[pulseaudio-discuss] Mono Upmixing Only

[David Henningsson]

On 04/04/2013 05:41 AM, Alexander E. Patrakov wrote:
> 2013/4/3 Tanu Kaskinen <tanuk at iki.fi>:
>> On Wed, 2013-04-03 at 10:20 +0100, Carl Gridley wrote:
>>> Wow, wasn't expecting code so quickly - thanks!
>>>
>>>
>>> I've been away for a bit. Is this working and done? Do I need to get
>>> PA 3 and a patch on my system or is it all good and in the queue for
>>> an offical release in the future?
>>
>> You need to apply the patch yourself, because my solution was not
>> considered good. It seems that this was only discussed in IRC, not on
>> the mailing list. A bug report has been filed, which contains the
>> explanation why the solution was not accepted:
>> https://bugs.freedesktop.org/show_bug.cgi?id=62588
>
> I'd say that the "no should mean no" argument is at least incomplete,
> in the sense "more discussion is needed". First, did the opponents
> present a use case when a pure "no" (with all associated side effects)
> would yield the correct result for all use cases?

I think the primary objection was not that there was a use case or need 
for the mono-upmix-only option. The objection was mostly that having a 
"enable-remixing = no" option that actually did remixing would be quite 
confusing.

> Second (semi-trolling, don't take too seriously) remixing by channel
> copying and linear summation is so old-school! No modern 5.1 receiver
> does it by default when fed a stereo signal over SPDIF or HDMI. They
> have a US-patented upmixing algorithm from Dolby inside, using Z
> transformers and even non-linear elements, and most so-called "stereo"
> records are produced with that "matrix decoder" in mind. So, once this
> fancy upmixer is completely reverse-engineered, it should be an option
> at least in the countries where the patent is invalid. And yes, this
> does mean that the "enable-remixing" option is a misnomer, as it is
> not a boolean.

The first match on Google on the topic yielded this result:

http://www.sersc.org/journals/IJSIP/vol2_no4/7.pdf

It looks like the first option, "Passive Surround Decoding" yielded 
pretty good listening results and seems not too difficult to implement. 
The other steps - time delay, low-pass filter, phase shifter - also 
needs somebody to implement them though.

Also; I guess in some cases distributions and others have been reselling 
patent licenses, so even in patent-covered countries there might be a 
possibility to use such an algorithm.


-- 
David Henningsson, Canonical Ltd.
https://launchpad.net/~diwic