The technology allows for both situations... One paper showed 8 voice coils located in one 10 inch speaker and a previous paper showed 8 separate piezos. The bottom line is that their shaping to compensate inadequacies of the speaker/voice coil mismatch is done using shaping filters, shoving all of the noise into higher frequency realms. Their output spectrum showed a higher background noise at lower frequencies. I'm not sure what kind of end user effects this would have. It would be interesting to compare waveforms between a piano and piano driven through these different speaker configurations to some different analog configurations.
On first reading this article, I was thinking that they were talking about a multi-voice coil speaker where each voice coil was a separate piston, driving a different physical sector of a single speaker cone area. In fact, the conference paper reveals the design to be simply a standard (single piston) speaker voice coil structure broken up on the same coil-form into equal chunks of fewer turns, thus lower impedance for each separately wired section.
Since the class-D principle (DAC-less conversion) is well established art, perhaps Intels interest in this approach is the multi-coil speaker itself as a means of developing high audio power directly from logic-level Vcc. This would alleviate having to run a power amp at battery voltage (12V or more). Driving eight small 0.5-ohm coils with 1V each could produce the same magnetic motive force into the speaker cone as a single 4-ohm coil being driven at 8V. This effectively could be considered direct digital drive, which Intel could build into a chipset that doesn't have to see more than 1.5V supply voltage and yet produce almost 2W of speaker drive power.
Agreed Brad - these are conference papers and not subject to the level of peer-review of journal ones. Their first one shows CMOS buffers driving voice coils directly which would be less efficient than even a class B 'analog' amplifier.
I can see no engineering justification whatsoever for what they're up to here.
Ah, well, if so, it's a derivation of the multiple speaker drivers.
Multiple voice coils in a single speaker make for a more complicated speaker design, where the transducer has to oscillate efficiently over a wider frequency range.
Or, as has also been done in the past many times, you can have "coaxial" speakers, in which multiple coils AND multiple transducers are mounted on one axis. So it more or less looks like "one speaker."
bert, I believe it is multiple class D outputs driving a plurality of voice coils in the same driver. The approach to reducing power consumption is fairly valid, but it's imo about the only advantage. And there are other ways to pretty much the same end. The manufacturing of multiple voice coil transducers is highly nontrivial as well.
On this same general topic, some time ago I saw a kind of cool scheme for producing low bass notes without having to have very large drivers. The idea consisted of driving small, high frequency drivers at ultrasonic levels, and then relying on the beat frequencies between the drivers to create the audio signal.
Probably plays havoc on any nearby dogs. Anyway, I was thinking this new scheme had something along those lines. Would be a cool way to get big sound out of small devices.
I'm not sure if I'm reading this correctly, through the hype. My take is, Trigence has created a scheme whereby Class D amps are driving speaker systems consisting of multiple drivers? What traditional Class A, B, or AB amps do, through traditional crossover networks, when driving speaker systems.
My speakers at home consist each of 5 drivers, from woofer to piezoelectric tweeter. But it's likely that the crossover network won't work correctly with just any Class D amp - I wouldn't be surprised.
So if you have separate Class D amps, one for each driver, then you can even optimize each of the amps. For instance, the woofer-driving amp could get by with a much lower pulse rate than the piezo-driving amp. But the woofer-driving amp might require higher voltage pulses.
And sure, if there isn't much energy in a particular frequency range, then the amp serving that frequency range would put out little or no energy. Exactly the same happens with crossover networks, after all.
Using multiple H-bridges to drive multiple voice coils is cool and even novel, but I agree it's not "direct digital drive" -- at least not any more so than any other Class D amplifer. The direct digital drive verbiage is pure marketing hype.
I had not heard the Bruno Putzeys reference before, but I like that -- the symbol domain. Another way to say it would be the numeric domain. But as you said, timing and voltages and also currents are crucial. We're not just talking about binary numbers magically turning into acoustic pressure here.
From the AES paper, it sounds like their true IP is in the ML-NSDEM, which can choose to scale back the number of voice coils being driven, based on audio signal amplitude.
That really is pretty innovative, so there's no need to over-hype it and make it sound like something other than an improvement on the traditional Class D concept.
Identifying the outputs of switches as being "digital" ("not building anything analog") is a misnomer and an example of the prevalent conflation of "switching" with "digital". You are, emphatically, already NOT in the symbol domain (as Bruno Putzeys calls it) ---timing and voltages are crucial.
But shifting the burden of "conversion" to multiple voice coils is novel, if not necessarily practical. Does the company really think such transducers will be easy and economical to make, and that the resultant system will be cheaper and/or higher fidelity? Pardon my skepticism here, but people have enough trouble making single voice-coil transducers with a single pair of flexible leads. And what about all the silicon needed to produce the separate drive signals?
Sounds like a marketing gimmick to me.