225WAD - 2nd Order Ambisonics Decoder with Multiple Inputs Real Ambisonics
Needed Software :
- FreezeRMS group
- DCKiller (included in package)
- RMS Buddy (included in package)
- Panorama (included)
- Wig Ambi Decoder 2nd order (included)
You can download the package here : 225WAD-MIRA
The layout :
The first thing you notice is the multiple fileplayers. This method depends on
the output from other methods. As an example we use the output from CCG as we
discussed in CCGLCR. The same workflow is used here : upsample your stereo to
32/96. Run your stereo through CCG and you end up with stereo32/96.wav,
stero3296-C.wav and stereo3296-S.wav. Downsample these back to 32/44.1 and put
the files in the right player. Don't forget you need to use the 32 bits stereo !
The settings don't differ from 225WAD :
For an explanation of these settings, see the 225WAD guide.
Worked out example :
Upsample your stereo to 32/96
Load your stereo into CCG
First downsample back all files to 32/44.1. I kept the names here the same (!)
Load the layout and load stereo3296 in the first player, 3296C in the center
player and 3296S in the rears player (the downsampled ones):
After the first run you get these values in the upper RMSBuddies :
Make the inputs as loud as possible with even made RMS values :
Run with the gains and this setting for the decoder :
The outcome is in the RMSBuddies in the bottom :
The problem with any Ambisonics method, is that changing the gain of one
channel, also influences the other channels and that the gain is not linear,
which makes it a bit of gambling for the right values. But when you have done
this a few times, you recognize patterns and know with which value to start.
Before reaching 0 Db PEAK in the rears, we can add + 4 Db in the "mastergain >
pregains" and run it again :
You see that the + 4 Db, is not enough and the other values changed too. Let's
see what happens with +5 Db :
And voila ! The rears look good. They are -1 Db compared to the fronts and the
center also has a desired value of 0.5 Db louder than the fronts. The only
one that needs a small adjustment is the LFE. We do that in the decoder to make
it a bit softer as it is now. Change the default value of 0.4 to 0.3 :
It's a tiny bit too soft, but acceptable. If you wish you can make it 0.32 :
Record the song with these values......
and make a DTS CD WAV from them. Listen to it on your HTS and compare with other
versions from other methods. Excellent separation, discrete channels and a
perfect soundfield ! This song does excellent with this method !
There's one setting that can have a huge impact on the total sound and that's the
"Ambi Decode Center". When you check it, the center is run through the Ambi
process, but when you uncheck it, only the corners are Ambi. On some voices this
makes quite a difference, but...... the gains are even harder to adjust. At the
first run with corrected gains on fronts, center and rears, you will get some
heavy clipping in the center. We fed it with a non clipping signal, so the
clipping takes place in the process and thus we need to lower gain there. But it
is far more difficult to get the right values, Like the previous example first
adjust the settings for the rears, then those of the fronts and finally the LFE
and the center. You can get strange and extreme values. Let's compare the gains
for a non-ambi center with those for an ambi center :
You see the giant difference for the gain in the center ! TRIAL AND ERROR !
Some times you get values in the rears that have a far louder peak than the
fronts but the RMS is lower. Do not change that. It's one of the charms of this
method that 3 Db lower rears in RMS with a peak almost 0 db, still will sound
like excellent separation. It's nice when you get like this example almost the
"perfect" values, but most of the times it's more like I just described : too
low rears in RMS, but higher peaks than the fronts. In the sound of the surround
this will means that some instruments come really loud out of the rears and when
those instruments are not used, the rears only add to the ambience.
But with all my experiments I did I found out that there is one thing you should
not do : DO NOT MESS WITH THE FRONTS EXCEPT IN PREPROCESSING !!
And : let the Ambisonics always be the last process in the chain. Changing gains
or whatever postprocessing, will destroy the Ambisonics soundfield.
For the rest it's just a matter of trial and error to get to the right values,
but it get's easier with every time you do it. The best way to do this, is to
start with the rears, then the fronts and then the center + LFE.
In this guide we used the inputs from CCG, but you can also use inputs from other
methods that are based on separation only. It then acts a bit like a second order
Ambiwrapper, but with a complete different sound.
Happy Encoding !!
Layout and Guide by The Dynamic Duo ©
The Dynamic Duo : Kempfand & Eye of Horus
It's not allowed to re-publish this method !