are even ditching the inductor. If you are running your amp without feedback i.e. open loop, then of course you don't need it anyway.
If you do put it inside the feedback network in a conventional linear amplifier, you have to deal with the load capacitance + output inductance ringing which can be quite severe up at 100 kHz +. Then there's the issue of what load capacitance range to design for since you have no control over the kind of speakers the customer will use.
Naah.
Last edited:
There has to be limitations to that because the inductor on the output is to reduce the carrier signal, which you can't remove. I suppose feedback could reduce it to a minimum. That or maybe feedback for the audible sound is improved by not attempting corrections above the audible level? Either way I'm not so sure it's accomplishing what you're thinking.
Zobels are RC network with specific function. Where you put the zobel will depend on what you are trying to achieve (stability related or impedance related)
It can, depends on what you are trying to achieve. Feedback is to correct the input signal by sensing the output. You can choose at which point on the output you want to take the feedback from. Directly from the transistor output leg, after small resistance, after output inductor, after LC filter (in case of class-D) or even after the speaker load! (Something like Trans-Amp by Joe Rasmussen)... The only caveat is the stability issue because of (stray) capacitance/inductance.
I know that you might have got the answer by now, but if you build or at least read the trans-amp thread, you will know better the benefit and disadvantage of different ways to take the feedback signal for comparison with the input signal...
While that isn't wrong, it's a bit loose of a definition. I think that would be most confusing to people that don't understand it. It's more like it has two inputs, from this perspective, one is the reference and one is the correction.
As Keantoken said, having the zobel in the feedback means it's in series, instead of parallel. His explanation was great.
Maybe there is a way to stabilize an amplifier inside the the feedback, but it would likely be excessively complicated and ridiculous compared to a small cap and resistor. Would it correct the cause of the problem? That's another question because even if your amp is stable due to a special feedback, that doesn't mean it's not having an oscillation, it just isn't over-compensating to the point of instability.
That was not supposed to be regarded as definition.
Have you ever wondered why some people don't understand something? It can be because it is too complex. Simplification is a common way to break complex parts into smaller number of simple parts.
What is an "input"? What is a "reference"? What is the "correction"? You introduce parts that require further explanation but you didn't include the explanation (like Kean did).
Now look at my statement: Feedback is to correct the input signal by sensing the output.
Everybody knows more or less what a "feedback" is. "To correct the input signal" tells us that the input can be wrong and can be corrected. "Sensing the output" tells further that the input correctness has something to do with the output...
Details of the simplified statement can be searched but the simplified statement must be "complete" by its own. But that is not even the point because I know Kindhornman knows what a feedback is...
If you talk about a series zobel L//R, the object is to drop off the output level at higher frequencies, to avoid loading the amp too much there.
If you include it in the feedback loop, the effect of the feedback is to make the zobel disappear; the feedback 'tries' to keep the output up even at higher freq.
So that doesn't seem to be a smart thing to do.
Jan
If you include it in the feedback loop, the effect of the feedback is to make the zobel disappear; the feedback 'tries' to keep the output up even at higher freq.
So that doesn't seem to be a smart thing to do.
Jan
The reason for the output inductor is not well understood here. The major component is the added output resistor (1-10 ohms) that buffers the load capacitance from the output. The inductor is simply added to bypass the series resistor at low frequencies to keep the output impedance at audio frequencies low. The output inductance is usually between 0.5 and 5uH, 2uH being typical.
I usually avoid the resistor-inductor combination in my power amps. This takes a bit more internal compensation to keep the amp stable.
I usually avoid the resistor-inductor combination in my power amps. This takes a bit more internal compensation to keep the amp stable.
Fully agree John, and if you can do without it with no adverse effects, more power to you!
But IF you use it, then wrapping feedback around it defeats the purpose - pretty daft
Jan
- Status
- Not open for further replies.