Tue Jan 06, 2015 3:50 pm
As a Hauptwerk user who is also knows something about the mathematics of data compression (as a big part of my day job) I felt the need to chime in on this.
I had not heard of this Meridian MQA technique so I had to search online to find out. There are two problems - first, in mathematics, the acronym MQA usually stands for "modified quasilinearization algorithm" which is a technique used to pseudo-optimize control algorithms. It turns out the Meridian is NOT using this definition. Instead, their MQA stands for "Master Quality Authenticated!" Once I figured that out, I learned that because this is proprietary technology, I was not going to find out anything real about how it works.
However, I can say a few things ARE true. There are two major kinds of compression algorithms: lossless and lossy, though I hate the latter term. All algorithms are optimized to preserve the most important parts of the original signal relative to the application in which they are used. The Meridian MQA appears to be a lossless algorithm, which means that in the absence of noise and with infinite latency, it will reproduce the original signal to the accuracy of the sampler that is used in the system.
Having said that, there are lots of such lossless algorithms that have been optimized for audio signals - meaning that they completely reproduce the original signal under the same constraints. The main figures of merit are the compression ratio (how much smaller they make the signal in terms of bits) and latency (how much delay they introduce.) I can find no published results on the Meridian algorithm to compare. However, the compression ratio is limited by the entropy in the original signals (how random they are). For audio, this results in about a factor of ~two compression - regardless of the algorithm. Some algorithms trade these off, for example a very low-latency algorithm might achieve only 40% bit reduction.
The compressor used in HW certainly has reasonable latency, so I do not expect switching to a different one will make any significant change to the responsiveness of HW. Because all the various algorithms have about the same compression ratio (and because there is no magic here) I don't see this new one resulting in any significant change to the amount of RAM or disk space required for an organ.
The claim to fame for the Meridian algorithm seems to be something other than compression ratio or latency. They are claiming some sort of authentication mechanism that can trace the various operations performed on the file so that by the time you hear it, you can trust the intervening steps. This might have value in developing HW sample sets, I suppose.
Les