Audio Processor in the IMB Server

[Mark Gulbrandsen]

How often do amplifiers actually fail? Sure, in ATMOS installations, there is a higher statistical rate. At the same time, in ATMOS installations, there are easy means to mask it.

I am thinking for a good while now about easy ways to detect individual speaker failures in classic parallel surround setups. Easy in the sense that it allows them to be checked often (enough).

I think the statistical rate in Atmos is higher just because there are more amplifiers. It certainly can't be because the mixes are too dynamic. The 13 movies I heard in Atmos were pretty average

[Marcel Birgelen]

I am thinking for a good while now about easy ways to detect individual speaker failures in classic parallel surround setups. Easy in the sense that it allows them to be checked often (enough).

I'd say that if you perform a per-channel analys on a sine-sweep on a scope with a fully functional, calibrated system from a fixed point in the room, it should be relatively easy to detect if one or more speakers have failed, simply by comparing the amplitudes to the original. Detecting which speaker has failed is a bit more challenging. Maybe with a mobile rig with a shotgun microphone and a bunch of good isolating headphones?

Back when I was building radio stations, we'd use the older equipment to build a parallel backup station. We did a system in Los Angeles where an FM station had two transmitters and two antennas with no antenna switch. On changeover to the standby transmitter, the standby would be brought up at 50% power, then the main dropped. The standby would then be brought up to 100% power. To switch back, the reverse was done. There were separate audio processors and studio to transmitter links for each transmitter. The only thing in common was the tower. Some stations have auxiliary sites so not even the tower is in common.

I guess as long as the sources of the transmitters are close together, you can do a mostly issue-free transition between both transmitters. If they're geographically apart, you'll create all kinds of uncontrollable interference patterns when both are broadcasting at the same frequency.

On amplifier output relays, I'm thinking now that class D amplifiers that rely on pulse duration modulation would not require such a relay. It would be very easy to not enable the output transistors until the PDM circuitry is up and running.

I don't think I've seen a lot class D amplifiers in cinemas lately, most of the newer gear is still class AB or G/H, which is essentially an improved class AB.

Class D amplifiers have kind of a history of not being very good at their job, mostly due to sub-par component design. I consider them to be of the Niquist-cheat-design (just like any DAC ) and I guess a lot of their quality depends on the quality of the low-pass filter, which seemed to be difficult to get right for high power applications.

That being said, I've listened to some recent Class D amps which sounded absolutely fine. I guess we'll see more Class D amps in cinema, once we start seeing active speakers making their inroads, as the increased efficiency is much more key there.

I guess the PDM circuits don't require any kind of "stabilization", but I'm not sure about the low-pass filter, which will probably still contain some kind of capacitors running at considerable power.

Speaking of pulse duration modulation, I heard of a very clever idea a few years ago. A typical pulse duration modulator generates a single pulse of varying width at a certain frequency, either through analog techniques (triangle or sawtooth wave driving an analog comparator) or digital techniques (a digital counter driving a digital comparator). The interesting technique is to use a random number generator instead of a counter. The random number generator drives a digital comparator that provides an output if the random number is below a threshold corresponding to the instantaneous audio sample. As the sample rises, the random number is below the sample more often. The result is a high frequency series of pulses whose DC average corresponds to the audio sample. But the ripple frequency is much higher than with the other method, so lower cost filtering can be used. I guess, though, that as the sample voltage increases, the output pulse frequency decreases as more of the random pulses combine to longer pulses.

Sounds indeed like a clever design, but I see a problem here though. The design depends on reliable random numbers with a good entropy and those are hard to get on most embedded platforms, since it's hard to get good entropy sources.

I think the statistical rate in Atmos is higher just because there are more amplifiers. It certainly can't be because the mixes are too dynamic. The 13 movies I heard in Atmos were pretty average

I'm pretty sure it's a simple numbers game. But the impact of a single amp failing in an Atmos system is also less than that of a classical 5.1 or 7.1 system, unless it hits one of the primary stage speakers off-course...

[Steve Guttag]

I think I can tie a couple of parts of this together. QSC's DPA series of amplifiers (and CX-Q) are Class-D amplifier. Without a doubt the DPA 4.3 and 4.5 series had a MUCH MUCH MUCH higher failure rate than traditional linear amplifiers (at least from QSC). The DPA 4.2, while higher than a DCA failure rate, was nowhere near the rate of the bigger 4.3 and 4.5. The next generation of QSC's Class-D designs happened on the 8-channel variants, the DPA 8.4 and DPA8.8. Those have not proven to be problematic. The fixes applied to those are what is in all of the current DPA xKx series (i.e. DPA 4K4 or DPA 4K8). It is probably a discussion worthy of its own topic to discuss the overall sound quality of Class-D versus traditional Class A/B and Class H not to mention each design's strengths and weaknesses.

At the moment, I don't have any more apprehension putting in a QSC Class-D than their linear amps, with respect to unit failure.

Class-D is definitely where future amplifier development is going. Dolby's DMA is class-D and a lot of amps I see targeting A/V industries are Class-D.

[Marcel Birgelen]

Looking a bit further and maybe in-line with this topic of where stuff may be heading: With digital getting ever closer to the edge, I even expect that the "traditional amplifier" can essentially be eliminated. The "Class D design" can be modified in such way that the DAC is already included in the design and with a high-power output, can then drive the drivers in the speakers directly. You could actually tune the DAC/AMP for every driver inside the speaker. Crossover and compensation for every driver can be done digitally, right inside the DSP in front of the DAC.