| Page 1 of 2 | 1 | 2 |
|
|
|
Finding clicks
I've recently been experimenting with using Audacity to deal with clicks in
digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf
wrote: I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Jim The click fixer in CoolEdit (many incarnations, and there is a shareware version among them) has a good reputation, and I've used it successfully. The product was subsequently bought by Adobe and has morphed into Audition - and become bloatware while abandoning the best features. You will need to adopt the dreaded windows to use it, I'm afraid. d |
Finding clicks
On Sun, 07 Sep 2014 14:29:07 +0100
Jim Lesurf wrote: I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Jim Many years ago the BBC (I think) developed a system that worked by playing a track *backwards*. The clicks still presented themselves as sharp edged pulses, while the music was a slowly rising signal. -- W J G |
Finding clicks
In article , Don Pearce
wrote: On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf wrote: [snip] The click fixer in CoolEdit (many incarnations, and there is a shareware version among them) has a good reputation, and I've used it successfully. The product was subsequently bought by Adobe and has morphed into Audition - and become bloatware while abandoning the best features. Does it *find* the clicks automatically? Fixing them is easy. You will need to adopt the dreaded windows to use it, I'm afraid. It would be easier to experiment with making use of the approach for auto-finding clicks if the above has one, and use my own software. Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf wrote:
BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Agreed. I've never found it worthwhile. However, on the few records I've digitised with Audacity and found click removal necessary, the clicks have all had enough amplitude to spot by eye once playback provided the approximate location. Maybe I've been lucky, but so far that have all been a single high amplitude wave cycle and have been simple to remove after zooming in far enough. -- martin@ | Martin Gregorie gregorie. | Essex, UK org | |
Finding clicks
In article 20140907144744.351a420b@debian,
Folderol wrote: On Sun, 07 Sep 2014 14:29:07 +0100 Jim Lesurf wrote: [snip] Many years ago the BBC (I think) developed a system that worked by playing a track *backwards*. The clicks still presented themselves as sharp edged pulses, while the music was a slowly rising signal. In effect, that's what I'm experimenting with at present. My first experiments scan though looking at the level. Then finding places where the peak levels drop a great deal in a short time. Thus picking up events with a sharply falling trailing edge. I've tried combining this with the peak level and crest factors. It works for the most obvious clicks. But not for the small ones whose size is *not* much bigger than the musical waveforms. So it shows clicks that are also clear to see with Audacity, but misses the smaller hard-to-see examples. So it is useful, but limited in value. Hence I'm thinking of trying the same approach as above, but to the first or second differential of the waveforms to change the relative scaling of quick events (with a lot of HF) to the surrounding music. BTW I also recall the old Hi Fi News cover showing some LP replay systems at their pressing factory. These looked strange because they were playing the LPs 'backward'. They were being used to look for faults (clicks) so went backwards for the same reason as above. Anyone buying EMI classical LPs at the time may not have been astonished that 2 out of 3 of the decks shown had a big red 'fault detected' light lit up. 8-] That seemed about right to me at the time. About 2/3rds of the EMI classical LPs I bought then had to be returned due to the added rifle shots! Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
In article ,
Don Pearce wrote: You will need to adopt the dreaded windows to use it, I'm afraid. Or grab a glass of WINE! I still run CoolEdit Pro 2.1 under WINE on Slackware, the last version from before Adobe stuck their bib in. -- --------------------------------------+------------------------------------ Mike Brown: mjb[-at-]signal11.org.uk | http://www.signal11.org.uk --- news://freenews.netfront.net/ - complaints: --- |
Finding clicks
On Sun, 07 Sep 2014 15:03:37 +0100, Jim Lesurf
wrote: In article , Don Pearce wrote: On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf wrote: [snip] The click fixer in CoolEdit (many incarnations, and there is a shareware version among them) has a good reputation, and I've used it successfully. The product was subsequently bought by Adobe and has morphed into Audition - and become bloatware while abandoning the best features. Does it *find* the clicks automatically? Fixing them is easy. You will need to adopt the dreaded windows to use it, I'm afraid. It would be easier to experiment with making use of the approach for auto-finding clicks if the above has one, and use my own software. Jim It has several modes. There's full auto where you just let it loose, a directed one where you can set thresholds and a manual one where you find the clicks, surround them with a pair of cursors, and some algorithm - spline or whatever - connects the two ends together. d |
Finding clicks
I can illustrate the real challenge here with an example.
