|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
60.227.51.246
In Reply to: here's the picture posted by Werner on January 29, 2007 at 22:44:31:
Your graph shows content above 25kHz, around -72dB.Yes, there seems to be a notch around 20-30kHz, but there could be many reasons for this.
If your claim that this is an upsampled recording is true, there should be nothing above 24kHz (ie. artefacts should be below the digital noise floor of -144dB, not -72dB).
Follow Ups:
"so this is a 48kHz recording upped to 96kHz by mere sample replication!"If you go from Fs 48kHz to 96kHz by pairwise replicating samples you fill in the band between 24kHz and 48kHz with the first image, somewhat attenuated by the sinc(f) function. And that closely matches what we are seeing here.
Except that in this file (I checked in an editor) the samples are not pairwise identical, but near-identical (making it all the more mysterious). So obviously something went wrong somewhere with this datastream.
You could argue that the fault lies with my Flac-> Wav conversion (foobar), but I used the same flow on all of the files, and the others came out just fine.
I am not suggesting that HDTT are cheating (why would they?), but rather that there might be something wrong with that particular file.
bring bac k dynamic range
Anything's possible? maybe the Flac had something to do with it? haven't experimented with Flac that much this is the first time, maybe somebody with more knowledge about the program can comment on this situation, even if it is possible??
I can assure you I don't upsample any of my releases there really isn't a need for it, and i never master at 24/48, so all this is very strange to me.
Thanks
Bob
HDTT
I opened the file in Adobe Audition, and applied the "Spectral Frequency Display" view, and you are right - the spectral pattern is curiously symmetrical.And the samples do appear to be valued in pairs.
This post is made possible by the generous support of people like you and our sponsors: