Need advice about mixing 5.1 down to stereo

[hello_hello]

I posted a link to the post with the original audio attached so you could compare the compression.
But it's here. The first sample "1 - MeGUI Downmix.m4a"

The only reason I also encoded the sample again using RockSteady was to make it fair. Originally I created those samples by encoding the whole audio each time, then I extracted the same section from each. To test R128Norm I simply re-encoded the extracted downmixed sample (which had been ReplayGain-ed), not the whole thing, so I thought I'd better do the same again with RockSteady.

The RockSteady version is created using my particular settings. Although I'll admit as RockSteady and R128Norm produced a pretty similar result, it did give me a little confidence my settings aren't completely silly.

I'm not even sure I'd say RockSteady lowered the dynamics more, as such. The reason I mentioned the first gun shot in the samples is because relative to the original, it's the same volume for the RockSteady version, but R128Norm actually increased it. Have a listen to the original and you'll see what I mean.

Keep in mind though, it's a little hard to compare compressed audio to the original version because you need to adjust the volume to compensate for the compression. My solution was to downmix the audio, then run ReplayGain on it (the MeGUI mixdown sample). I did the same thing after re-encoding with RockSteady compression. Obviously I didn't need to for the R128Norm version because it effectively compresses and applies ReplayGain at the same time. They're all very close to a target volume of 89dB. I'm not sure there's a better way to compare them.

I think RockSteady works on multi-channel audio but to be honest I've never tried to test it properly. Compressing multi-channel audio isn't something I'm ever likely to do myself.

[hello_hello]

There's another Winamp plugin called LoudMax which I've only tried briefly over stereo audio. The Winamp plugin version (unless it's changed recently) gives you nothing more than a single slider to increase the compression. Despite that though, I recall it working quite well and sounding quite good. I was happy with RockSteady so I never got around to playing with it more, but I thought I'd mention it in case you want to.

[Detmek]

Thanks for original sample. I listen it and I also did a Null-test of both of your normalized files against original. Here is screenshot of analysis.
It seams that R128Norm had more effect on quieter parts compared to RockSteady. This goes along my impression that R128Norm file sounded louder at the beginning compared to RockSteady.

I will need more time to analyze those results and compare it with what I hear.

P.S. I have LoudMax as VST plugin on foobar but I never used it. I don't even remember when it was installed. I may try something with it, too.

[hello_hello]

I agree. The R128Norm version is definitely louder during the first part, but that's using my settings for the RockSteady version. There's no reason why you couldn't crank it's amplification up a bit. Looking at the waveform, it's seems the first gunshot is definitely louder in the R128Norm version (compared to the original).

A lot of the difference may come down to the size of the window used to determine the volume. If the two DSPs use different size windows then I assume during sections where the volume changes very quickly and continuously, each DSP will interpret the average volume differently. Visually, it appears RockSteady is compressing more. The R128Norm waveform looks closer to the original. Whether that's good or bad depends on the result you prefer, I guess.....

I didn't even know there was a VST wrapper for foobar2000, but your previous post encouraged me to look. I found this one which appears to work well. Is that what you use? And while I'm asking...... now I know you can run VST plugins with foobar2000, do you have any recommendations?

Anyway.... I'm not sure about the LoudMax DSP yet. I tried the VST version. It definitely works, and it definitely sounds okay.... I'll need to play around some more to decide if it's compression works well for video/soundtrack audio. For the moment, there's a sample attached. The LoudMax threshold was set to -15dB, the output to -1dB, then I ran ReplayGain on the output file and converted it to AAC. As you can see from the waveform, it definitely compresses (I split each sample to mono and removed the right channel for the the screenshot). The first section sounds quieter than RockSteady, while the second sounds pretty similar.

These are the original waveforms, not a null-test.

[Detmek]

Sorry hello_hello, I was busy this days. I tried LoudMax and it does the job. Its just does not do this job automaticaly. I probably need to set gain sliders manually. I will try to experiment with R128Norm+Limiter and LoudMax (both have VST versions that I can use with foobar) to see which one does better job. By looking picture you posted LoudMax seams like a winner but we don't look sound, we listen to it.

About VST plugin recomendations, I don't use much plugins. I have BootEQmkII and You Wa Shock! for music (though YWS doesn't work well under Windows 8.1 and I have tired to use Izotope Ozone 5 but it is too much complicated for my light needs. I am not actually mastering anything.. I also have Bauer stereophonic to binaural plugin when I convert music for listening on my phone. For movies I only need some good compressor, nothing more. So, R128Norm or LoudMax, will see after more tests.