[78-L] CEDAR

[Chris Zwarg]

At 21:08 09.10.2008, you wrote:
>Hi!
>
>I do not have the black box version, I have a computer (PC) with 
>nothing but the software version of CEDAR installed on it.
>
>Unlike my normal computer (as in the Apple Mac upon which I'm typing 
>right now), it IS in the same room as my gramophone, so what you say 
>doesn't apply to my situation.
>
>My current problem is that, just to get the hang of the thing, I'm 
>practising by accessing a rather noisy track from a CD, trying to 
>process it with CEDAR, through its three processes, but can hear very 
>little difference between the before/after versions. Indeed, only one 
>of the three processes seems to make any difference at all.
>
>The CEDAR people kindly popped in once, so I know that it's wired up correctly.
>
>Earl.

Is your sample track guaranteed to be flat EQ stereo (no RIAA, no graphic EQ, no treble-cut, no mono mixdown, no nothing)? 

I have but very little first-hand experience with CEDAR (used it only once when transferring a handful of sides at a Paris studio where everything was already setup properly so I didn't have to think much about the settings), so I'm doing a bit of educated guessing here, but I have heard from various users that CEDAR is very critical when it comes to pre-processed tracks. You will possibly get very little benefit from CEDAR if you use it on a commercially mastered CD track. Even slightly filtered noise, although still *sounding* like 78 noise, can have very different physical properties (esp. phase-distortion from analogue EQ circuitry, a tube preamp with a "soft" impulse response, etc.) that causes CEDAR to mis-recognize it as part of the music and thus to pass it through unchanged. The rather inefficient decrackling/denoising heard on various CEDAR-processed CDs on tracks that could not be found as master pressings but were taken from earlier analogue copies points to this being a CEDAR-typical problem.

You can re-equalize not-too-badly filtered recordings *manually* by analyzing the spectral distribution of the noise and creating a "mirror" EQ curve that counteracts the treble rolloff introduced by the original engineer and flattens out other dips and bumps in the response, but that is quite a tricky thing to do, and definitely not recommended for a first experiment. I have found that such manual EQ re-engineering (a rather time-consuming procedure BTW!) greatly improves the decrackling performance of the AlgorithmixPro suite I am using for my own remastering work; quite likely the situation with CEDAR is similar.

Chris