If you look at the way the pixel array works to produce analog sound from a choreographed activation of a sequence of digital pulses moving from one extreme position to another you can see that a) it is important that every pixel element must be identical to every other pixel, b) must move at the same speed as everyone other pixel, c) must start and stop its travel in exactly the same time frame as every other pixel element.
For that to happen is the basis of the function. To ensure that capability every pixel element must be manufactured to be precisely identical to every other element, and etc. The problem is that in reality that is not always achievable within the controllable range of the term 'identical'. So there my be some tiny differences. Given the physics of DSR this will impede the performance.
Imagine a 1000 drummers standing in rank, with a conductor up front beating time with a big stick, and they are all attempting to play a rhythm in complete synchrony with his beat. Try as they like there will always be a few that are outside the strict timing. And you will hear them. The sound will be fuzzy. When you take a closer look at every drummer you will notice that some have longer or shorter arms than the others. And even when they are following the baton of the conductor perfectly they manage to hit the skin a fraction sooner or later than the others. The solution to this? Well we can't slice off pieces of the long arms or put extra slices into the shorter arms. So for the guys with longer arms you make their straps longer, so the drum is a centimetre or so lower. For the short arm guys their straps are a centimetre shorter. Sounds a lot better.
That is the equivalent of adjusting the electromechanical characteristics of the MEMS device. Maybe the correct strap length for every drummer is now written in a little book and can be controlled now by the software.
But maybe we could also replace the guy with the baton up the front with a control box with a heap of wires coming out, and each wire goes to a light hanging in front of each drummer. The light flashes on when the beat should start. But the controller is smart, it also has the ability to measure the delay time of each drummer's response to the light. And it can then adjust the signal start time for each drummer independantly, so that every drummer whacks the skin at exactly the right moment. And when all the drummers are marched in and stand ready in place it can call them to order and take them through their paces and measure and store all that information for every drummer in the band and use it to adjust the timing he follows. Dah-dah! Boom titty boom titty boom boom boom!
That is the proprietary software solution. Stored on every chip (or chip controller) at time of chip manufacture and initialisation. It presumably can also tell the odd pixel to sit down and put their sticks in their pocket because they just can't cut it.
"... Industry methodologies involved in mass producing semiconductors indicate that over time such variances can be drastically reduced if not eliminated, nonetheless the company decided to explore other possible means of negating this problem.
The innovation ultimately conceived involves expanding the electromechanical characteristics of the devices, that when combined with a proprietary software-only approach provide the capability of rendering any such noise inaudible. This solution which has been tried and tested on our prototyping platforms (MEMS GEN 1) has proven to offer the sought improvements to the clarity of sound even though it cannot be fully utilized without the electromechanical changes incorporated in the mass-production version of the chip.
..."
Do I know this is true? No. But it is so obvious that I read it as being implicit from every other detail I have read in the announcements and patents. So I am guessing these guys know a thing or two about their own specialty field, and that they have got way beyond that point long ago.
(20min delay)