http://jcgl.orpheusweb.co.uk/temp/ZoomCircled.png This shows the start of side 2 of an LP of Brahms 1st Piano Concerto (Barbirolli, Barenboim on EMI 1967) Its a lovely LP but has various various 'ticks' that are clearly audible in the quiet passages. The tick shown here at about 6.42 sec from the start is audible with the piano. Note the low modulation levels. The music is below about -25dB as recorded (0dBFS was about +17dBRIAA) and the tick is smaller in amplitude than the music. This one is relatively easy to find by ear-eye *but* you have to zoom the time and amplitude scales to be able to see it. If you don't the ripple at the bottom of the previous cycle looks like the cause because it sticks out of the displayed waveform, but it isn't. Other ticks are harder to find. But even this one seems a challenge to find by an 'automated' locator. Doing an automatic locator for loud bangs is easy. But then so is seeing them with Audacity! Question is if this kind of example can be detected by something of the kind I've mentioned. Ideally a program that generates a list of 'click candidates' that would find this but not be swamped with false positives. I suspect its almost impossible, but wonder what people think. Took me hours to do side 1! 8-] Its only something I'd do for 'special cases' where I really want to clean up as much as possible particularly enjoyable examples. ... and this is a 2 LP set. 8-] Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
In article , Don Pearce
wrote: On Sun, 07 Sep 2014 14:33:16 GMT, (Don Pearce) wrote: On Sun, 07 Sep 2014 15:03:37 +0100, Jim Lesurf wrote: Jim, I don't know if you can watch Youtube, but here's a short clip on the manual repair process. So far I've not bothered with YouTube TBH. However the problem I'm interested in is any algorithm for *finding* (and listing the positions of) clicks and ticks. The repair is the easy part, although I'd always do that manually so I can the waveform before and after. Sometimes a careless repair is worse that the original. :-) Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf wrote:
LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Washing the record would be a good start. |
Finding clicks
In article , mechanic
wrote: On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf wrote: LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Washing the record would be a good start. I experimented with that a while ago. I found that it may have helped in some cases, but in others it made no real difference. Many of the clicks or ticks on old discs seem to be due to scratches or wall damage.[1] Particularly for the 2nd hand discs. Overall, I decided the time was better spent on de-clicking. However I do carefully clean and prepare the discs before playing. Preener, Zerostat, and the brush of a dust bug. Its a tedious ritual, but OK given that from then on I'll use the cleaned digital file. I've also learned the habit of listening carefully a multiple of 1.8 sec after any tick... :-) Jim [1] Or for EMI discs in particular, scratches or dirt on the stamper for discs made back circa 1970s. -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
On Sun, 07 Sep 2014 16:44:21 +0100, Jim Lesurf
wrote: In article , Don Pearce wrote: On Sun, 07 Sep 2014 14:33:16 GMT, (Don Pearce) wrote: On Sun, 07 Sep 2014 15:03:37 +0100, Jim Lesurf wrote: Jim, I don't know if you can watch Youtube, but here's a short clip on the manual repair process. So far I've not bothered with YouTube TBH. However the problem I'm interested in is any algorithm for *finding* (and listing the positions of) clicks and ticks. The repair is the easy part, although I'd always do that manually so I can the waveform before and after. Sometimes a careless repair is worse that the original. :-) Jim I think it is worth a look if you can find a way. The repair is not the straightforward thing I thought it was. d |
Finding clicks
On 07/09/14 14:29, Jim Lesurf wrote:
I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... {...} There is a Linux application called Gramofile which claims to do what you want. I have used it in the past, but only for digitising recordings, not de-clicking them. The software hasn't been updated since 2001 but some distros still seem to include it. The website at http://www.opensourcepartners.nl/~costar/gramofile/ has some details of the algorithms - including the ones that didn't work. -- Dave |
Finding clicks
On Sun, 07 Sep 2014 14:29:07 +0100, Jim Lesurf
wrote: First, let me remind you of my own findings, which I've linked here several time befo http://www.macfh.co.uk/JavaJive/Audi...storation.html Note particularly the section about using a binary chop method to locate the clicks faster by eye, and the screen grabs combined with audio examples showing the clicks being removed. I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! Yes, no problem spotting these in the waveform, but not always easy to fix - for example, some scratches in piping records cannot be just interpolated, although this improves things significantly, often an audible bump remains. However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Yes, my site contains a sample of several of these in a very short section of music (it's actually Curved Air, for the afficionados), and the clarity that results from fixing them all. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. I've tried many such programs, and basically none of them really work. Either they miss too many genuine clicks and/or they mark too many false positives, to be worthwhile. BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. Yes, that sounds about right. Depending on condition, I used to reckon about one or two sides could be done in an evening's work. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... My page describes the software that I was using. It's really quite old now, but at the time was quite expensive. -- ================================================== ======= Please always reply to ng as the email in this post's header does not exist. Or use a contact address at: http://www.macfh.co.uk/JavaJive/JavaJive.html http://www.macfh.co.uk/Macfarlane/Macfarlane.html |
Finding clicks
On 07/09/2014 14:44, Don Pearce wrote:
