Skip to main content
Topic: LossyWAV transcoding - personal test, part 1 (Read 5255 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

LossyWAV transcoding - personal test, part 1

I finally did a little ABX testing of lossyWAV's transcoding abilites. Since I was not too confident in my ability to tell any differences, I made the setup as easy as possible, but soon found that it was still nearly impossible to tell a difference. So I used only problem test samples. File A was LAME 3.98 -V5 transcoded from lossyWAV 1.1.0b --portable, File B was LAME 3.98 -V5 sourced from lossless.

Here are the results from round 1:
Code: [Select]
00595_ms    1/3    failed    total guesses
Atem-lied    0/2    failed    total guesses
badvibel    7/12    failed    started well, 3/3
bibilolo    6/8    failed    started well, 5/6
            
            
castanets    6/8        started well, 6/7
trumpet    0/0        couln't tell a from b; maybe too tired


I did not hide the results, because I was expecting to only do 8 trials each. I went higher with badvibel because I thought I could tell the difference but then remembered that is an invalid method, so that score would not have counted anyway had I succeeded. I got closest with bibilolo and castanets, but technically can't claim victory on those. Seeing the rapid success of my early trials on those two perhaps made me overconfident and more quick and careless as they progressed. Even still, the issues I was hearing were not that annoying to me.

I need to take a break before I can do any more. But so far it's encouraging, that I might be able to use lossyWAV --portable and be confident that transcodes (for me) will not suffer any more than a straight encode from lossless.

The equipment: Sennheiser HD 280 Pro (closed), Edirol UA-3 sound card (USB, does not resample - tested with udial)

edit to add lossyWAV version

LossyWAV transcoding - personal test, part 1

Reply #1
Many thanks for the testing - interesting results so far....

I take it that you are using lossyWAV 1.1.01b?

[edit] Oops.... reference to development beta removed. [/edit]
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848| FLAC -5 -e -p -b 512 -P=4096 -S-

LossyWAV transcoding - personal test, part 1

Reply #2
Sorry I forgot to mention the version - it's LossyWAV 1.1.0b, since it was the version posted to the first page of its thread. Would there be a change to --portable in 1.1.1b?

LossyWAV transcoding - personal test, part 1

Reply #3
Beta 1.1.1b is only the second beta in the development process for version 1.2.0. Development so far as been limited to modifications to the spreading-function (an averaging process). Further developments will include the incorporation of SG's noise shaping method (if I can get my head round the process properly....).

If you wish to test a beta version, please wait until 1.1.1c (tomorrow) as I am happier with the revised spreading function.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848| FLAC -5 -e -p -b 512 -P=4096 -S-

LossyWAV transcoding - personal test, part 1

Reply #4
I think for now I'll continue the rest of the samples with 1.1.0b, and in a later stage maybe I'll try 1.1.1c on samples that come close to positive ABX-ing (bibilolo, castanets).

LossyWAV transcoding - personal test, part 1

Reply #5
Here are my results for the rest of the samples I have:
Code: [Select]
eig	                     7/8	started with 7/7. slight added hiss. but i still felt like I was mostly guessing*
harp40_1             0/0 couldn't tell a from b
herding_calls            0/0 couldn't tell a from b
keys_1644ds     0/3
livin_in_the_future    0/0 couldn't tell a from b
S37_OTHERS_Martenotwaves_a 1/3
triangle-2_1644ds 1/3
Under the Boardwalk 0/0 couldn't tell a from b
fuious                 6/8 barely perceptible added hiss/coarseness


*The eig result is baffling. In each trial, I was listening for the slightest perception of added hiss/noise, but I had an utter lack of confidence in each choice. Such a lack of confidence that I was continually surprised to be correctly identifying X each time. So the good result on it is misleading to some extent.

Overall, it looks very nice for transcoding from --portable. I'm still not sure if I'll go with it or with --standard, so I think I'll test bibilolo, castanets, eig, and furious with --standard (as well as 1.1.1c --portable) and see how it goes...

 
SimplePortal 1.0.0 RC1 © 2008-2020