Indeed. One would expect that you would have to make additional changes beyond simply moving the feedback take-off point.
would be surprisingly tricky if the class-D amplifier has pre-filter feedback. A matching filter is still required even if you use post-filter feedback (assuming the class-D output filter has a break frequency somewhere in the region 30 - 60 kHz).
Those interested with access to the AES e-library may wish to read this:
AES E-Library >> Switching/Linear Hybrid Audio Power Amplifiers for Domestic Applications, Part 2: The Class-B+D Amplifier
Yes.
Beat me to it!
Harry kindly sent the copy of his AES paper to me, and it helped me greatly to understand what problem I will encounter during this adventure.
Shame on me for not reading the whole thread more carefully and thank you for Your patience. There were answers to all of my questions in the thread but "Monja" in post #42 wrote that JLH analysed a very similar commercial design from Technics in 1989 and found a problem. Unfortunately the link to that particular article is dead.
Maybe You could do a quick build without tinkering with the feedback of the Class-D amplifier. I can't be the only member who wants to know how or if it worked.
This is based on my assumption that tinkering with the Class-D amplifier is only necessary if You want many watts at high frequencies.
The problem JLH found was: "However, there is an inherent problem,
which is that the load is connected
to the OV line, but the floating power
supply is not. Since this is only returned
to this line through the class B amplifier,
it follows that this latter amplifier is in
series with the load at all times."
This will not be a problem if the class B amplifier is instead a class-D amplifier as this thread is suggesting.
In the article the class B amplifier is unity-gain and its input is taken from the output of the class-A amplifier. High-power unit-gain non-inverting amplifiers are scarce.
Maybe it would be easier if it was an high-power, unity-gain, inverting amplifier but then you would need a preceding inverting stage and then you can just as well take the input from the input to the class-A amplifier and let the class-B amplifier have the same or somewhat higher gain than the class-A amplifier.
Again you interpret it in wrong way. JLH did not talk about using Class D, only Class B and he is talking about Class B.
How you've found that there will not be problem if Class B was used instead Class D???? JLH did not suggest that!
My english is not perfect but this time I think You misunderstood me. I never meant to imply that JLH suggested a Class-D amplifier.
I wrote "This will not be a problem if the class B amplifier is instead a class-D amplifier as this thread is suggesting." I should have phrased it differently. I meant if a Class-D amplifer is used instead of a Class-B amplifier. With thread I meant the thread we are writing in. What else it is called?
Interesting stuff! But will it work for Class-A? Both use PWM to make the Vcc+ and Vcc- track the output signal but on from from 0 and up for the positive rail and from 0 and down for the negative rail.
That EEEngine is actually class-AB+(Class-B•D) - the tracking part that supplies power to the main class-AB output stage is made of a parallel combination of a linear (class-B) and switching (class-D) part.
Indeed. Both will still have high linear-stage power losses for non-resistive loading (i.e. output current and voltage out-of-phase with each other) - like you get with a speaker!
Dadod's proposed method will allow the "+ve" supply rail of the linear output stage to go negative with respect to output ground, and visa-versa for the negative rail. In other words, assuming perfect matching of his linear and switching part gains, and no ripple in the floating low-voltage supply, the Vce of the linear-stage output transistors will be constant under all conditions, i.e. regardless of relative phase of output current and voltage.
Vce is not constant in a regular class-A amp but constant Vce is should be an improvement over a regular class-A.
Over lunch I came to the conclusion that the Labgrupp TD solution could be made to work for class-A operation., at least in theory.
What struck me was that the upper PWM can be regarded as a voltage controlled current source and the lower PWM as a voltage controlled current sink. The upper will never have to sink current and the lower will never have to source current. In that the requirements for the two are very different from what a class-D amplifier has to do.
I have absolutely no idea of how this would workout in real life with real life loads.
Might be interesting to think of the problem differently. Rather than switch the supply rails, you can correct for the shortcomings of the class D with error correction using a small, wide bandwidth class A amplifier. Since the error between the input Vi1 and the feedback voltage Vi2 is very small within the bandwidth of the system, you only need a small error correcting current through Rsum to remove any distortion.
You may argue that this is not a class A system, but I think you could make the case that if A3 is class A, the system could be classed as a class A amplifier with a class D simply providing the high power 'heft'.
High current back to back diodes across Rsum would be required in a practical solution so that during start-up or potential failure modes, the output of A3 is not dragged beyond its rails. By using Rsum rather than trying to sink the speaker current into and out of A3, you can keep A3 very low power. however, you would need to account for the class D output current x Rsum voltage drop so it would have to 'ride' the main signal current across Rsum and then add or subtract the error voltage from this.
Anyway, fun project for a rainy Sunday afternoon
You may argue that this is not a class A system, but I think you could make the case that if A3 is class A, the system could be classed as a class A amplifier with a class D simply providing the high power 'heft'.
High current back to back diodes across Rsum would be required in a practical solution so that during start-up or potential failure modes, the output of A3 is not dragged beyond its rails. By using Rsum rather than trying to sink the speaker current into and out of A3, you can keep A3 very low power. however, you would need to account for the class D output current x Rsum voltage drop so it would have to 'ride' the main signal current across Rsum and then add or subtract the error voltage from this.
Anyway, fun project for a rainy Sunday afternoon
Last edited:
Hello Dadod, I would also recommend reviewing the Swedish company Lab.Gruppen Class TD® amplifiers.
Here are some service schematics:
Free Labgruppen Diagrams, Schematics, Service Manuals :: Schematics Unlimited
And their patent USPTO #5,200,711
US Patent Full-Text Database Number Search
From their service schematics fP2200 they mention the use of flyback converts, a bit different topology.
Good luck with an interesting project!
Here are some service schematics:
Free Labgruppen Diagrams, Schematics, Service Manuals :: Schematics Unlimited
And their patent USPTO #5,200,711
US Patent Full-Text Database Number Search
From their service schematics fP2200 they mention the use of flyback converts, a bit different topology.
Good luck with an interesting project!