High bandwidth switching amplifier

[Snickers-is]

I am currently working on a circuit that seems to be capable of very high bandwidth compared to class D.

I have noticed that class D often starts to some bandwidth dependent distortion well inside the audible range. Most of them stop at 40-80kHz, and the typical result of this is that they also get phase distorsion from around 1-5k, distorsion in frequency response in the same area, rising output impedance (if not corrected heavily with feedback) and some instability problems at the highest frequencyes when being run with capacitive loads.

I think moving the bandwidth up with a factor of 10 or so could mostly solve all this. The design I am working on seems to be capable of that without creating more noise or loosing efficiency.

Does anyone know of any current switching amplifier design capable of this? There is no reason for me to put down a lot of work in something that other people have solved in a similar or a better way allready.

The goals for this are:
-Low output impedance in the entire audioband without heavy or non linear loop gain.
-Power bandwidth of at least 300kHz.
-RF noise similar to, or less than an ICE-power.
-Flat frequency response and phase.
-low impedance capacity to at least 40kHz without stability problems.

Thanks

[Pano]

The Tripath amps typically switch at about 1 MHz. That’s your factor of 10, right there. It seems to work well, so your are on the right path.

[Snickers-is]

Factor of 10 compared to what? Most typical class D designs switch at 300-600kHz, and the power bandwidth is mostly lower than 1/10 of the switching frequency.

What is the typical power bandwidth of a Tripath 1MHz?

[phase_accurate]

A power bandwidth of 300 kHz is indeed not very usual for switching amplifiers. It is not necessary either IMO but it could make a good sales argument.
Is your design a pure switching design or is it a linear/switching hybrid ?

Regards

Charles

[Snickers-is]

Since the real problems of class D amplifiers are related to the bandwidth limitations a higher bandwidth should be of great interest.

It is a pure switching design.

[Pano]

Originally posted by Snickers-is
Factor of 10 compared to what?

Compared to your 1st post of "Most of them stop at 40-80kHz"

There are some power bandwith charts on my site:
See this Page

That's for the small chips run at 12V with the typical filter. I have not tested the bigger chips.

[Snickers-is]

40-80kHz referred to the bandwidth limitations. 1MHz at the Tripath is, as far as I have understood, the switching frequency. I guess it is possible to push it to 100kHz bandwidth.

Interesting measurements you have there. Have you made some similar HF (20kHz) measurements on these amps?

[Snickers-is]

And by the way, do the tested amplifiers run full bridge so they can take advantage of a -12-0-+12V swing or are they limited to -6-0-+6V?

[Pano]

Originally posted by Snickers-is
40-80kHz referred to the bandwidth limitations.

Oh, OK. I thought you meant switching frequency, sorry.

Have you made some similar HF (20kHz) measurements on these amps?

Yes, I'll have to dig them up. Of course they don't look nealry as good, some of that is due to the nature of measuring class-d signals. But a lot of amps suffer in the top end, no matter what the topology.

The amps are bridged, (BTL). So they get more power.
Another possible advantage of BTL amps is that an output filter may not be needed. As the RF is the same polarity on each line, there should be no RF current flow. The speaker sees onlly the difference, which is AF.

However, with speaker wires over about 10 inches, you'll raditate a lot of RFI! Sheilded speaker cable, maybe? =)

[Snickers-is]

Do you really believe shielding is enough? Even with grounded shielding the ground network might become an enormous antenna.

Anyway, I believe you see the advantages of making the natural bandwidth higher?