Divine Art Record Company wrote: > The biggest difficulty I find with Dcart is locating small clicks on the > waveform display, as the zoom function seems to shift the display sideways. I'm still amazed that most manual click removal tools don't employ wavelet-derived frequency-time-space visualization to locate the small clicks left over after the automatic declicking/decrackling tools do their magic (certainly, the settings of such tools can be configured to effectively remove 100% of the impulses, but then, as Divine Art noted, the sound is horribly distorted as a result.) Small audible clicks which are hard to locate and visualize in the waveform stick out like sore thumbs in frequency-space (wavelet mode, not FFT mode) and can be trivially selected with the "click" (pun not intended) of a mouse, almost machine-gun like in fashion -- like shooting fish in a barrel, so to say. Pristine Sounds 2000 is a wavelet-based frequency-space editor I've experimented with for removing clicks in 78 material (I can manually remove 10 to 20 in a *minute*, almost machine-gun like fashion.) Unfortunately, it has the fatal flaw that even though it *exactly* determines the location of the small clicks, the tool it uses to reconstruct them does it in frequency-space (rather than just marking the exact time and directly editing the waveform by "cut and reconstruct" of the samples) -- thus each edit changes the waveform far outside the click area (well over a whopping 200 samples on both sides of the click!) If there are just one or two clicks in a tenth of a second timeframe, the repair is not too noticeable, but if there are several in such a time frame (and this is common in the more noisy 78s), then their combined removal results in a very noticeable unevenness ("pumping") in the background hiss, which is very annoying and almost as bad as the clicks themselves (the reconstructed music sounds great, though -- but variable/pumping background hiss *is* quite annoying.) "Cut and reconstruct" should be to simply cut out the several samples in the waveform itself where the click is located (which can be from several to a few dozen samples wide) and then reconstruct the "gap" based on high-order frequency analysis of BOTH sides of the gap thus assuring the reconstructed portion has the correct spectral content and time synchronization/harmonization -- no samples outside of the removed gap should be changed. There are algorithms to do this high-order reconstruction (i.e., Sonic Solutions employs such an algorithm, and manual click removal and reconstruction using SS produces superb results as George Morrow's exemplary 78 restoration work shows -- unfortunately SS does not have frequency-space visualization to quickly locate the clicks thus one has to laboriously scan the waveform with squinty eyes to try to find the click.) As an aside, what should *never* be done to "fill in the gap" is simple interpolation, either a straight line connecting the ends of the gap or simply zeroing out the samples -- this in effect replaces one type of noise with another type of noise, and I can audibly hear such "linear" fixes -- they are almost as annoying as the original click. Anyway, I've been in touch with the developer of Pristine Sounds, but at the time I talked with him, he was totally unmoved by my suggestion since the intention of his tool is to do complex frequency-space editing for advanced waveform processing (such as filtering out certain musical instruments, etc.) and he simply saw no "market" for optimizing it as I describe above for 78 source click removal. Nevertheless, PS 2000 demonstrates how such a killer manual 78 click removal tool would work. It would revolutionize how 78 material is restored, and allow one to fully declick the more noisy 78's in a fraction of the time it would take to do using the traditional method of trying to locate the clicks by scanning the waveform (as an aside, it means that one would be able to more gently use the automatic declickers and decracklers, thus resulting in even less distortion of the wanted sound.) I oftentimes find with full manual declicking that final denoising is not necessary, and even when it is called for, one can use it much more gently. I believe the key to getting great results out of older 78 restoration is complete (and properly done) *manual* declicking. Removing 300 to 500 clicks in a recording is not an issue when one can machine-gun them out in a matter of an hour or so -- wavelet-based frequency-space visualization allows one to do it. If the Diamond Cut people are reading this, hopefully it will spur them on to investigate the use of wavelet-based frequency-space visualization for manual click removal and reconstruction. I'd do it myself if I could, but my programming skills are zilch and I don't have enough resources, or time, at the moment to setup a venture and hire the programmers to build the tool (I'm working on other ventures at the moment.) Note, CoolEditPro and such tools do have frequency-space visualization modes, but to get them to work properly for manual click removal is difficult, and even when setup properly they do not even *remotely* come close to the convenience, speed and ease demonstrated by Pristine Sounds 2000 (and anyway, both their frequency-space visualization and wave reconstruction algoritms suck royally.) (As just noted, too bad PS 2000 does not do the right thing when it comes to the removal and reconstruction of clicks.) Jon