AVS Forum banner
Status
Not open for further replies.
1 - 20 of 94 Posts

·
Registered
Joined
·
181 Posts
Discussion Starter · #1 ·
I've set up a blind test in order to check, in a controlled way, if kmixer effect over audio streams is really as evident as some people claim.


The test consists in being able to tell in a blind, statistically reliable way, an unprocessed file from that same file after being bitmangled by kmixer.


In order to perform the test, you need to download the original unprocessed file and the kmixer processed file. The files are near-30 sec. wav files, compressed with lossless Monkey's Audio compressor (ape format). The original can be downloaded here , and the kmangled file here . They are around 2.3 MB each.


In order to set up the blind test, you must also download this zip file , and extract its contents in same directory where the ape files are stored. Execute then the 'test_setup.bat' file, and it will decompress the ape files and create a series of wav files. It will generate 'A.wav' and 'B.wav' reference files (A is original, B is kmangled) and others named 'Xnn.wav', that you must try to identify just by listening. There are more explanations over the procedure during the process.


By default, the test consists of 8 trials. In order to achieve a realiable identification (95% confidence, or probability of being guessing below 5%), you must identify at least 7 of the 8 wav files. You can use less trials (5 minimum) or more, just edit the bat file. When checking the results wou will read how many correct identifications are needed.


In order to listen to the files with bit-perfect precision, you must use Win9X, or Win2k or WinXP and a player/soundcard that is inmune to kmixer. Winamp or foobar2000 plus kernel streaming or ASIO plugins will be fine in most cases (not with Soundblasters, sorry). You can also burn the wav files and listen to them with you favourite hi-fi equipment, if you wish.

Technical details:

The original was extracted with 100% accuracy from Chris Isaak Wicked Game cd-audio, using EAC. The kmixer bitmangled file was obtained by playing the original file on a Revo soundcard through its SPDIF out, with foobar2000 0.7RC13 using directsound, and recorded with the SPDIF input of an Audiophile 2496 soundcard with CEP 1.2a. No editing (but cutting out unneeded parts of the song) was done at the files. Same procedure, but using kernel streaming output instead, yielded a bit-perfect recording of the original. The computer used was a P3 600 MHz with WinXP pre-SP1.


Edit: more tips about how you can perform the test at the posts #13 and #43.
 

·
Registered
Joined
·
283 Posts
Grateful if you could explain why this won't work with Soundblasters.


I'm using an Audigy 2 and kernel streaming within Foobar2000. I'm upsampling to 48khz within Foobar and it certainly sounds better than the standard method, so why won't I be able to use these files?


Thanks


geoff
 

·
Registered
Joined
·
181 Posts
Discussion Starter · #3 ·
I just intended to say that SoundBlasters can't achieve bit-perfect playback at 44.1 KHz. If you resample, you won't get bit-perfect playback.


FB2k resampling is very good quality, though, so you can try the test and see if kmangling previous to resampling (not the real situation though) does a difference in your case.
 

·
Registered
Joined
·
37 Posts
Thanks KikeG for taking the time to set up that test - excellent idea.


I can't pick the difference (and neither can my wife for what thats worth). It would be interesting if a few other people can post whether they can or not.


I ran the test using the ASIO plugin on Winamp. I had the digital out of my Revo soundcard connected to a relatively high-end CD player (Audio Analogue Maestro) and then on to a reasonably decent amp and speakers.
 

·
Registered
Joined
·
551 Posts
Thanks **** for making this test available, creating kmangled data and a program to make test materials. :D BTW, what condiguration did you use to make that kmangled sound image? I guess one of M-Audio cards.


Anyway, I can't pick the difference too with RME Digi96/8 PST, headphone amp, and HD600. ;)
 

·
Registered
Joined
·
37 Posts
Wormzer -- no I am not sure.


All I did is follow the instructions in the FAQ thread on this forum. I am definitely using the ASIO plugin with Winamp, and I am definitely getting output from the SPDIF of the Revo. So it appears to be working, but I am unfortunately unable to run the tests suggested in the FAQ to check if the output is bit-perfect because I am outputting to a CD player with a digital input, not a receiver. All the tests to check for bit perfect output rely on outputting a DTS file to a receiver (which I don't have).


If anyone can shed any light on whether I am doing the right thing or not or suggest another way to check if the output I am getting is bit-perfect or not I would be grateful.
 

·
Registered
Joined
·
3,510 Posts
If we use the analogue outputs on sblive, is there still a problem listening to these 2 files? Isn't the kmixer resampling only done when using the digital output? Also, kernal streaming gives me audio in Foobar with sblive+analogue outs, does this mean it's bypassing kmixer?


Questions aside, the kmixer wav has significantly less bass and there is a definite difference in the two clips' overall tonality and soundstaging.
 