The click fixer in CoolEdit (many incarnations, and there is a shareware version among them) has a good reputation, and I've used it successfully. The product was subsequently bought by Adobe and has morphed into Audition - and become bloatware while abandoning the best features. The click filter in GoldWave works excellently. It's not worth trying to do it by hand. But it is worth doing a difference between the raw and declicked files, which gives 20 minutes of silence and clicks that you can then add to other music to give the authentic LP sound. :-) -- Eiron. |
Finding clicks
On Sun, 07 Sep 2014 16:44:21 +0100, Jim Lesurf
wrote: In article , Don Pearce wrote: On Sun, 07 Sep 2014 14:33:16 GMT, (Don Pearce) wrote: On Sun, 07 Sep 2014 15:03:37 +0100, Jim Lesurf wrote: Jim, I don't know if you can watch Youtube, but here's a short clip on the manual repair process. So far I've not bothered with YouTube TBH. However the problem I'm interested in is any algorithm for *finding* (and listing the positions of) clicks and ticks. The repair is the easy part, although I'd always do that manually so I can the waveform before and after. Sometimes a careless repair is worse that the original. :-) I can certainly vouch for that effect when such declicking tools are used indiscriminately! Like Don, I've been using CoolEdit. In my case the Pro version 1.0, since before the turn of the century (from around 1997). I still use it to this day but I've let the audio processing jobs stagnate these last few years. :-( CoolEdit Pro does have some fairly comprehensive click and pop removal tools (and the usual noise elimination facilities as well, of course!) but these do need to be applied with some care. I suspect that there has been very little improvement over the past 17 years in this regard. One classic trap you can fall into is to use automatic click removal over a whole track or even a whole LP's worth when it contains brass wind instrumental sections since the autodetection tends to treat this type of waveform as 'clicks'. Indeed, this would make a nice test of the automated click removal. Just select a section with brass (trumpets or whatever) and apply the declicking process then compare the before and after. If you find the processed audio is rather lacking power in the trumpet section, that will just confirm the lack of progress I've surmised. Like any complex tool, an audio editing program like CoolEdit, will take some practice to get used to its limitations and to learn its strengths (and features/foibles). I initially used it to control and monitor the digitisation process whenever I transcribed vinyl and tape recordings into standard wav files. Using the older ISA soundcards (SB16, AWE64GOLD), the clip indicator on CoolEdit's meter proved a very useful feature to let me see at a glance whether any clipping had occurred whilst I'd had my back turned and so alert me to the need to check and decide whether I'd have to take another pass at a lower level (a very modest amount of real FSD clipping was acceptable but in many cases, this could have simply been the result of a loud pop or click in the source recording which, of course would be excised before attempting to normalise the digital capture). Unfortunately, this neat clip indicator feature was defeated by the initial crop of PCI soundcards[1] where the manufacturers it seemed, all to a man, had slavishly followed the 'reference design' offered by the sound chip maker and had managed to overlook the fact that jumpering the 6dB sensitivity reduction option on the ADC had the entirely foreseeable consequence of the input buffer amp clipping some 3dB below FSD due to reliance on the 5v line rather than a 10v (derived from the 12v rail) source as I suspect was the case with the older ISA cards. The aim of input noise reduction, whilst laudable enough in its own right, unfortunately was a major error in the design of these early PCI cards. I soon developed a strategy for dealing with such de-clicking/de-popping processing. Essentially, scan the whole waveform by eye for any loud obvious spikes, home in on them to ascertain what they actually are, select a narrow window bracketing the click and apply the declick filter, check the result was acceptable, undoing it if need be and try again with different parameters or else hand edit the samples or even simpler for a very short transient (around 1 ms or less), snip out that section entirely. Repeat and rinse until the whole waveform was cleared of major clicks and pops before applying normalisation (always, of course, auditioning such edits before moving onto the next). Usually, at that point, I'd save the wav file for further processing later as I deemed necessary (mostly a case of cleaning up the noise during the quieter passages, most often, during the intertrack pauses). This basic level of processing only took 5 to 10 minutes per album's worth so that much seemed worth doing straight away before moving onto the next album. Generally, I'd digitise several LP's worth per session or, in the case of the reel to reel recordings, either both sides of a 7 inch reel of LP tape (96 minutes per side @3 3/4 ips) or else one side of a 10 inch reel, up to 3 1/4 hours' worth. My aim was to archive the material into a state of digital preservation that was cleaner than if it had been auditioned directly. Even if I never got around to any further cleanup work, they'd still be good enough for playback. I could try improving it any time and, that is the problem with digital media, it can so easily induce a tendency to "Mañana". :-\ [1] For many years, I felt I was in a wilderness of 'consumers blind to this obvious deficit' / manufacturers who didn't give a flying ****. This deficit wasn't just limited to PCI soundcards alone, even the on-board MoBo sound chips suffered this affliction. The situation seems to have improved somewhat over recent years, at least as far as recently manufactured MoBos and USB adapters are concerned. I don't know how long it took before the industry finally spotted their "Schoolboy Howler" and corrected the design. I suspect it took something like a decade for them to finally sit up and take notice. -- J B Good |
Finding clicks
On Sun, 7 Sep 2014 14:47:44 +0100, Folderol
