I disagree entirely.
Disregarding Cherry's compound BJT attempts, the use of MOSFETs with grounded sources and a floating supply delivers exemplary results - with reliability as a bonus. I have also implemented this with an AD797 but the difference between that an a run-of-the-mill 5534 is negligible in terms of distortion due to the added open loop gain supplied by the output stage. Noise was improved slightly as expected with the AD797 but stability needs attention in both cases.
I have already stated that for MOSFETs with source connected to the case that the floated supply is a very neat solution, allows the case to be earthed directly to the heat-sink, no thermal washers, mechanically foolproof, safe.
If it's done with lateral FETs, as I expect, then the result should be very robust indeed.
I did not say that the floated supply is bad, only that I am not yet convinced.
Surely the supply capacitance to earth must appear as a load?
Usually capacitive loads are not desirable, especially as we try to push up the gain crossover frequency to reduce distortion.
Best wishes
David
Whether it's a single ended amplifier runing from a single polarity supply or a push pull amplifier that is running from a single polarity supply, they both have to drive the ground connected load via a large DC blocking capacitor.
When you move to a dual polarity supply push pull amplifier driving a ground connected load, then you will find that the load currents come via the PSU capacitors. They have the same effect as the DC blocking capacitor of the single polarity supply versions.
In all of these topologies the load current effectively comes through big electrolytics.
Move to one of the grounded output topologies and you find the same current flow arrangement. The big electrolytic capacitors have to pass the load current.
Effectively all these apparently very different topologies ALL pass the load current through big electrolytic capacitors. It as if the load is always in series with the capacitors with respect to the amplifier.
When you move to a dual polarity supply push pull amplifier driving a ground connected load, then you will find that the load currents come via the PSU capacitors. They have the same effect as the DC blocking capacitor of the single polarity supply versions.
In all of these topologies the load current effectively comes through big electrolytics.
Move to one of the grounded output topologies and you find the same current flow arrangement. The big electrolytic capacitors have to pass the load current.
Effectively all these apparently very different topologies ALL pass the load current through big electrolytic capacitors. It as if the load is always in series with the capacitors with respect to the amplifier.
My usual case for simulations and comparisons is @ 100 W.
That's about nominal full power with just one pair of output transistors, makes it easy to draw.
The number is only a benchmark at the moment, no simulation yet, but I don't see why the AD797 + CE shouldn't do it.
Best wishes
David
My concern is not with the CE amp, my initial impression is that it's an excellent idea, that's why I want to study it more closely.
My concern is with the floated supply implementation, that is the detail that looks vulnerable to noise coupled from the mains.
Is there a schematic of the ATC that we can both refer to in discussions?
Would make it easier to discuss which nodes couple to where, and so on.
Best wishes
David
I think you have missed my point.
The issue is not about current thru the power supply capacitors, the problem is that the floated supply moves with respect to earth, so that now the amplifier load includes the stray capacitance of the supply to earth.
This does not happen when the supply is earthed.
Best wishes
David
This does not have to be a problem.
With 'conventional' amps, worst case stability is usually with just 1-10n on the output. Once you get to 100n, stability usually gets better.
This is one interpretation/explanation of Cherry's favourite 'Zobel' network which has the usual R//L in series and then 100n across the speaker terminals. No 'conventional Zobel'.
I haven't analysed Vanderkooy's stuff in detail so be warned I might be pontificating out of the wrong orifice.
But I'd be more worried about RFI & other Mains crud getting into the feedback loop than of stability.
Attachments
It's obvious that these are all special cases.
One can consider the matrix with all the possible interconnections of the various sections.
So there may be some local feedback around the VAS and OPS (a la Cherry) combined with some around the Input and VAS (a la JLH and Cordell), some around the VAS only, etc
All of these can vary with frequency, all independently.
Then it should be possible to mathematically optimize the whole catastrophe to minimize the distortion.
That's what I have been at work on for a promised article for Jan Didden, and I finally have a break-thru.
In the end I didn't solve it with some kind of "variation of functions" calculus, I just did what a real mathematician would do and stared at it until the answer occurred to me.
Unfortunately I am not a real mathematician so it took a couple of years
The result is unexpected, the minimum can be reduced to ZERO.
It is a kind of error correction, at least mathematically.
That's like the discovery of perpetual motion, seems that it can't be true, but it looks OK so far, there are some stability questions I need to study.
Of course it is subject to non-linearity in the feedback networks (resistors and capacitors), so never truly zero.
But it looks that it should be possible to achieve very low distortion indeed.
Or perhaps more to your taste, and mine, very low distortion with a simple circuit, because it's so insensitive to circuit non-linearity.
Now to implement it and experiment.
Best wishes
David
Yes, those are typical worst case values, the problem is that I have no idea what the stray capacitance value of a power supply to earth is likely to be.
Obviously depends on the construction, layout, physical size of components etc.
I don't like the uncertainty.
Also I am not sure if the power supply stray capacitance can be isolated by the output network, need to see a circuit.
