Can extended frequencies be created out of thin air based on the existing source? I have several old jazz and classical recordings re-released on CD. The original sources were extremely old vinyl or even cylinders! That means that nearly all of them have the highest frequencies roll off from 4k - 8k.
What I would like to do is a type of uncoded SBR where higher frequencies can be recreated based on the source I have. In virtually every case, the recorded instruments produce the initial pitches/timbres below the cut-off frequencies. So I am wondering if it would be possible to perhaps take from 2k or 4k on up and high pass filter it out. Then reproduce 2-4 octaves (overtones - yes I know there are more to oeverotnes that just the octaves but this is an example) above and add them into the original. I would need to know at what db level each octave would have to be added to mimic the original frequencies.
Is there any software that might do this already or any research in this area? Any hints on how to start such a project, if possible?
Full Version: Recreating Frequencies Above 8k
QUOTE(detokaal @ Apr 14 2005, 07:49 AM)
. So I am wondering if it would be possible to perhaps take from 2k or 4k on up and high pass filter it out. Then reproduce 2-4 octaves (overtones - yes I know there are more to oeverotnes that just the octaves but this is an example) above and add them into the original. I would need to know at what db level each octave would have to be added to mimic the original frequencies.
Adding overtones is also known as harmonic distortion. Each instrument (maybe each player of each instrument) would have a different profile. You could look for something meant to emulate tube amp sound.. minor amounts of even harmonics.
I remember using an audio editor with a "High Frequency regeneration" function (at least I think that's what it was called, but Google can't find it!).
It did what you suggest (SBR without knowledge of the original sound) and it sounded horrible and metallic.
However, there is continued research in this field (according to Google) - maybe we'll get something useable one day.
btw, a lot of early recordings (not cylinders or acoustic recordings, but certainly some 1930s electrical recordings) do have a reasonable amount of high frequency content, but it's often filtered for CD release to reduce the noise. Bad CD releases often have a sharp low pass filter at 4-6kHz, while the musical content can easily go up to 8-12kHz, especially on later releases.
Even on earlier releases, _not_ filtering the HF noise out allows our ears to "imagine" HF content that's not really there!
Cheers,
David.
It did what you suggest (SBR without knowledge of the original sound) and it sounded horrible and metallic.
However, there is continued research in this field (according to Google) - maybe we'll get something useable one day.
btw, a lot of early recordings (not cylinders or acoustic recordings, but certainly some 1930s electrical recordings) do have a reasonable amount of high frequency content, but it's often filtered for CD release to reduce the noise. Bad CD releases often have a sharp low pass filter at 4-6kHz, while the musical content can easily go up to 8-12kHz, especially on later releases.
Even on earlier releases, _not_ filtering the HF noise out allows our ears to "imagine" HF content that's not really there!
Cheers,
David.
QUOTE(2Bdecided @ Apr 15 2005, 10:22 AM)
btw, a lot of early recordings (not cylinders or acoustic recordings, but certainly some 1930s electrical recordings) do have a reasonable amount of high frequency content, but it's often filtered for CD release to reduce the noise. Bad CD releases often have a sharp low pass filter at 4-6kHz, while the musical content can easily go up to 8-12kHz, especially on later releases.
They should rather apply decent noise reduction instead of cutting all HF content away. Noise reduction tools have become very effective today, the only restoration tools which really work.
Since when have people making commercial CDs always done what they should have done? 
Some of these releases are produced very cheaply, and more importantly, even the output of a good denoiser sounds cleaner to most people if you remove the high frequencies afterwards (especially if there was very little top end to start with).
Many people would rather have a dull sounding track without noise, than a brighter fuller sounding track with noise. So they get what they want!
It would be interesting to release some CDs with all the noise of a 78rpm disc intact and watch market reaction - I don't think it would be favourably received!
Cheers,
David.
Some of these releases are produced very cheaply, and more importantly, even the output of a good denoiser sounds cleaner to most people if you remove the high frequencies afterwards (especially if there was very little top end to start with).
Many people would rather have a dull sounding track without noise, than a brighter fuller sounding track with noise. So they get what they want!
It would be interesting to release some CDs with all the noise of a 78rpm disc intact and watch market reaction - I don't think it would be favourably received!
Cheers,
David.
I have some CD's like that. "The Cream of Fred Astaire" is one. I think a couple of those may be ripped from a film soundtrack rather than 78, but they are definately not noise reduced.
One problem is that optical sound tracks don't respond to restoration as well as 78s - at least all the original de-clicking algorithms failed miserably on optical sound track kind of noise.
Where there are clicks and thumps, they're much longer than those found on 78s. Where there is "constant" background noise, it sometimes varies with the content more than that found on typical 78s.
Also, the high frequency limit on optical sound tracks usually more severe than you'll find on a record from the same year.
For all these reasons, a sharp low pass filter may be the best that can be done - for now. Though sometimes there's a better print of the film elsewhere that won't need so much noise removal - cheap releases will sometimes just use whatever they find first, even if it's damaged, and a much better copy exists in an archive or collection somewhere.
Cheers,
David.
Where there are clicks and thumps, they're much longer than those found on 78s. Where there is "constant" background noise, it sometimes varies with the content more than that found on typical 78s.
Also, the high frequency limit on optical sound tracks usually more severe than you'll find on a record from the same year.
For all these reasons, a sharp low pass filter may be the best that can be done - for now. Though sometimes there's a better print of the film elsewhere that won't need so much noise removal - cheap releases will sometimes just use whatever they find first, even if it's damaged, and a much better copy exists in an archive or collection somewhere.
Cheers,
David.
SebastianG has tried to manually create an SBR-like effect once:
http://www.hydrogenaudio.org/forums/index....showtopic=28258
http://www.hydrogenaudio.org/forums/index....showtopic=28258
This may not be the professional approach you are looking for, but try things like 'DFX' and 'Wow Thing' for WinAmp. They have a bandwidth extender.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.