From: liz@poppyrecords.invalid.invalid   
      
   Liz Tuddenham  wrote:   
      
   [...]   
   > The basic principle is to split the sound spectrum into ten bands, each   
   > about an octave wide.  In each band a click or crackle is identified by   
   > comparing the vertical and horizontal vectors of the stylus movement -   
   > if there is more than a certain proportion of vertical movement, a   
   > switch opens and momentarily disconnects the signal.  The band is then   
   > filtered again, so that the harmonics due to switch clicks are removed   
   > and damped resonance in the filter covers any short gaps in the sound.   
   > All ten bands are then recombined to restore the original spectrum.   
   >   
   > My previous attempt (the MkII) used inductors in the resonant circuits   
   > and became so cumbersome and difficult to align that I abandoned it.   
   > The latest version uses 'state-variable' filters based on op-amps and a   
   > cascade filter configuration that makes the centre frequiencies far less   
   > critical and allows the use of ordinarly 5% tolerance components.  The   
   > overall pass band is within 2 dB of flat from 20 c/s to 20 Kc/s and the   
   > ripple due to filter overlap is less than 0.5 dB.   
      
   I've now been using the analogue computer for a few days on a variety of   
   different 78s (and 80s)l, so I have got a better feel for how it works.   
   On high frequency crackle it is very effective, the crackle disappears   
   and there is no noticeable deterioration of the sound (at least, to my   
   ears).  I am hearing details on my records that I never knew had been   
   recorded; it really 'cleans the dirt off the window'.   
      
   With the crackle removed, a much quieter lower frequency burble becomes   
   apparent.  It is difficult to say whether this was on the record anyway   
   and was masked by the crackle, but I suspect that isn't the case.  My   
   guess is that it is the remaining lower frequency components of the   
   crackle after tha high frequency components have been removed.  Crackle   
   is wide-band and wasn't subjected to the low-frequency cut that was   
   applied to the recorded sound, so after equalisation any residual low   
   frequency components of the crackle wil be more prominent.   
      
   Greatly increasing the overall senstivity of the de-clicking circuits   
   seems to remove the burble but mutilates the higher frequencies by   
   grossly over-processing them.  I may be able to overcome this by   
   selectively increasing the sensitivity of the lower frequency channels   
   whilst keeping the sensitivity to high frequencies unchanged.   
      
   There is plenty of experimenting still to be done but the results so far   
   are very promising.   
      
      
   --   
   ~ Liz Tuddenham ~   
   (Remove the ".invalid"s and add ".co.uk" to reply)   
   www.poppyrecords.co.uk   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)