wrote: On Sun, 07 Sep 2014 14:29:07 +0100 Jim Lesurf wrote: I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Jim Many years ago the BBC (I think) developed a system that worked by playing a track *backwards*. The clicks still presented themselves as sharp edged pulses, while the music was a slowly rising signal. This echoes the method used to reduce the effects of group delay inherent in the analogue magnetic recording process where they'd dub the master onto a distribution copy tape with both machines running in reverse. It would make 'square waves' look like 'square waves' once more, rather than triangle waves. However, since both sound indistinguishable from each other, it was a moot point as to whether this was worth doing just for its own sake. However, if you needed to create disposable distribution copies and the machines could both run in reverse, this would prevent piling up more such group delay distortion, indeed, it reduced it even if the distortion wasn't entirely cancelled out so no bad thing. -- J B Good |
Finding clicks
On Sun, 07 Sep 2014 16:40:06 +0100, Jim Lesurf
wrote: I can illustrate the real challenge here with an example. http://jcgl.orpheusweb.co.uk/temp/ZoomCircled.png This shows the start of side 2 of an LP of Brahms 1st Piano Concerto (Barbirolli, Barenboim on EMI 1967) Its a lovely LP but has various various 'ticks' that are clearly audible in the quiet passages. The tick shown here at about 6.42 sec from the start is audible with the piano. Note the low modulation levels. The music is below about -25dB as recorded (0dBFS was about +17dBRIAA) and the tick is smaller in amplitude than the music. This one is relatively easy to find by ear-eye *but* you have to zoom the time and amplitude scales to be able to see it. If you don't the ripple at the bottom of the previous cycle looks like the cause because it sticks out of the displayed waveform, but it isn't. Other ticks are harder to find. But even this one seems a challenge to find by an 'automated' locator. Doing an automatic locator for loud bangs is easy. But then so is seeing them with Audacity! Question is if this kind of example can be detected by something of the kind I've mentioned. Ideally a program that generates a list of 'click candidates' that would find this but not be swamped with false positives. I suspect its almost impossible, but wonder what people think. IIRC (its been quite a while since I last did such processing), the declick function in CoolEdit Pro lists the number of clicks and pops it finds in a selection before allowing you to apply the removal process itself. If the click/pop count looks suspiciously on the high side you can try the effect and audition it afterward (I've assumed you would have already auditioned it beforehand). If there's no obvious improvement, or worse still, a degradation, you can simply undo the action. I do recall, however, that I tended to avoid auto repair and manually deal with the quieter sections where such noise would be a real intrusion (after dealing with the grosser, obvious by eye, clicks and pops). Here, when the small selected portion was largely the same low level amplitude, I found I could get away with using auto-declicking in most cases. Any recalcitrant clicks that escaped their well deserved fate I would home in on and manually edit the sound samples, if need be. Any other larger clicks hiding amongst the louder passages were usually undetectable by ear. In any case, I figured this would be a problem best left to my grandchildrens' progeny to solve. :-) There's only so much you can do before the benefit to effort ratio falls to a vanishingly small value where you begin to question your very existence. Hell! If I was content to listen to this stuff (warts and all) before, the result I've got so far aught to be more than enough to improve my listening pleasure. Enough already! Just give it a rest and be happy! \-) If you mess around enough with such processing, you'll find out what that last paragraph is all about soon enough. Took me hours to do side 1! 8-] Its only something I'd do for 'special cases' where I really want to clean up as much as possible particularly enjoyable examples. ... and this is a 2 LP set. 8-] I think you're already getting a notion of what I was going on about two paragraphs back. I think we all start off with an idealistic zeal for 'perfection' (at least that was true enough in my case) before the realism kicks in when the enormity of the task finally sinks in. Just deal with the most obvious defects and leave the rest for future generations to deal with when they might have access to better tools by which to complete the task. After all, you've already completed the most important task of digitising it in the first place even if you never process it any further than topping and tailing the tracks. -- J B Good |
Finding clicks
On 08/09/2014 03:16, Johny B Good wrote:
On Sun, 07 Sep 2014 16:44:21 +0100, Jim Lesurf wrote: In article , Don Pearce wrote: On Sun, 07 Sep 2014 14:33:16 GMT, (Don Pearce) wrote: On Sun, 07 Sep 2014 15:03:37 +0100, Jim Lesurf wrote: Jim, I don't know if you can watch Youtube, but here's a short clip on the manual repair process. So far I've not bothered with YouTube TBH. However the problem I'm interested in is any algorithm for *finding* (and listing the positions of) clicks and ticks. The repair is the easy part, although I'd always do that manually so I can the waveform before and after. Sometimes a careless repair is worse that the original. :-) snip I soon developed a strategy for dealing with such de-clicking/de-popping processing. Essentially, scan the whole waveform by eye for any loud obvious spikes, home in on them to ascertain what they actually are, select a narrow window bracketing the click and apply the declick filter, check the result was acceptable, undoing it if need be and try again with different parameters or else hand edit the samples or even simpler for a very short transient (around 1 ms or less), snip out that section entirely. Repeat and rinse until the whole waveform was cleared of major clicks and pops before applying normalisation (always, of course, auditioning such edits before moving onto the next). snip Yep, that's more or less what I've done. I'd also 'sew' the wave in something like Audition, and remove the peak. Only takes a couple of seconds. However, I'd only tend to get involved at that level with scratches. Which of course are considerably easier because they pop (ha) up at fixed intervals. For crackle and pop a decent clean, and live with what's left. Adds to the ambience :-) -- Cheers, Rob |
Finding clicks
In article , Johny B Good
