original is 51 with blank LFE & mono surrounds.

Limitations: software decoders only, and DPL1 was not tested since hard to find free software decoders.

Conclusion: Although DS/DPL1 decoders might be more authentic to period limitations,  DPL2 preserves the original better, Even when the audio is DS encoded, since DPL2 implementations have 20-2000ish bands of steering vs DS of 0 and DPL of 1. Not to mention, DPL2 decoders are downmix-compatible, unlike DPL1 decoders.

DS-like decoders you can find everywhere, and example of DPL2-like decoder is freesurround.

Sometimes the best things in life are free, and just cuz it's old/rare doesn't mean it's good.

May you send me the untouched 5.0 file? Wanna make some experiment.

I guess it's 4.0 with two copy of surround, right?

(2021-10-24, 05:30 PM)spoRv Wrote: May you send me the untouched 5.0 file? Wanna make some experiment.

I guess it's 4.0 with two copy of surround, right?

(2021-10-24, 12:13 PM)junh1024 Wrote: original is 51 with blank LFE & mono surrounds.

It is a simple test with ambient/wide music.

I would attach the file except it doesn't work for me.

L: https://www.fromsmash.com/w8HYp-3S6s-gt

I used your file to make some tests, but I did not like it because left and right contains music, and it's difficult to discern results in a clear way - you don't know if an encoder add small unwanted sounds on other channels because it would be masked by that music.

So I made a comparison using only spoken channel names; you can find it here: http://blog.sporv.com/surround-test-1/ - take a look, listen to the clips and let me know what do you think.

(2021-10-27, 03:05 AM)spoRv Wrote: I used your file to make some tests, but I did not like it because left and right contains music, and it's difficult to discern results in a clear way - you don't know if an encoder add small unwanted sounds on other channels because it would be masked by that music.

So I made a comparison using only spoken channel names; you can find it here: http://blog.sporv.com/surround-test-1/ - take a look, listen to the clips and let me know what do you think.

Interesting write-up. Out of curiosity do you have a blog post detailing "your way" of software decoding?

(2021-10-27, 03:05 AM)spoRv Wrote: I used your file to make some tests, but I did not like it because left and right contains music, and it's difficult to discern results in a clear way - you don't know if an encoder add small unwanted sounds on other channels because it would be masked by that music.

So I made a comparison using only spoken channel names; you can find it here: http://blog.sporv.com/surround-test-1/ - take a look, listen to the clips and let me know what do you think.

While your tests shows how they will perform at their best, on sterile content & will look good for any decoder that has any steering, even as low as 1 band.

My test has music which better represents a real-world scenario (of a movie soundtrack) , but also is channel calls, so you can tell how well steering performs on a busy mix. This will make decoding more challenging, so is more representative of "typical" content, and a probably a better indicator of performance on movie soundtracks, which is why I did it.

Either way, it goes to show "HD matrix surround™", falls *significantly* short of its suggested performance & free DPL2 methods such as freesurround perform best generally.

PS: for your “my way” decode, is that manual copypasta of waveforms?

---

---

(2021-10-27, 07:01 AM)junh1024 Wrote: While your tests shows how they will perform at their best, on sterile content & will look good for any decoder that has any steering, even as low as 1 band.

My test has music which better represents a real-world scenario (of a movie soundtrack) , but also is channel calls, so you can tell how well steering performs on a busy mix. This will make decoding more challenging, so is more representative of "typical" content, and a probably a  better indicator of performance on movie soundtracks, which is why I did it.

Either way, it goes to show "HD matrix surround™", falls *significantly* short of its suggested performance & free DPL2 methods such as freesurround perform best generally.

PS: for your “my way” decode, is that manual copypasta of waveforms?

Sorry for the long delay, but it took some time to prepare an answer!

First thing first, I moved your posts (and the following ones) into a new thread, because I think it's worth it.

Then, back to the answer: you are right, the test I made proved nothing, as it should be conducted on a real world source; so I took some (a lot of, actually) time to make a proper test, this time using a clip with very hard material... I wrote a long article that you can read here:
http://blog.sporv.com/i-was-shot-today-s...parison-2/

tl;dr: while the FreeSurround is not that bad as surround upmixer when listened through a surround setup, it fails (at least in the worst case scenarios) when listening to it as downmix, while the hardware decoders do the same good job on both surround and downmix occasions.

And, about the so-called "my way", it fails more or less like FreeSurround - but a bit less "phasey" sometimes - and no, it is not a "copy&paste" even if, looking back at my other comparison, it could seem so!

After all, it seems I stand corrected: decoding an encoded Dolby Surround track using hardware decoder gives a worthwhile result, better than free software surround upmixer; can't say if paid software upmixers could rival or best those old hardware decoders; if they could, they would decode tracks in a matter of minutes, while hardware would take the same track time + the needed sync time; still, the latter could be found for few bucks while the former cost several times more.

By the way, all files (including source) are available to download, so anyone who would like to spend some time decoding it with any kind of surround decoder, would be it hardware or software, free or paid, is more than welcome! Don't forget to post your results here, though!

Thanks for doing some more tests. freesurround has some settings which might improve it. Can you please try:

• speaker: 51 (original method)
  

This might fix the phasiness on 20.

Tried; phasiness remains; it seems to switch and mash up some contents between L and SL - infact, the phasey "Oui, Bwana" first found on the SL is now on both SL and L channels!