Current state of the art Class D?

[kipman725]

Originally Posted by Dave Zan

.
What I did eventually settle on is to use a hybrid, a Class H amp with Bode Maximum Feedback type control loop but switch mode controlled rails for efficiency.
Despite all the discussion of SOTA Class D, the hybrid scheme seemed a better balance.

For highest fidelity this does seem to be the best approach as the complications of output filtering is avoided. I'm using Yamaha P series amps which also have voltage variable rails on a bipolar AB? amp. Infact the best performing amp I know of also uses this structure:
Review and Measurements of Benchmark AHB2 Amp | Audio Science Review (ASR) Forum

[Dave Zan]

Originally Posted by Mello4

Which already published patents? European?

Yes, European Patent Office, it is in the header information, section (30) for priority information shows an EP number and prior date.
There is also Patent Cooperation Treaty information in the header, I haven't kept up with this since I left the Patent office but I believe it just shows the use of the PCT to simplify paperwork for applications in other countries.

Best wishes
David

[Dave Zan]

Originally Posted by kipman725

...this does seem to be the best approach..
amp I know of also uses this structure:
Review and Measurements of Benchmark AHB2 Amp | Audio Science Review (ASR) Forum

The Benchmark is indeed impressive but a little different from my idea I think.
It has a switch mode power supply but then a conventional linear modulation of the Class H rails.
I propose switch mode control of the Class H power rails, more efficient but some potential for the switch noise to sneak into the output.
But I have the benefit of the Power Supply Rejection Ratio of the linear amp to reduce it.
So I'm maybe 60 dB ahead of class D that typically has no PSRR at all and still works remarkably well.

Best wishes
David

[kipman725]

The marketing info seems to say its rails are modulated by the SMPS:
The AHB2 - A Radical Approach to Audio Power Amplification - Benchmark Media Systems
it wouldn't have an efficiency advantage over conventional AB if the SMPS generated fixed rail voltages. For your SMPS are you using current mode control?

[Jazzfan]

I've had a Benchmark DAC 2 HGC and AHB2 amp for six years and they are brilliant.

[Dave Zan]

Originally Posted by kipman725

...info seems to say its rails are modulated by the SMPS:...
it wouldn't have an efficiency advantage over conventional AB...

I think you misinterpret the Benchmark brochure, it says
"For this reason, there are very few amplifiers that use multiple supply rails. ...In contrast, the feed-forward error correction used in the AHB2 is so effective that multiple rails can be added"

So it's a multiple rail Class H and that's where the efficiency improvement comes from over a Class AB.
If it was switch mode controlled rails, as I propose, they wouldn't need or use multiple rails.
Those multiple rails are driven by a SMPS but that is a separate issue.
This stuff is less than obvious because there is some inconsistency of nomenclature on Class H (rails that track vs switched rails) and some brochures less than clear.
But the main surprise was that initially I expected amps to work just as you propose, with SMPS modulated rails but I checked schematics and haven't found any yet.
I was uncertain about Benchmark at first but I believe my current explanation is correct.

For your SMPS...current mode control?

My rails are controlled similarly to Bruno Putzeys' Class D output.
Phase shift in the filter combined with natural delay in the comparator makes the system self oscillate.
This is independent of the power supply, switch mode or not.
Does that answer the question?
Possibly I won't even use an SMPS.
I have 3 phase power to my house so I could use a 3 phase power transformer and not require any "bulk" capacitors at all.
Haven't decided on this yet.

Best wishes
David

[kipman725]

ah I see yes I agree it does sound more like a switched rail approach, easier to design the power supply.

In my previous attempts at envelope tracking supplies I have used poly phase buck converters with an inner current loop and an outer voltage loop.

[Dave Zan]

Originally Posted by kipman725

...In my previous attempts...I have used poly phase buck converters with an inner current loop and an outer volt...

Nice, my rail trackers are essentially buck converters and it did occur to me that poly phase would reduce ripple and noise.
But simulations so far look pretty clean with just 1 phase so I hope to avoid the extra complexity.
Were yours for audio or some other application with more difficult requirements?
I must admit that I still don't have a satisfactory feel for best loop control in this case.
I understand control theory reasonably well and have some results about how nested loops can be treated as an equivalent simple loop.
But I don't yet see how it plays out in the situation you describe.

Best wishes
David

[kipman725]

In simulation I have done up to MHz envelopes for LTE (poly phase EPC 8000 series devices) , project didn't get funded beyond simulation. One thing I'm good at is getting high speed switching and instrumentation of said high frequency switching, usually I'm able to get faster than datasheet turn on/off times out of GaN devices. When it gets really fast you switch to resonant sine wave gate drive and take the conduction loss hit, I think this project was hard switched though.

Single phase should be fine for audio until you get to high currents and high step down ratios. However almost all the time you can get better performance by going to polyphase and a smaller solution size. Perhaps I'm missing something but I struggle to find off the shelf controllers so have had to make my own for a few projects and this instantly makes the project not commercially viable in size and cost (I just use the excuse that in production we will just use an ASIC).