wrote: On Sun, 07 Sep 2014 16:40:06 +0100, Jim Lesurf wrote: I do recall, however, that I tended to avoid auto repair So far I've avoided it entirely and intend to go on doing so. My interest at this point is wrt assessing how to generate a 'list' of locations which may have a click. Then examining them and making the decision, case by case, and using manual methods as and when I think needed for each example. I think you're already getting a notion of what I was going on about two paragraphs back. I think we all start off with an idealistic zeal for 'perfection' (at least that was true enough in my case) before the realism kicks in when the enormity of the task finally sinks in. Well I knew from the start that my default was to do no click editing at all unless I felt it was needed as a 'special case'. So far I've done hundreds of my old LPs and not bothered with any click removal. All I've done is some snipping of long lead ins or outs. Plus doing things like making mono files if the discs was mono. In practice most of my LPs are ones I bought decades ago. I would then return any with bad defects to the shop for a replacement. Then kept them carefully. So they are generally fairly free of annoying defects because of the effort I went to to avoid them back then! More recently I've been experimenting with buying some 2nd hand LPs. I found a source of cheap Jazz LPs and many of these are close to being free of audible clicks. Many are things like the old RCA 'Black and White' or 'Tribune' ones transferred in the 1970s from 78s. So they have lots of surface noise anyway. Hence no real need to de-click them at all. However I've also experimented with a few Classical LPs 2nd hand and found some that were as 'good as new'. But of course some others aren't. If they are poor and run-of-the-mill content I just write them off as a donation to charity. :-) But a *few* LPs have a special status from my POV. Three examples: Play Bach No 1. Teldec pressing. This is a *superb* recording. Makes a good test LP for the tracking ability of my V15 as well! In general no clicks or ticks. But it did have some. So I decided to clean them away. The result is very nice indeed. Barbirolli EMI LP of Sibelius Tone Poems. Superb recording and music. But lots of clicks. Since I love the sound of this I spent time removing all the clicks I could deal with with. Again, excellent results. Beethoven Triple Concerto. Oistrakh/Richter/Rostropovich. Like the Barbirolli. These are examples of digital transfers I expect to play often as they are so good. LPs I only play rarely seem less worth working on. Just deal with the most obvious defects and leave the rest for future generations to deal with when they might have access to better tools by which to complete the task. After all, you've already completed the most important task of digitising it in the first place even if you never process it any further than topping and tailing the tracks. Indeed. With most of my transfers I've adopted the view that I can 'fix them later if I really want to'. However being able to generate a *reliable* list of most of the audible click locations would speed up both the decision about how much work - if any - to do, and how long that work then takes. The loud clicks and bangs are easy to decide about. The problem with the smaller ticks is *finding* them to be able to deal with them. Fixing them is easy and quick *once* I've located them. So automating the location process is what would save time. In turn that would make it easier to do more discs, more thoroughly. But I know I'm asking what may be an impossible question. I have no problem with the 'fixing'. Just with *finding* the damn things when they can 'hide' in the music waveforms. Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
Well, not totally free, but I've had good results with Goldwave. I let it
loose on its passive setting first, and this gets rid of the most annoying ones. Then you have the issue you describe. What I tend to do now is to decide by temporarily doing a part I've copied out in several ways and see what sort of effect I get. If the record is sizzling, ie the clicks are many, trying to do the substitution usually results in an audibly worse effect such as a gurgle in the sound due to so many repeated samples. If its just an od one or a thud, sometimes using a more aggressive targeted process in that region can help. Not always though. Suck it an see. Thuds are the worst in my view, as no way to detect them. The very committed might look at the waveform and manually mess with it, but is it really worth it? I do also clean the record with warm water and fairy liquid in a Knowin cleaner and play them wet, as this reduces surface noise and puts a lot of the much in suspension. Make sure the stylus is cleaned though as a mess of dried crap tends to build up! Surface noise and rumble. This process really depends on how much its annoying. Often very light noise reduction can be a help though in quieter areas, it can make the sound have a watermark of the noise in it making it sound a little odd. I tend to only use this if its really needed in the quiet areas, and do not use it on the loud bits as it is masked. Fade outs show the adverse effects. Goldwave allows you to tweak the overlap and the Fourier transform parameters to make it as inaudible as possible on a test area. It can be quite time consuming of course but can make some remarkably good sounding results. I rejected audacity as it was not up to the job, but each to their own. The goldwave I use is the old version not the new multi track all singing and dancing one. Brian -- From the Sofa of Brian Gaff Reply address is active "Jim Lesurf" wrote in message ... I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
Snipped
The situation seems to have improved somewhat over recent years, at least as far as recently manufactured MoBos and USB adapters are concerned. I don't know how long it took before the industry finally spotted their "Schoolboy Howler" and corrected the design. I suspect it took something like a decade for them to finally sit up and take notice. FWIW.. This company specialised in 78 restoration and developed systems to do that some time ago now.. http://www.cedar-audio.com/ -- Tony Sayer |
Finding clicks
Jim Lesurf wrote:
1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. Try AFDeClick http://www.andreas-flucke.homepage.t...index_eng.html It might help to understand what (and what not) it does ... http://www.andreas-flucke.homepage.t...ick/about.html Anyway, like I mentioned above, just try. Keep on rocking. Muck |
Finding clicks
On 2014-09-07, Jim Lesurf wrote:
