Tom Fine wrote:

> Finally had a chance to listen to both MP3 files. One man's opinion 
> here, but the de-thumping was too radical to my ears. It creates 
> annoying gaps in the music and makes the surface noise that's left pump 
> so it's actually more noticeable. Did you try a notch filter at the 
> loudest frequency of the thumps? I'd bet there's not much necessary 
> music content down that low in the bandwidth of a tight notch. That 
> might make the thumps less noticeable than the gaps and pumping are with 
> the thumps removed.
> That said, it's pretty amazing you were able to paste that thing back 
> together and play it. The little ditty isn't half bad either.

As much as I hate to admit it, I've had CEDAR's de-thump and re-touch processes 
for some years but problems with how the earlier SADiE hardware and software 
platform handled things caused me to put it on the back burner.

With the new SADiE v5 software and related hardware, the problems were 
eliminated, and this example was my first effort at applying the processes 
under these conditions.  This was primarily for demonstrating the de-thump 
process and yes, further improvements could definitely be made by applying 
additional processes like CEDAR's NR-4.

Over that roughly 3 minute track, there were a little less than 200 individual 
de-thump edits and some were arguably better than others, but on the whole, the 
result shows what can be done with severely damaged source materials.

This process is manually intensive since it can't be automated and hence is an 
expensive proposition that demands a decision on whether the rarity of the 
source material justifies the cost of the work to be applied to it.  The 
example disc was one of a very few of these known to exist, so it was 
justifiable, certainly as a test subject.

CEDAR's de-thump process, as I said, is manual and involves marking the extent 
of the thump on the workstation oscillographic display, then telling CEDAR how 
many cycles of thump exist in the marked area.  De-thump then looks at material 
on either side of the marked area and constructs what it thinks would be the 
missing material in the marked area, and substitutes that in the space.  If the 
operator is wrong on the number of thump cycles, it will affect what it 
substitutes to one degree or another, and audibly to the critical listener.

All told however, the process is pretty remarkable and allows corrections to be 
made without the timeline being affected.

... Graham Newton

Audio Restoration by Graham Newton,
World class professional services applied to tape or phonograph records for
consumers and re-releases, featuring CEDAR's CAMBRIDGE processes.