·
Registered
Joined
·
283 Posts
KikeG


Thanks for the explanation re Audigy 2. It makes sense to me now. Of course you won't get bit perfect playback at 44.1 since the Audigy insists on resampling to 48khz; but as I've found and you've agreed, if you do the resampling in Foobar the results are much better than if you let the Audigy do it.


geoff
 

·
Registered
Joined
·
181 Posts
Discussion Starter · #10 ·
Quote:
Originally posted by kazushi
BTW, what condiguration did you use to make that kmangled sound image? I guess one of M-Audio cards.
I don't understand what you mean exactly. If you refer to Revo control panel, I put the card in digital mode, set all output levels to max, and disable all DSPs and crossover.
 

·
Registered
Joined
·
551 Posts
Quote:
Originally posted by Jesse S
If we use the analogue outputs on sblive, is there still a problem listening to these 2 files? Isn't the kmixer resampling only done when using the digital output? Also, kernal streaming gives me audio in Foobar with sblive+analogue outs, does this mean it's bypassing kmixer?
Even if kmixer doesn't resample, sblive itself is sesampling any 44.1kHz data into 48kHz. Like other suggested, the best way is to use foobar's resampler and output everything in 48kHz.


Maybe new 2ZS is not resampling regarding to their advertisement, but not sure.
 

·
Registered
Joined
·
181 Posts
Discussion Starter · #12 ·
Quote:
Originally posted by Jesse S

Isn't the kmixer resampling only done when using the digital output?
When it happens, it supposedly will happen using any output, whether analog or digital. However, there's evidence that in cards with hardware directsound acceleration, such as Creative ones, kmixer is avoided too.

Quote:
Also, kernal streaming gives me audio in Foobar with sblive+analogue outs, does this mean it's bypassing kmixer?
Yes, ks bypasses kmixer. However, when playing 44.1 KHz files (such as the ones at this test), Creative cards will perform its internal resampling to 48 KHz, and this resampling is of measurabily worse quality than kmixer one.

Quote:
Questions aside, the kmixer wav has significantly less bass and there is a definite difference in the two clips' overall tonality and soundstaging. [/b]
Well, the important part of the test is the blind one. Have you tried it? How many correct identifications did you get?
 

·
Registered
Joined
·
181 Posts
Discussion Starter · #13 ·
Some tips that could make the test easier:


- If you edit the 'test_setup.bat' file and add the option -xy at the end of the fileABX line, the program will generate pairs of A/B files, called Xnn and Ynn. At every pair you have to tell which is A and which is B.


- Same way, using the options -s and -e you can select just a part of the files to perform the test. Execute the fileABX program without parameters (or just from the explorer) to see the way of usage of this.
 

·
Registered
Joined
·
13,423 Posts
KikeG,


What difference does it make using either the 48kHz resampling in the SB Live card or 48kHz resampling in the receiver/prepro?


Seems to me that for this to be a valid test you would have to use "bypass" mode in the receiver/prepro to avoid the DSP.
 

·
Registered
Joined
·
13,423 Posts
Quote:
Originally posted by wormzer
http://www.avsforum.com/avs-vb/showt...evo+asio+spdif


Sounds like ASIO + spdif + revo is most definitely broken and has waining hopes of being fixed by M-Audio. Not sure how you're hearing sound this way, davvy...


matt
Matt,


As a result of my posting the tech support email address (and member input) the broken ASIO over S/PDIF has been moved back to top priority.
 

·
Registered
Joined
·
15,430 Posts
Quote:
Originally posted by davvy
I can't pick the difference (and neither can my wife for what thats worth). It would be interesting if a few other people can post whether they can or not.
and
Quote:
Originally posted by kazushi
Anyway, I can't pick the difference too with RME Digi96/8 PST, headphone amp, and HD600. ;)
There is a very good reason why you two can't hear the difference. Because they're the same. Mathmatically subtracting the one from the other gives you a file full of 0s. Anyone passing this blind test is a liar.
 

·
Registered
Joined
·
1,639 Posts
Quote:
Originally posted by Cliff Watson
Matt,


As a result of my posting the tech support email address (and member input) the broken ASIO over S/PDIF has been moved back to top priority.
:) Good news!!!


Cliff,

I just noticed my Lexicon reports 44.1KHz when playing music on my Revo via SPDIF. Shouldn't the Revo be sending out 48KHz? Is this be resampled twice. Once to 48KHz and then back to 44.1 by the card? I'm using the SqrSoft Advanced Crossfading Ouput v 1.75 configured for gapless playback.

SqrSoft
 

·
Registered
Joined
·
15,430 Posts
Quote:
Originally posted by Mad Chemist
:) Good news!!!


Cliff,

I just noticed my Lexicon reports 44.1KHz when playing music on my Revo via SPDIF. Shouldn't the Revo be sending out 48KHz? Is this be resampled twice. Once to 48KHz and then back to 44.1 by the card? I'm using the SqrSoft Advanced Crossfading Ouput v 1.75 configured for gapless playback.

SqrSoft
No, it shouldn't be sending out 48kHz unless you're playing a 48kHz file. Windows doesn't force the sampling rate to 48kHz on most cards, the hardware/drivers do. The whole AC97 compliance thing. Now, that doesn't mean the file isn't mangled by kmangler. It just means it hasn't changed the sampling rate.
 
1 - 20 of 94 Posts
Status
Not open for further replies.
Top