new class D modulation method?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Though most of the discussions moved to self-oscillating amps at the moment, I had an idea how to improve clocked amps. The common topic in many discussion was the shoot-through vs. deadtime problem. After some thinking, I found a way around this. I'll try to explain, and I also attached a picture.

The idea is to have 2 phase-coherent triangle waves, shifted by almost their amplitude (they may slightly overlap). After the comparator, this generates pulses, which never overlap, so shoot-through can't occur. Since the triangles may overlap, the deadtime problem related zero crossing problems don't occur that badly. The triangles can be carefully adjusted, to get the best possible zero crossing behaviour, and best THD, without any risk of burning the power FETs.

I already built the modulator, using a standard function generator (very nice, amplitude and frequency can be adjusted separately, and it's more stable than opamp generators). The triangle wave is decoupled with 2 capacitors and following potentiometer-resistor bridges, which allow to "shift" the two separate triangles up and down by a DC offset. The output of this and the comparator look very nice, but the power stage is not finished yet, so I don't know what the final result will be.

My question is: has anyone done this already? are there any nasty distortions coming from the phase shift at the zero crossing? How does the output behave, when the triangles overlap significantly? In this case, for small signals, we would get alternating positive and negative pulses, which would partly compensate each other.
Does it just go into some "kind-of-class-AB" mode, or does the slope of (let's say) a sine steeper or flatter than it should in that overlapped region?
 

Attachments

  • classd_modulation.gif
    classd_modulation.gif
    3.9 KB · Views: 583
wow, that was quick! Thanks for the answer, at least I know now what buzzword to look for. Is there anything special to know, when using this modulation? Some traps everyone runs into, or something? I had a long look around in this forum, but couldn't find anyone using this in a diy project. That made me a bit suspicious - either noone had the idea yet (unlikely), I didn't look hard enough (possible), or there is something utterly wrong with it (can't see what, but that's why I ask)
 
Hi,

John, I suppose you were thinking about Crown and their BCA amplifiers, but I will check Peavey's patents.

Since SFX has not proposed how the power stage would look like let me make some speculation. If he thinks of using classical half bridge and drive only one fet at the time (as suggested to eliminate shot-trough) then he is actually using two buck converters sharing the same inductor. In this case he is using opposite fet's diode as frewheeling diode. This is not a good choice since those diodes are rather slow. Another solution is Crown's BCA power stage (two discrete buck converters). In this case there is a difference in modulation scheme compared to the original one and I think that the patent is not applicable.

SFX, this is a very nice problem for Spice simulation. I think you could have your answers in an hour. Maybe it would be good if you would explain also proposed power stage topology.

Best regards,

Jaka Racman
 
Hi Jaka,

I'm pretty sure it was Peavey- Brian Attwood and a Peavey Designer. Wish I could recall more details.

Anyway, how’s your project going? I'm just finally completing the latest version of my Digital amplifier (hopefully this will become a product). Had to use PECL logic for the Clock and "Reference PWM" circuit - as the performance of the last design was limited by the Clock and CMOS logic used to generate the Ref PWM signal for my error correction stage.

I'm also RELACTATLY thinking about starting a Hi fi Brand using my technology, the last company I consulted for felt my designs where to high performance for there target market - seems I can never win!

If I do (start a brand), I'm thinking of selling directly thought the internet - although I still have to craft a business plain - I'm just investigating / exploring options at the moment.

John
 
Hi John,

I did quick search on Peavey, but the only patent I found about phase modulation was already discussed before .

Thanks for asking. I have prototype PCBs ready, but have not yet started soldering. One reason is that TI discontinued DIR1703 receiver, another is that I have a similar job project right now which is taking almost all my energy (800V 1200A 12 phase buck regulator) and the third is UcD. I am currently waiting to order UcD 400 modules to have a reference. After that I will see if there is anything left to us ordinary folks to do. In a way, I see difference between UcD and PowerDAC as a difference between Russian and American space technology. In my younger days I had access to both western and russian military electronics. Me and my collegues of course ridiculed russian technology, but now in my older days I am starting to see elegance of their seemingly crude designs. It takes a great mind to accomplish given design task with a minimum amount of high tech equipment and materials.
Sorry for digression and thank you for your kind words of support in a heated debate a few months ago.

