I was fooling around with some plugins for Spark XL and got the idea to
try something out, curiosity being what it is. Remember Cornell's page
giving equipment recommendations:
http://birds.cornell.edu/LNS/recordingnature/html/recordingnature_techequip2.html
In particular note it's Fig.1, which purports to show that MD is
unsuitable for recording tropical birds.
We have discussed how this was a test on a early version of ATRAC, but
have always assumed it was a valid test of that machine. Anyway, I'd
assumed that and I decided to duplicate the test and test my ATRAC 4.5
equipped HHb Portadisc on it. So, I first produced a aiff 44khz 16bit
file containing the two test frequencies, 6.5 & 7.5khz set with a little
headroom to avoid clipping. I then digitally played that to the
Portadisc and it recorded at a level of -10dB on the meter. Then I
digitally transferred the playback from the Portadisc back into another
aiff file. Using Spark XL's realtime sonogram facility I've produced a
comparison file of before and after:
http://frog_recordist.home.mindspring.com/naturerecordists/MD_6.5_7.5.jpg
This really is half the original and half the MD recorded audio, look
just to the left of the 2.3 second point and you will see a slight
transition in color, where I paused in one file and started playing the
other. To the left of that is original, to the right is ATRAC.
So, did ATRAC really improve all that much? I did some more
experimenting, which resulted in the following comparison sonogram:
http://frog_recordist.home.mindspring.com/naturerecordists/Clip_6.5_7.5.jpg
Compare this to the Cornell Fig. 1.
How did I get this? Well several ways, but the one you see was a result
of taking the sample file and increasing the gain to give it a level of
+4dB (if it actually could have done that) by taking a file in peak and
applying a gain adjust. In effect applying 4dB of clipping. The exact
pattern you get varies depending on things like exactly how much gain
you apply, and to what, but all are some form of splattering the
original two frequency bands all over the frequency scale if you do it
carefully all digital. I even know why it stays more or less in bands.
As you clip, you can only clip sample by sample and they are 1/44,100th
of a second apart. When the samples are reconstructed to a analog
soundwave, or analyzed by sonogram you appear to get frequency jumps
that correspond to changes in wavelength timing of that amount.
Now the question is what about Fig. 1? Is this really the effect of the
early ATRAC? Or do we have a example of another beginner in digital
sound turning the gain on the recorder up to where the meter reads about
zero or a little above, just like he'd done for many years with analog
tape? And clipping? It seems to me what was proved way back then was
that digital clipping will mess up your tropical bird recording.
Anyone have one of the early ATRAC machines we can repeat the test and
see what we really get?
And anybody who's still setting their levels in the analog manner on
their digital equipment go back and look at that 2nd sonogram again. The
sound, btw, sounds harsher in the clipped sample, but the primary signal
still dominates. And according to my son with his better high frequency
hearing is very penetrating and annoying compared to the original.
BTW, anyone who can't get at my two jpgs, let me know. My ISP has
rearranged their server in a very annoying manner and I'm having to redo
my site. And if anyone has kept bookmarks direct to any of the pictures
or sounds on my site, those addresses will be changing, you will have to
go back in through the main page, which should end up with the same address.
Walt
________________________________________________________________________
________________________________________________________________________
|
