A I was thinking about why there is so much confusion about what MQA is, s much conflating of various technologies, and I think there is an easier way to describe it. I think I can point to well established parallels in another field.
MQA has three components.
A. Correct various errors in the original analog to digital conversion.
B. Correct various errors in the endpoint digital to analog conversion.
C. Reduce bandwidth requirements of a high resolution content.
These are really independent and could be applied independently. Their ultimate goal is similar, so they are discussed together.
A. Correcting for errors is commonplace in many areas. Not difficult to understand, and should not be controversial. When we do room correction, it is common to get a calibrated microphone (like the $75 MiniDSP UMIK-1); these are not perfectly flat because that is difficult and expensive, but the manufacturer measures them and provides a calibration file which allows the RC software to correct for the errors.
Another example: in photography, lenses introduce various errors, like geometric distortion (straight lines appear curved) or chromatic abberation (purple or green fringes). Eliminating them is costly and makes the lenses large and heavy. But software like Adobe Photoshop and Lightroom include correction for these errors: unlike our field, digital photographs have rich metadata including the lens and camera model, focal length for zoom lenses, aperture etc. and Photoshop has lens profiles that permits automatic correction. Very valuable: in product reviews you often read language like “the lens has CA at large aperture but this is of course easily corrected.” Some manufacturers build the corrections into their camera software which lets them make other engineering choices, not just lower price but small size and weight, faster autofocus, etc. Note that this correction does not involve upsampling; Photoshop includes upsampling, with different algorithms (like our filters) optimized for different goals, but that is an unrelated feature, on a different menu, you can apply lens correction or upsampling or both.
If you have older photos, maybe scans of film photos, and of not have built-in metadata, things are more difficult. Maybe you can measure the old equipment if you have access to it, or you can eyeball it.
That is a pretty direct parallel to MQA ADC correction.
B. In photography we also routinely compensate for limitations on the output side. Photoshop has profiles of printers, inks and papers, and can adjust colors and sharpness for different specific output channels and for specific psycho-visual or artistic goals.
To extend the parallel, printers and printer drivers often include their own corrections for similar goals. You can ignore the problem and leave it to the printer manufacturer, you can use a combination (e.g. sharpen for the print purpose but let the driver do color adjustment), of turn off all device processing and take complete charge in the software.
C. The origami technique is a form of compression. It works in a way different from most existing techniques. The goal is to support high quality sound at bandwidth that is higher than the goal of MP3 twenty years ago, but still low enough to work for streaming.
We can see a few conclusions from this way of thinking about it.
None of the steps alters the sampling rate. It is not upsampling. This is why we see MQA content at all resolutions, even 44k.
It might be worthwhile to also do upsampling, or increase the bit depth. We know the benefits and limitations of these techniques. But they are independent of the compensation for AD and DA conversion flaws. Again, just like in photography.
If we understand clearly what the techniques and goals are for each of the three steps, we can then evaluate, for each of the three techniques:
- Is it worthwhile? Does it address a problem that matters?
- Is it effective, does it achieve its goal?
- Does it operate without harmful side effects, from a technical and sound quality perspective?
- Are there unattractive commercial effects?
I think the Photoshop parallel illustrates why these should not be difficult to understand, or controversial. Compensating for upstream or downstream aberrations are well established techniques.
Will MQA lead to all systems sounding the same? Lens correction in Photoshop has certainly not led to all lenses being the same, although it has led to certain aberrations being less important and tradable against other attributes.
There is also the question of why these techniques are combined in one solution. E.g. why can’t we apply the origami compression without the time smear correction? Or time smear without origami? And why not include others, like room correction? I think the answer to these is partially about consistent goals, and partially about commercial viability. Ultimately these are MQA business decisions, which will be successful or not. But the field is certainly open for anybody to compete with alternate solutions, with different engineering and business trade offs.