I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Apparently my response never got posted. So try again. Clicks are caused by defects on the record. beween the cartridge and the output is the RIAA filter, which is essentially an integrator (actually and integrator followed by a high frequency single pole boost). This means that a sudden displacement of the stylus back and forth, gets converted into what is essentially a step funtion-- ie the effect of the click gets distributed in time. Thus what one wants to do is to apply an inverse RIAA curve to the output and look at the the result. The clicks should now be far more localised-- ie their effect should be far more concentrated, and removeable. Ie, apply the inverse RIAA (essentially a differentiation followed by a bass boost-- Ie, flat to 500 Hz, then a fall at 6dB/octave to 2000 Hz, and then flat again above that if I remember the RIAA correctly.) note that this means that there is a total of about 50dB change from low freq to high, which means that you have to be using at least 24bit, and preferaqbley 32 bit processing of the signal in order not to get clipping, or introduce excess noise. So use sox say to impliment the inverse RIAA, then use audacity to look for those spikes, and remove them, then use the RIAA on the result. Note that one could just take the derivative, but that would still leave a finite spreading due to the treble/bass boost. Jim |
Finding clicks
Sometimes playing it backwards for detection actually works better than
forwards. Brian -- From the Sofa of Brian Gaff Reply address is active "William Unruh" wrote in message ... On 2014-09-07, Jim Lesurf wrote: I've recently been experimenting with using Audacity to deal with clicks in digital recordings made from old LPs. I suspect I'm not the first to do this or encounter the following! Hence I'd be interested in feedback on what follows... LPs in very good condition only have a few clicks, and these can be easy enough to find and fix. Particulary if they are loud 'rifle shots' that stick out clearly on something like Audacity's waveform plots! However other LPs can have many many clicks per LP side. This can make finding and fixing most of them fairly time-consuming. In particular when a small 'tick' is hiding as a small alternation to a larger and complex audio waveform. It becomes a bit like looking for a sapling in a forest! For some old classical LPs there may be lots of these which are audible as the music can have long low-level sections, meaning that clicks it would be impossible to hear with loud Jazz, say, show up against quiet classical. Because of this I've been experimenting with ways to scan a wave file looking for clicks. Using tricks like looking at the first or second derivative of the waveforms which appear rise and fall quckly to emphasise short sharp clicks out of the steady music background. However I'm wondering about two things. 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. 2) To what extent this is simply a waste of effort beyond finding the most obvious clicks. i.e. That there isn't a simple and reliable algorithm for this and it ends up being quicker and better to use ears and eyes and Audacity. So far I have the impression that (2) comes into force pretty quickly as the clicks vanish into the waveforms. But I thought I'd ask as I suspect others have explored this already. :-) BTW At present simply using ear/eye/Audacity I seem to find that the 'hard cases' where I'm searching for many tiny 'ticks' can mean about 0.1 rate working. i.e. About 200 - 300 mins of work per LP side for classical if I really want to clear even the faintest ticks I hear. Fortunately, LPs that tend to spend most of the time at higher levels are much quicker as the music drowns out the smaller ticks. BTW2 Having experimented I haven't found the declicking 'effect' of Audacity to be much use. I've just been using the 'repair' instead. But maybe I'm missing something here... Apparently my response never got posted. So try again. Clicks are caused by defects on the record. beween the cartridge and the output is the RIAA filter, which is essentially an integrator (actually and integrator followed by a high frequency single pole boost). This means that a sudden displacement of the stylus back and forth, gets converted into what is essentially a step funtion-- ie the effect of the click gets distributed in time. Thus what one wants to do is to apply an inverse RIAA curve to the output and look at the the result. The clicks should now be far more localised-- ie their effect should be far more concentrated, and removeable. Ie, apply the inverse RIAA (essentially a differentiation followed by a bass boost-- Ie, flat to 500 Hz, then a fall at 6dB/octave to 2000 Hz, and then flat again above that if I remember the RIAA correctly.) note that this means that there is a total of about 50dB change from low freq to high, which means that you have to be using at least 24bit, and preferaqbley 32 bit processing of the signal in order not to get clipping, or introduce excess noise. So use sox say to impliment the inverse RIAA, then use audacity to look for those spikes, and remove them, then use the RIAA on the result. Note that one could just take the derivative, but that would still leave a finite spreading due to the treble/bass boost. Jim |
Finding clicks
On Wed, 10 Sep 2014 09:19:44 +0100, "Brian Gaff"
wrote: Sometimes playing it backwards for detection actually works better than forwards. Brian In the digital world, the idea of playing in any direction has no meaning - you don't detect steep edges that way, you differentiate and look at amplitude. d |
Finding clicks
In article , William Unruh
wrote: On 2014-09-07, Jim Lesurf wrote: [snip] I understand the argument about RIAA being quasi-integrating, etc. Its one of the reasons behind my thinking that looking at the first or second derviative would help. So use sox say to impliment the inverse RIAA, then use audacity to look for those spikes, and remove them, then use the RIAA on the result. Note that one could just take the derivative, but that would still leave a finite spreading due to the treble/bass boost. Wary of that because 'mending' a differential waveform might lead to a dc offset problem when you re-integrate the result. So I'd use a dx/dt or d2x/d2t to *find* and list click locations. But do any editing on the actual audio file recorded using RIAA. Avoids the problems of dealing with the real response curve being rather complicated. Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
