Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Sound Card Challenge (Read 8015 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Sound Card Challenge

Maybe some of you remember my soundcard challenge at www.kikeg.arrakis.es, where I challenged people to detect a sound clip that had been passed several times over my soundcard.

The thing is that I only received one response from one person (from here, by the way), that I know has very good equipment, and his ratings were just random. This initial test had some eq. over the files, but later I changed the test and eliminated all eq. or extra digital amplification.

Following the recent trend about ABXing equipment, I think it would be interesting to know if any of you can at least tell the clip passed 3 times (without any additional eq. or the like), through the soundcard (M-Audio Audiophile), from the original clip.

The identification of each clip is:

wick_1.ape: passed 1 time
wick_2.ape: passed 3 times
wick_3.ape: passed 2 times
wick_4.ape: passed 0 times (original)

Every file is 2.56 MB.

I think it is a quite a tough challenge, what are your impressions?

Sound Card Challenge

Reply #1 –
I compared wick_4 and wick_2, but I couldn't prove any differences, while ogg -q8 is obvious with castanets.wav.

I had to listen through onBoard-sound (for now), maybe the differences are easier to detect on hi-end equipment.

(Funny, that you identify the files on this board; do you trust us?  )

Sound Card Challenge

Reply #2 –
Quote
(Funny, that you identify the files on this board; do you trust us?  :P )


Well, I consider the challenge closed, since only one person sent me their results in three months, even when quite many people have downloaded some of the files. I have to update the web page.

Sound Card Challenge

Reply #3 –
I compared wick_2 vs. wick_4 and got:

6/6
10/12
13/18

If I had used the 28-trial profile, I would have passed, since it would have stopped at 6/6.

But actually I can't really be sure I heard anything!  I randomly chose a different starting point at 13.5 seconds, and thought I could make out that there was a slight difference, perhaps only in volume.

I used wavgain to determine that wick_2 is 0.2 dB louder than wick_4.  I already know that 0.2 dB is right at my threshold for being able to detect volume, so my guess is this is what I heard, if I actually heard anything at all.

ff123

Sound Card Challenge

Reply #4 –
Strange, since, according to CEP, there's a RMS level difference of just 0.04 dB taking into account the entire unweighted frequency spectrum.

However, taking into account only a 500 Hz band centered on 1 KHz, the difference is of just 0.01 dB.

The only differences remaining then would be a slight attenuation on high frequencies (over around 14-15 KHz) , and a slightly higher background noise and distortion.

But, your results are quite conclusive, I thought nobody was going to be able to ABX it.

If CEP RMS calculation is correct, my bet is that the perceived differences are due to  the slight attenuation of high frequencies, even when it's only at relatively high frequencies.

However, interesting results,  I guess I should analize more thoroughly the objective differences and/or a possible level imbalance.

Sound Card Challenge

Reply #5 –
Or it could be that the wavgain calculation isn't completely correct, especially for very small differences.  However, I do note that it appeared to work ok for fewtch's turntable sample.  I'll try to take a closer look at your files tonight.  Does wavgain remove DC offsets, and could that be causing it to miscalculate?

Also, while I got as much as 8/8 correct on this sample, each correct answer was somewhat surprising to me, because what I thought I was picking up was such a slight difference that I'm not even sure I know how to describe it or if it was even real.  Also, if I got 8/8, then that means I got 5/10 on the second half.  So I may have gotten extremely lucky in the beginning.

ff123

Sound Card Challenge

Reply #6 –
wick_4 and wick_2 sound the same to me.

Sound Card Challenge

Reply #7 –
I encoded wick_2 and wick_4 to mp3 using lame aps, and then analyzed it with mp3gain instead of wavgain.  After encoding to mp3, the files were at exactly the same volume.  Moreover, when I decoded back to wav, the files were still at the exact same volume when using wavgain.

So using wavgain on the original files gives a different answer, for some reason.

ff123

Sound Card Challenge

Reply #8 –
Quote
Also, if I got 8/8, then that means I got 5/10 on the second half.  So I may have gotten extremely lucky in the beginning.

Could you try to ABX them again, just to be sure? 

Tonight (it's morning at here now), I will post frequency response differences between both files, I will also post the wick_2.ape file with just  this frequency response corrected to match the original file, to try to evaluate if it's that what's affecting perception.

Sound Card Challenge

Reply #9 –
ok, I'll try this again when I get another quiet time (probably late tonight)

ff123

Sound Card Challenge

Reply #10 –
The frequency response mismatch between wick_2.wav and wick_4.wav is this:

log frequency scale:



linear frequency scale:




I've corrected frequency response of wick_2.wav, you can download it here: wick_2_corr.ape

Now, the frequency response mismatch is around or below 0.1 dB.

log frequency scale:



linear frequency scale:

Sound Card Challenge

Reply #11 –
I can't repeat the ABX.  There's nothing for me to latch on to.  The last time, I thought I might have been hearing a difference in something, but now there's nothing.

ff123

Sound Card Challenge

Reply #12 –
OK. Thanks, ff123.

Sound Card Challenge

Reply #13 –
What EQ did you use?

I thought you said you're NOT going to EQ your output in this trial?

Sound Card Challenge

Reply #14 –
Quote
What EQ did you use?

I thought you said you're NOT going to EQ your output in this trial?

The original files are not equalized, but KikeG posted a flattened version later for ff123 to compare:

Quote
I've corrected frequency response of wick_2.wav, you can download it here: wick_2_corr.ape

Now, the frequency response mismatch is around or below 0.1 dB.

Sound Card Challenge

Reply #15 –
Quote
What EQ did you use?

CEP 1.2a FFT filter, hand-made eq. curve, 1024 FFT points hanning windowing.

Sound Card Challenge

Reply #16 –
Quote
CEP 1.2a FFT filter, hand-made eq. curve, 1024 FFT points hanning windowing.


Linear phase or minimum phase?

So, are the results so far are supposed to be interpreted as 'nothing that a digital EQ can't fix'?  But a digital EQ in real time costs $$$ outside a computer, you know

BTW, Naoki calls his EQ 'an accurate equalizer for winamp 2. With 16383th order FIR filter'--does that stand for 16383 FFT points? :eek:

Sound Card Challenge

Reply #17 –
Quote
So, are the results so far are supposed to be interpreted as 'nothing that a digital EQ can't fix'?  But a digital EQ in real time costs $$$ outside a computer, you know

So far it's not even clear if there is anything to fix. 

Sound Card Challenge

Reply #18 –
Quote
Linear phase or minimum phase?

BTW, Naoki calls his EQ 'an accurate equalizer for winamp 2. With 16383th order FIR filter'--does that stand for 16383 FFT points? :eek:

Linear phase.

Quote
So, are the results so far are supposed to be interpreted as 'nothing that a digital EQ can't fix'?  But a digital EQ in real time costs $$$ outside a computer, you know


No, this time I just wanted to know if the possible perceived differences were just due to frequency response differences, or there were other things.

Quote
With 16383th order FIR filter'--does that stand for 16383 FFT points?


I'd say that means 16384 points.