Error forward correction for Class D

[hahfran]

About 4 companies sell both half-and full bridge class D modules
from 100 watts to 1000 watts at reasonable prices for ex. a
1000 watt full bridge configurable as 2 250 watt half bridge comes
for 300 Euros, of course without power supply.
However these devices cannot compete with analog HiFi amps.
Based on available literature on error feedforward correction I figured out it is possible to have a close error nulling with a particular realization of error feedforward , the current dumping solution.
This error correction method is widely in use in radio frequency power amplifiers etc. where negative feedback is not applicable.

Such error nulled class D are a perfect solution for active shelf speaker boxes where analog amps take up too much "acoustical" volume.

[rtarbell]

Hello hahfran,

In regards to error feedforward and the current dumping principle for class D, look up "SMALA" (switch mode assisted linear amplifier). I *think* the SMALA concept is based on the error feedforward correction, and today there are several IEEE papers on the topic (the "Class K" amplifier, for example).

[bwaslo]

However these devices cannot compete with analog HiFi amp

I think you will find quite a few people that will disagree with that statement, also.

[hahfran]

Originally posted by rtarbell
Hello hahfran,

In regards to error feedforward and the current dumping principle for class D, look up "SMALA" (switch mode assisted linear amplifier). I *think* the SMALA concept is based on the error feedforward correction, and today there are several IEEE papers on the topic (the "Class K" amplifier, for example).

hello,
thanks interesting but no SMALA is apparently not an error feedforward correction design. I looked a bit deeper in the theory.
The class A amp is way above my restricted DIY means if the principle of error nulling ( and thus omitting the class D LC output filter ) is to be achieved: ultra-fast , very high open loop gain, very high gain bandwidth product around 2-5 watts up to 10 MHz...
( for 250 watts class D output ) and of course vanishing distortion in this range of operation.
someone with lab equipment experienced in PCB high frequency layout can certainly accomplish such an class A amp but I cannot
that would be wasting money and time.
Nevertheless it were interesting because of theoretical advantages
at least 80% efficiency at full power, zero distortion up to fifth harmonic no bulky radio wave emitting output filter...

[jcarr]

hahfran: Over the years, I have been quite interested in the Sansui and Denon approaches to FF-EC, and I was also considering a similar topology some time ago (Class D + Class A/AB feedforward error correction). I talked to another EE friend about this, and he mentioned that Yamaha had put some effort into this area, and that they had a patent on the technique. I think that in Japan it was called "Parallel Hybrid" or something like that.

I believe that you will need some kind of combining network between the primary output stage and the error-corrector (even if the primary stage operates in class AB or B). Therefore rather than attempting to banish the class D LC output filter entirely, it may be more efficient to work on optimal integration of the combining network and LC output filter.

BTW, the error corrector in the Sansui (for a power amp rated at 130W into 8ohms) consisted of a 2SK213/2SJ76 pair, and these are lateral MOSFETs in a TO-220 package. They have good frequency extension, but not so much power-handling capability. IIRC, the error corrector had a voltage gain of 1.

In any case, I agree that this is a worthwhile area to pursue.

regards and hth, jonathan carr

[rtarbell]

"I think that in Japan it was called "Parallel Hybrid" or something like that."

--> FYI, I did a quick search on "Parallel Hybrid" and Amplifier, and I found this paper from IEEE:

"Modeling of a parallel hybrid power audio amplifier"

by

Lorenzo Chiesi, Emilio Lorenzani, Giovanni Franceschini
Alberto Bellini

Just as an observation, in most of these designs I see, the output capacitor of the Class D amplifier is omitted (I think the idea is that the class AB amplifier is supposed to "take over" the filtering duties that the capacitor normally does).

[Steven]

Originally posted by hahfran
...
Based on available literature on error feedforward correction I figured out it is possible to have a close error nulling with a particular realization of error feedforward , the current dumping solution.
...

See "HIGH EFFICIENCY AUDIO POWER AMPLIFIERS, design and practical use" (http://www.ub.utwente.nl/webdocs/el/1/t000000d.pdf) from Ronan van der Zee, esp. section 4.3.2, but the whole book is interesting.

Steven

[jcarr]

Steven: A very informative read. Thank you!

regards, jonathan carr

[darkfenriz]

I don't know about any commercial or DIY class D error correction amp.
Vanderkoy and Lipshitz mentioned experimentig with current dumping in the original quad405 article

[hahfran]

Originally posted by Steven


See "HIGH EFFICIENCY AUDIO POWER AMPLIFIERS, design and practical use" (http://www.ub.utwente.nl/webdocs/el/1/t000000d.pdf) from Ronan van der Zee, esp. section 4.3.2, but the whole book is interesting.

Steven

Thank you yes this an error feed forward current dumping design.
Imo given the average DIY audio amateur a class D amp is in terms of PCB design above the means especially if switching frequency is in the range of 500 kHz.
However recently I've seen a special offer a class D amp 60 watts in 8 ohms size of a cigarette box for 99 Euros. No data given except minimum load is 6 Ohms. That were interesting to use this device
in a current dumping configuration because the requirements for the class A are moderate.