Fons' solution for mixing 1st and higher orders

[e deleflie]

Can I ask you guys to comment on Fon's proposed workaround (from
sursound) (I've copied the essential below ... but I'm sure you've
read his post on sursound)?

If this is indeed a possible solution ... then I am going to change
the Universal Ambisonic spec to specify one and only one component
combination, being 3P.

And I will also change the specification for decoders to include Fons'
proposed strategy.

There would also be the need to include the process for mixing 1st
order with Universal Ambisonic.

... and then we have a good solid easy to implement processing format.
Compression (for distribution), in the first instance, can just
involve omitting higher orders.

Etienne

quote:
"The decoders are dual band, max-rV for LF, max-rE for HF.
The same matrix is used for both LF and HF, but for the
HF part there are additional per-order gain factors
applied to the input signals (so the effective matrix is
indeed different). It is these gain factors which are
shown below. The leftmost one is for the 0-order signal,
going up to the right.

First order, regular octagon: 1.960 1.420
Second order, regular octagon: 1.650 1.400 0.825
Third order, regular octagon: 1.414 1.307 1.000 0.542

Now look at the ratios of the first two in each line.
They are different for each decoder. This means e.g. that
if you mix a first order signal into a third order one,
the third order decoder will use it *assuming* it is part
of a third order signal, and consequently apply too much
gain to the first order components.

You can get this right if

- you know the decoder that will be used,
- and apply shelf filters on the lower order
signal before mixing it with the higher one."

[Oliver Thuns]

If you apply lossless compression, it could make sense to keep the 1st
order and the 3rd order signal separate. The size of the two signals
separately compress is maybe not much bigger than the compressed mix
of both signals.

An analogy for better understanding:

You have a song with voice and guitar. FLAC can compress the WAV file
to 60% of the original size. When you compress the voice track and the
guitar track seperately with FLAC, the added size of both FLAC files
should be around 60% to 80% (let's say 40% for the guitar and 30% for
the voice). I haven't tested it, it's just an assumption.