DIYHiFi.org

http://www.newscientist.com/article/dn8 ... atter.html

I needed approx 2 minutes to differentiate betwen sounds, but I was cheating, since I was listening in a loop with 10 repetitions. Since I do not consider myself to have a very acute hearing I am interested how others differentiate betwen sounds. To me difference was not in the first 40 miliseconds.

Best regards,

Jaka Racman

Interesting
I replayed the samples a few times and I think noticed the difference in less than a minute, but just because the text suggested there is a difference. In a normal situation both sounds would pass as the same sound.

The french sound has a faster attack followed by a constant decay and the hindi has a sudden "click" followed by a slight pause in the beginning. The last section of the sustained "dahs" sound equal, which is intriguing: It appears that both sounds were taken from a single sample of the old computer speech devices. Remember the Commodore Amiga? Or Kraftwerk?s Music Non-Stop? Or even Stephen Hawking! So, the french version sounds like the hindi sample, but truncated at the beginning.

I don?t think the test is accurate with this method if their target is to find how well we find subtleties in different languages. For that it would be best if they recorded a same person speaking in both languages.

If it?s just to find how accurate one?s brain is in miliseconds, they could get away with any other type of sound.
Keyboard magazine once did a test to find why an analog Minimoog was generally regarded by musicians as "punchy" and a digital Yamaha SY22 was regarded as "slow". So they recorded a percussive bass note with the least attack time both synths could play and plotted it. The results showed that the Yamaha had a longer rise time than the Moog, the difference being only a few miliseconds.

I don?t know if a musician is born with more white matter or if it?s the training that leads to it. In the pioneering days of digital sampling where state-of-the-art instruments like the Emulator One had no more than 256k of memory, musicians like the Art of Noise (or should I say programmers) went to great lengths to take the best use of so little time to play with. Often samples of human voices, or drums or guitars were heavily truncated just before the meaning or the character of the instrument wasn?t recognized anymore just to save memory space. My first professional sampler (1986) had a whopping 2Meg but I had to use the technique to take advantage of its multi-timbral capabilities in a MIDI tapeless studio.

Here?s a picture of the french waveform in Adobe Audition, togheter with photographer reflected in dirty screen.

Hindi waveform.

This is just to show how the mind can be "tricked". Notice the little peak on the Hindi. There?s no "click followed by a pause" ( followed by the final sustained sound) afterall apparently. It?s the spike that makes the first miliseconds sound quieter when in fact they have the same amplitude as in the french. The spike is acting as a delayed transient. The french waveform is much more uniform, which gave me the illusion of a faster attack transient. Curiously it looks like the profile of an airplane wing.

So, on paper something happened that was subjectively interpreted in a different way. Now it doesn?t matter anymore if I found the differences. It?s more interesting how the differences were perceived. Fascinating food for thought or what?

Best regards

Hi.

as I said my hearing is not very acute and I perceive difference as pitch of "a" wovel. One "a" has higher pitch IMHO.

Below is a picture that shows difference of both signals. I do not know where my perceived pitch difference comes from.

Best regards,

Jaka Racman

Jaka,

Is it the top curve that is the difference? What is the time scale of that wrt the original sounds?

Jan Didden

_________________
He who is about to speak the truth, should keep one foot in the stirrup - Dzengiz Khan

Yes, top curve is the difference and bottom is the original Hindi sound. You can invert one sound and mix both files together.

Time scale is the same for both files.

Best regards,

Jaka Racman

Hello Jaka,
Good work. I did as you described and got the same results. Looks like I goofed when I stated that besides the spike, the initial cycles had the same amplitude. A little more care in watching the draws immediately reveals that it isn?t so. My mistake. But I was correct when I found the sustained sounds (the "a" vowels) were the same. The difference prooves it. Accurate hearing eh?
Well, so it shows that both sounds are composittes.



Now that you mention, the french "a" vowel seems higher in pitch.

Howcome is this possible? Not counting suggestion, of course. Again my explanation refers back to the late eighties when sample memory was too expensive. Roland Corp. cited some psychoacoustic principle that says it?s more the initial portion of a sound that makes it recognized by humans, not the timbre. So, instead of sampling entire sounds, the Roland D50 used only very short samples of attack transients which were then followed by easily synthesized single-cycle waves. To emulate a flute, only the intial "chiff" sound was sampled and the rest was a triangle wave. The arpeggiating harp in Enya?s Sail Away is a D50 sound. It seems realistic enough. Only the "pluck" portion was taken from a real harp. The rest is just a rapid decaying square wave in 75% duty cycle, but it sounds like a real harp. Now, switching the "pluck" sample for a "pick" sample, we?d have a guitar, even if the rest of the sound is still the same square wave.

In the case of the Dahs, it has to be in the first 40 miliseconds that one sound is higher in pitch than the other, the phenomenon taken by Roland inducing us to believe it?s the entire sound, which is mostly the "a" vowel, where it would be more natural to find differences in pitch, not in the initial "Dhhh" sound.

Cheers