Best regards,

Jaka Racman
 
Class D modulation scheme

I built a 60W admplifier 2 years back with this same modulation idea (thought then it was original :bawling: ).

The only problem - that is evident in simulations - is the inherent nonlinearity caused by the PS pumping through the flywheel diodes in the output mosfets. This does not happen with conventional modulators for there is allways a conducting mosfet.

I corrected this using a high pass LC network in parallel with the regular low pass feeding the load. Added benefit is you dump harmonic energy contents (you are generating a square wave at the carrier frequency but end up with audio, that wasted energy must go somewhere, usually back to the power supply) to a resistor (about 20 ohms is fairly good) instead of forward diode loss.

The prototype worked fairly well and sounded good. I left it aside for some time but will come back to the project for high power H bridge woofer/subwoofer amplifiers.

Interestingly, though I started with a high speed, high performance opamp for the modulator, I later settled on a simple RC integrator. Strong feedback takes care of linearizing things.

Rodolfo
 
Jaka Racman said:

Since SFX has not proposed how the power stage would look like let me make some speculation. [...]

Sorry, I should've explained - you're right, I was going to use a standard bridge with one inductor. Actually, I'm planning to use an H-bridge, driving the diagonally opposite MOSFETs with the same signal from the modulator. In my pictures, the gray rectangles above the centerline drive the high side of the bridge, the ones below the line drive the low side of the bridge (or for H-Bridge bridge, upper rectangles left-high & right low, lower rect. left-low and right-high).

I use high power Schottkys for the freewheeling diodes, and probably combine them with an RC highpass to save them from nasty spikes. I already built a simple BCA amplifier (see here).
It worked pretty well, but after a while, I think one of the freewheeling diodes burnt (they were terribly underrated, but I couldn't get any better ones at the time), and took the rest of the power stage with it :bawling: . I thought I might try something new as a replacement, but maybe I'll just built up the BCA again. Just thought it'd be nice to need only 2 instead of 4 inductors...

Cheers,

Felix
 
Hi Felix,

I see two possible problems with your approach. First, there is added mosfet capacitance and inductance (in effect a resonant tank) to your schottkys, so switching waveforms might be less clean than in BCA. Second, at light modulation filter inductors will be running in discontinious conduction mode. Vin/Vout vs duty cycle D transfer function becomes severely nonlinear in that case, beeing also dependant on some factor K=2L/Rload*fswitching. If you are interested you might take a look in the following article: Slobodan Cuk and R.D. Middlebrook "A General Unified Approach to modelling Switching DC-to-DC Converters in Discontinious Conduction Mode" IEEE Pesc 1977.
Your modulation approach is also possible with one ramp and shifted treshold of comparators. Look at old Silicon General SG1731 data sheet.

Best regards,

Jaka Racman
 
Well I braved searching though old papers and found these possibly interesting patents, some are directly relevant, others just food for thought

4,992,751 Audio Amplifier with Phase Modulated Pulse Width Modulation
4,600,891 Digital Audio Amplifier Having a High Power Output Level & Low Distortion
4,479,175 Phase Modulated Switch mode Power Amplifier & Waveform Generator
5,657,219 Opposed Current Power Converter (Crown BCA)

Jaka,

My uncle was a quite famous political dissident in the CZ, this made life very difficult for the family – which meant my mother (we) was not able to return to visit CZ even when her farther died. As a child I grew up in the West, but when we did visit the “East”, even at my young age I was struck how creative the engineers in the East where. When we visited CZ, I also remember my farther smuggling across the boarder well hidden box’s of “High Frequency” power transistor (in the ties of our Austin 1100) which where used to construct “Ultra sound” machines for the main hospital in Bruno (which until they built there own, hadn’t even seen / heard of an ultra sound machine before). I can still remember almost 30 years on, the Doctor / professor almost having a heart attack when my farther presented them to him.

John
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.