In article , Muck Krieger
wrote: Jim Lesurf wrote: 1) Anyone know of decent free software that already does something like this well and can list a good set of 'click candidate' times in a wave file. i.e. low levels of 'misses' and 'false alarms' even with classical music. Try AFDeClick http://www.andreas-flucke.homepage.t...index_eng.html It might help to understand what (and what not) it does ... http://www.andreas-flucke.homepage.t...ick/about.html Anyway, like I mentioned above, just try. The above says it uses an 'algorithm' but gives no details of the algorithm itself. Its also just a DOS exc when downloaded. No added info on the program itself. When processing data I like to know the details of the process. Have a dislike of 'black boxes'... :-) I also use RO and Linux. Gave up any Windows use years ago, and don't bother with Wine, etc. Life's too short. :-) It also seems to just 'fix the clicks' rather than generate a list of candidate instants for me to examine and decide upon. FWIW Even the simple 'click lister' I wrote as a quick experiment seems to find the main click events without too much trouble. That's just using rapid falls in peak level. The challenge is smaller ticks which hide in the audio. So, when I can, I'll have a go at a program using differentiation first and see how I get on. The files I record are 96k/24 BTW. Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
In message , Don Pearce
writes On Wed, 10 Sep 2014 09:19:44 +0100, "Brian Gaff" wrote: Sometimes playing it backwards for detection actually works better than forwards. Brian In the digital world, the idea of playing in any direction has no meaning - you don't detect steep edges that way, you differentiate and look at amplitude. Isn't that the analogue way too? -- Ian |
Finding clicks
On Wed, 10 Sep 2014 09:19:44 +0100, "Brian Gaff"
wrote: Sometimes playing it backwards for detection actually works better than forwards. Quite likely to be true in most cases. Although CoolEdit Pro will allow you to reverse the wav file to facilitate this, it's not clear as to whether or not this would improve matters (the declicking algorithm may be using this method in the first place - possibly even a combination of bacwards and forwards scanning). However, since I can't recall seeing a description of the declicking algorithm using such a tactic, it'll certainly be worth trying out. -- J B Good |
Finding clicks
|
Finding clicks
On 07/09/14 16:44, Jim Lesurf wrote:
In article , Don Pearce wrote: On Sun, 07 Sep 2014 14:33:16 GMT, (Don Pearce) wrote: On Sun, 07 Sep 2014 15:03:37 +0100, Jim Lesurf wrote: Jim, I don't know if you can watch Youtube, but here's a short clip on the manual repair process. So far I've not bothered with YouTube TBH. However the problem I'm interested in is any algorithm for *finding* (and listing the positions of) clicks and ticks. The repair is the easy part, although I'd always do that manually so I can the waveform before and after. Sometimes a careless repair is worse that the original. :-) FWIW I've spent many happy hours recently doing just this. Like you I suspect I trust only my ear as the final arbiter of a click and auto click removers not at all. Initially I tried overwriting, actually with EZpatch rather than Repair, but about half the time that just turned a click into a thud. So I just redrew the damaged bits manually. I found you get to recognise the shapes that need repair and on a couple of really noisy records I didn't bother listening, just scrolled through looking for the shapes - fortunately just a couple of records. I found a feature of Audacity is that if you scroll forward at too much zoom it keeps on scrolling for an indeterminate distance, you need to be 3 zooms out from seeing the individual points (or sometimes 4, depending). But it sure is a labour of love, I haven't spent quite so long in proportion as you but then my standards may be lower. Still I've resurrected stuff I haven't played for decades, recordings of Menuhin, David Munrow and my ancient trad jazz (mono, some 10" - yes 33 1/3 vinyl). Brilliant. -- Dick Georgeson Whenever you find that you are on the side of the majority, it is time to reform. -- Mark Twain |
Finding clicks
In article , Johny B Good
wrote: On Wed, 10 Sep 2014 09:19:44 +0100, "Brian Gaff" wrote: Sometimes playing it backwards for detection actually works better than forwards. Quite likely to be true in most cases. Although CoolEdit Pro will allow you to reverse the wav file to facilitate this, it's not clear as to whether or not this would improve matters (the declicking algorithm may be using this method in the first place - possibly even a combination of bacwards and forwards scanning). In essence 'looking in the rear view mirror' is what my simple method does. It fined the peak value in each successive block of data of a few tens of milliseconds and checks to see if the 'current' block is much smaller than the one before it. Sign of a rapidly ending 'event'. But as I'd expected, it works for the obvious loud clicks, but can easily be confused unless you set the 'trigger' levels high. In effect it doesn't do much better than using Audacity and looking for spikes that stick well clear of the rest of the waveform. It is useful as a quick estimate of how many loud bangs are present, though. Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
In article , rmg
wrote: On 07/09/14 16:44, Jim Lesurf wrote: But it sure is a labour of love, I haven't spent quite so long in proportion as you but then my standards may be lower. Well for almost all LPs I don't bother at all. For some I just take out the loud explosions. Only for a few special items am I willing to spend hours doing a 'search and desroy' of as many clicks as I can cope with! Too much leads to madness... oops, sorry, too late. 8-] Still I've resurrected stuff I haven't played for decades, recordings of Menuhin, David Munrow and my ancient trad jazz (mono, some 10" - yes 33 1/3 vinyl). Brilliant. Yes. I find Mono Jazz actually often works well. Many of the smaller 'ticks' are on just one groove wall. So when I convert to mono the click to music ratio improves by about 5 or 6 dB. Hence often just removing the loud bangs and converting to mono helps a lot. Alas, stereo recordings at lower levels are more exposed to any small problems. :-/ BTW I've been told that my box of The Beatles Mono LPs should reach me tomorrow. I'm hoping that they will be in better condition than the LPs EMI used to churn out in the 1960s! Should be for the money! Rarely buy 'new' LPs, but in this case I weakened and made an exception. Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
