A new kind of Class D amp ?

[ingrast]

Steven:

I have no experience with those pointed by you.

I did build a 60W class D amplifier almost 2 years ago as testbed for circuit and topologies lending to a PSU-less amplifier, got even to the point of designing a PCB but put the project in back burner for the time being, to pursue others.

If there is interest, I can post details of several design issues and results.

Rodolfo

[ingrast]

Oops...

Sorry, somehow I duplicated the post while navigating past pages in the browser.... edited it out to avoid confussion.

Rodolfo

[subwo1]

Originally posted by Steven
Some other interesting switching amplifier brands/topologies, that use output transformers witch subsequent switching:

http://www.jam-tech.com/main.shtml
http://www.nphysics.com/whitepaper-classnaudio.htm

And another self oscillating one with feedback:
http://www.powerphysics.com/

I have no experience with these amplifiers. Anyone?

Steven

The last paragraph, of 2nd site linked, mentions what I did when experimenting on the "class A version of a class D" amplifier several years ago. I used a 6N137 to feedback a reset signal to the power supply to keep it syncronized with the audio input.

[analogspiceman]

Originally posted by Steven
And another self oscillating one with feedback:
http://www.powerphysics.com/

I have no experience with these amplifiers. Anyone?

Steven,

Mark Smith of power physics was a graduate student who studied at the University of California (Irvine) under Keyue Ma Smedley. Mark Smith extended Dr. Smedley's earlier work on unipolar one-cycle control for power supplies to apply to class d audio amplifier applications.

While he was a grad student they produced quite a few papers together that were presented at various switched mode power conversion conferences. One can follow the evolution of one-cycle control by reading Smedley's first papers on the subject, then their co authored papers and finally Smith's white paper from the PowerPhysics web site. Many of their paper's are available as pdf's from Dr. Smedley's web page: http://www.eng.uci.edu/faculty/smedley/index.html

Although one-cycle control worked great "out of the box" for unipolar power supplies, the wide dynamic range and four quadrant nature of class d lead to challenging stability and distortion issues that were only solved three or four design iterations and published papers later.

Their present control scheme works by toggling on one mosfet of the output pair for a predetermined time and then toggling on the other for as long as necessary for the integral of the unfiltered output voltage to exactly equal the command voltage. Thus, the error will always be zero at the beginning and ending of each complete switching cycle, hence the descriptor "one-cycle control". This type of control scheme would normally lead to a widely varying operating frequency, but they apparently feed forward both the input and output voltages in order to dynamically adjust the length of the "fixed" portion of the cycle. This keeps the operating frequency nearly constant.

Their performance curves look to be quite good and are probably not just marketing hype (their academic publications are top drawer). Haven't done a detailed comparison, but their product seems to be in the same league with the UcD and B&O ICEpower modules. As you noted, it seems their design would also qualify as yet a third type of self oscillating control to go along with the hysteresis and phase shift types. (Hey, I just answered one of my own questions from another thread.)

Regards -- analog(spiceman)

[Jaka Racman]

Hi Analog,

I have some mixed feelings about one cycle control. I did some simulations some years back, and it seems to me that one cycle control has the same drawback as peak current mode control and that is instabilty at duty cycles larger than 50% in buck type of converters. I have also read early patent about one cycle control in Class D amplifiers and I have a gut feeling that it has no advantage over simple hysteretic self oscillating design. Constant frequency of oscillation can be done in hysteretic oscillators by modulating amplitude of the hysteresis band (already patented). How well can a simple hysteretic control work can be seen by looking at the graphs of IR 500W reference class D amplifier .

It is interesting how some of the most brilliant academic work newer finds it's way into the industry. Think of Dr. Slobodan Cuk's work for example.

Best regards,

Jaka Racman

[Steven]

Gentlemen,

Thanks a lot. Some more interesting papers to read.

Steven

[ingrast]

Gentlemen:

While very interesting and most probably excellent performing class D solutions, I have a feeling asynchronous or self-oscilating designs are rather a dead end concept.

I guess the trend to a fully digital signal chain up to the power stage will be consolidating, given the fact digital signal processing is predictable and inexpensive in comparison.

In concept I guess the reference design to emerge will have A/D conversion from the output, DSP in the feedback path with predictive adaptation - may be even neural network adaptation - and synchronous with the program source carrier.

I for myself am planning to work in this line next year if time and resources allow.

Rodolfo