Nor I but the article did not impress me much, doesn't have much detail itself, not much analysis or much new, as far as I can see.
Well, it's a conference piece not a PhD so I shouldn't be too critical.
Yes, I am worried about this too, as I've already posted.
Best wishes
David
Last edited:
The supply is not earthed.
The amplifier is isolated.
It's only the requirement that all exposed conductive parts be connected to the protected chassis that makes the amplifier not totally isolated.
But prior to making this Safety connection the amplifier is "floating".
Last edited:
It is clear that the floated supply stray capacitance becomes part of the load, whatever that load is referenced to, which I called earth for convenience.
This was even in my 1st year university text, the rather brief Alley and Attwood, on p. 305 of the 1971 edition.
"the shunt capacitance between the power supply ground or chassis would also limit the hi-frequency response of the amplifier".
Please look this up, I don't even want to quibble about it.
Best wishes
David
It depends on what you define as the "load". A conventional description shows that is worth quibbling about...
Current flows through the stray capacitances in floating or non-floating power supply arrangements - however large or small they might be arranged to be. If we assume those capacitances are balanced between the two amplifier halves (the "n" and the "p" output devices) which is not far from realistic in my experience then...
With floating supplies an additional "common mode" current flows through the stray capacitances proportional to the amplifier voltage appearing across the loudspeaker. That voltage is formed from the "differential mode" current that is the difference between the currents flowing in the amplifier halves.
The amplifier load is still therefore that of the loudspeaker. The common mode stray capacitance does not load the amplifier output, rather it provides an additional load on the power supply. The high frequency output of the loudspeaker is not compromised.
Imagine for example a class-B output stage with a floating supply... When one half is switched off no current flows through the output device yet (nominally) identical currents flow through the stray capacitances from both amplifier halves to "ground".
There may well be a small difference between the stray capacitances apparent for each amplifier half. That difference will result in a differential current appearing in the load but this I would suggest is normally very small.
Also to note is that I have used this configuration for high power current drive applications that lack (what would presumably be) the benefit of a low impedance node at the power supply and I have never experienced a high frequency roll-off.
I cannot find this particular edition but the copy of Alley & Attwood I did find contains a section discussing shunt capacitance in single ended, small signal amplifiers with and without transformer coupling. The HF limitations discussed in this book are therefore out of context in the subject of this thread for the reasons I have stated in my last response and earlier in this thread.
Ok, the edition I have is hardcover, John Wiley and Sons, printed in the USA.
In case you have an edition with different layout, the quote is from
"Chapter 12 POWER AMPLIFIERS"
Section "12.4 COMPLEMENTARY-SYMMETRY AMPLIFIERS"
It refers to a picture - "12.15. A complementary symmetry push-pull amplifier"
It's a floated supply CE amp, exactly what we discuss in this thread so it's definitely in context.
The full quote is the second last para. on p305, starts with "The circuit...12.15 has a serious limitation".
Best wishes
David
Last edited:
As I said above there is a potential limitation. But with a loudspeaker load and a well-engineered amplifier that potential is evident only as a non-serious term in an equation - at least as far as the amplifier output is concerned.
It is also worth noting that the necessary provision of a separate power supply for each output stage can alleviate problems occurring elsewhere in real amplifiers, particularly when using class AB/B. I would suggest a better appraisal would show the floating supply arrangement to be advantageous even if a bit more expensive to implement...
And here to clear up the other point of confusion previously alluded to in this thread is a very basic diagram of the distortion nulling technique employed by ATC. The diagram has been adapted for simplicity here and is devoid of biasing components so please do not expect it to work as shown - it is shown to explain the concept only!
The source Vin represents the output of the low voltage amplifier gain stage(s). The supplies Vcc are assumed to be floating. Each half of the output stage has a net gain of approximately -R2/R1. The nulling variable feedback resistor 2.R2 then allows balancing of the net output stage gains - at dc at least.
At higher frequencies, balancing the RgCgs time constants by an additional variable cap has also been found advantageous but only in a particular non-switching application. In a conventional class AB amplifier I do not expect this to offer much advantage.
As said before, this arrangement is very simple and very effective. The open loop gains R2/R1 and that of the driver stage (a simple op-amp in the ATC version) can be optimised at the prototype stage.
View attachment OutputStage.pdf
The source Vin represents the output of the low voltage amplifier gain stage(s). The supplies Vcc are assumed to be floating. Each half of the output stage has a net gain of approximately -R2/R1. The nulling variable feedback resistor 2.R2 then allows balancing of the net output stage gains - at dc at least.
At higher frequencies, balancing the RgCgs time constants by an additional variable cap has also been found advantageous but only in a particular non-switching application. In a conventional class AB amplifier I do not expect this to offer much advantage.
As said before, this arrangement is very simple and very effective. The open loop gains R2/R1 and that of the driver stage (a simple op-amp in the ATC version) can be optimised at the prototype stage.
View attachment OutputStage.pdf
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.