Better resampling algorithm (Secret Rabbit Code)?

[Xenakios]

Originally Posted by Noirceur

I use a software called ABX for real blindest. http://lacinato.com/cm/software/othe.../othersoft/abx

The global results show my ear dont find lot of difference between Isotope resampling and libsoxr. But it find difference between Reaper resampled and other.

Well, I guess that's an interesting single result. More people testing this would be needed, though.

edit : This talk about Reaper's resamplig algorithms being "worse" has been going on for years but I think it's mostly, if not entirely, based on the analysis results that are shown on the website mentioned earlier in this thread among other forum threads here. But I haven't seen reliably gathered blind testing data about it. Measurements and geeking about those are one thing, but keeping things pragmatic is another. If a difference between various alternatives can't reliably be heard by blind listening, then it's not worth it for anyone (including the Reaper developers) to spend time and effort on it.

[Noirceur]

Personally, I don't find reaper resampling is "bad", to my ear it's very good, but not as good of the best resampler of the market.

I will be happy to see the next improvement ! Reaper 5 beta is already online ?

[Noirceur]

I just tested the improved resampling of Reaper V5 beta. Just compared R4 resampling vs R5 beta resampling on same music file with my ear. So, this first impression is 100% subjective, but still with ABX blindtest software.
Damn is sound so good, more "smooth" more "precise". I'm very happy now ! Can't wait the official release !

Note : OK sorry its not the good forum for speak about beta of reaper.

[evosilica]

Originally Posted by Noirceur

Personally, I don't find reaper resampling is "bad", to my ear it's very good, but not as good of the best resampler of the market.

It doesn't really matter, if there's something better, when you can't hear it, right?
Just like there's no need to make your bathroom mirror flat on the atom scale, because you can't see it.

If your final format is 44.1 16bit, then all the errors you see in these pictures will vanish in dither noise.

[clepsydrae]

Originally Posted by Noirceur

I just tested the improved resampling of Reaper V5 beta. Just compared R4 resampling vs R5 beta resampling on same music file with my ear. So, this first impression is 100% subjective, but still with ABX blindtest software.
Damn is sound so good, more "smooth" more "precise". I'm very happy now ! Can't wait the official release !

When you say that you can hear a difference using ABX software, you should say how many ABX trials you did, what percentage you got correct ("Accuracy"), and the "confidence" result that you got as well. What values do you get for these? You should also post the audio files that you used. Thanks!

[Noirceur]

I already do a ABX test, I don't want to doing again now, its required time and effort and concentration and its very fatiguing. If you want, will do it again but the next week and will post here more precise cue statistic !

What my previous test show : in 10 blind test, i find the reaper 5 resample 9 time. (9/10)

After the test, i checked if there are real difference between le last and the next resampling and the difference between the old and fixed filter was about -80 dB, and only at a very high frequency. I do this with difference phase on reaper.

The audio file was a music from Beck at 192khz resampled to 44,1khz with batch converter in "EXTREM" mode.

I will be happy to see your opinion about yours test too.

Sorry for my bad english

[clepsydrae]

Originally Posted by Noirceur

I already do a ABX test, I don't want to doing again now, its required time and effort and concentration and its very fatiguing. If you want, will do it again but the next week and will post here more precise cue statistic !

That's not necessary, if you remember that you got 9/10 that comes out to confidence 99% and accuracy 90%, which is a great result. But we do still need the sample files that you used so that we can hear what you were actually testing yourself with (if you're worried about copyright you could upload just a small section.)

The reason I'm asking is that if you were able to hear a difference you would be (as far as I'm aware) the only human on earth to ever be able to do so, so it would be a pretty amazing result.

So I'm wondering if maybe there was some mistake made in the creation of the files... if you could also upload the original 192k file so others could try to do the resampling as well, that would be helpful.

In the past, people have said they can hear a difference between X and Y, but then they upload their test files and it turns out they made some mistake with one that makes the difference very obvious. I just wanted to make sure that hadn't happened in this case.

After the test, i checked if there are real difference between le last and the next resampling and the difference between the old and fixed filter was about -80 dB, and only at a very high frequency. I do this with difference phase on reaper.

Interesting. Sounds like the conversion was good, then. Still, it'd be good if you can provide the files you used.

I'm also wondering what levels you were listening at? Meaning, did you turn the volume up very high? And were you listening to the whole track or did you focus on a small part of it?

Sorry for my bad english

No problem!

[evosilica]

Originally Posted by clepsydrae

The reason I'm asking is that if you were able to hear a difference you would be (as far as I'm aware) the only human on earth to ever be able to do so, so it would be a pretty amazing result.

Furthermore the equipment he must have used to accurately reproduced those artefacts does (as far as I'm aware) not exist on this planet, which would be just as amazing.

Seriously, this test was either heavily flawed, or did not even happen. >:|