ISTR that when I looked into this, about a decade ago now, I found
that at least some of the algorithms are based on Kalman filters: http://www.cs.unc.edu/~welch/kalman/index.html http://www.cs.unc.edu/~welch/media/pdf/kalman_intro.pdf On Wed, 10 Sep 2014 13:35:17 +0100, Johny B Good wrote: However, since I can't recall seeing a description of the declicking algorithm using such a tactic, it'll certainly be worth trying out. -- ================================================== ======= Please always reply to ng as the email in this post's header does not exist. Or use a contact address at: http://www.macfh.co.uk/JavaJive/JavaJive.html http://www.macfh.co.uk/Macfarlane/Macfarlane.html |
Finding clicks
On 2014-09-10, Jim Lesurf wrote:
In article , William Unruh wrote: On 2014-09-07, Jim Lesurf wrote: [snip] I understand the argument about RIAA being quasi-integrating, etc. Its one of the reasons behind my thinking that looking at the first or second derviative would help. So use sox say to impliment the inverse RIAA, then use audacity to look for those spikes, and remove them, then use the RIAA on the result. Note that one could just take the derivative, but that would still leave a finite spreading due to the treble/bass boost. Wary of that because 'mending' a differential waveform might lead to a dc offset problem when you re-integrate the result. So I'd use a dx/dt or You could always put in a 50 or 30 Hz cutoff in the RIAA curve. Some advocated that anyway. But those clicks put in a DC bias in the first place. d2x/d2t to *find* and list click locations. But do any editing on the actual audio file recorded using RIAA. Avoids the problems of dealing with the real response curve being rather complicated. But since the click has been spread out all over hells half acre by RIAA, that "fixing" either leaves loads of artifacts or also "fixes" a bunch of the real signal as well. The pre RIAA is the place to fix it. Jim |
Finding clicks
In article , William Unruh
wrote: On 2014-09-10, Jim Lesurf wrote: In article , William Unruh So use sox say to impliment the inverse RIAA, then use audacity to look for those spikes, and remove them, then use the RIAA on the result. Note that one could just take the derivative, but that would still leave a finite spreading due to the treble/bass boost. Wary of that because 'mending' a differential waveform might lead to a dc offset problem when you re-integrate the result. So I'd use a dx/dt or You could always put in a 50 or 30 Hz cutoff in the RIAA curve. Some advocated that anyway. But those clicks put in a DC bias in the first place. Erm. The mechanics and the RIAA don't pass down to dc. So what happens is a decaying offset. The results shapes are pretty clear. In my case I'm using a V15 in an old arm that has more mass than ideal. So the peak and fall at LF is at very low frequency, but not dc. d2x/d2t to *find* and list click locations. But do any editing on the actual audio file recorded using RIAA. Avoids the problems of dealing with the real response curve being rather complicated. But since the click has been spread out all over hells half acre by RIAA, that "fixing" either leaves loads of artifacts or also "fixes" a bunch of the real signal as well. The pre RIAA is the place to fix it. Again, looking at the shapes I can see the effects. Adding the 'fix' just puts in a plausible smooth interpolation anyway. To deal with it in the way you suggest would require an accurate 'de-riaa' that also precisely deals with the stylus and arm responses over the full range down to almost dc. i.e. much lower than 10Hz or so. Even measuring that isn't trivial. And it differs in the vertical and horizontal planes anyway. So you'd also have to convert the L and R to V and H first. So simply applying a reverse riaa preamp curve won't in practice be much better than a simple integrator if your concern is LF spread. Given that the mends I've made so far are generally inaudible except for severe events that clearly lose the waveform anyway. I'm happy enough despite the nice theory for preferring de-riaa. Life's too short. :-) Jim -- Please use the address on the audiomisc page if you wish to email me. Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html Audio Misc http://www.audiomisc.co.uk/index.html |
Finding clicks
On 2014-09-11, Jim Lesurf wrote:
In article , William Unruh wrote: On 2014-09-10, Jim Lesurf wrote: In article , William Unruh So use sox say to impliment the inverse RIAA, then use audacity to look for those spikes, and remove them, then use the RIAA on the result. Note that one could just take the derivative, but that would still leave a finite spreading due to the treble/bass boost. Wary of that because 'mending' a differential waveform might lead to a dc offset problem when you re-integrate the result. So I'd use a dx/dt or You could always put in a 50 or 30 Hz cutoff in the RIAA curve. Some advocated that anyway. But those clicks put in a DC bias in the first place. Erm. The mechanics and the RIAA don't pass down to dc. So what happens is a decaying offset. The results shapes are pretty clear. In my case I'm using a V15 in an old arm that has more mass than ideal. So the peak and fall at LF is at very low frequency, but not dc. Actually, the RIAA curve does go to DC. There was a very controvertial proposal to put another zero/pole at 50Hz to comensate for the cutter low freq resonance but the problem is that the cutters all have different resonances to for some it would make thing worse. Anyway, since your speakers cannot hear 30Hz, you could put it there-- the main thing is that the unRIAa and RIAA filter be complementary. If the spike from the record is before the RIAA then the RIAA filter will have spread it out all over the place, and "fixing" it after the filter will leave all that spread out residual in place. d2x/d2t to *find* and list click locations. But do any editing on the actual audio file recorded using RIAA. Avoids the problems of dealing with the real response curve being rather complicated. But since the click has been spread out all over hells half acre by RIAA, that "fixing" either leaves loads of artifacts or also "fixes" a bunch of the real signal as well. The pre RIAA is the place to fix it. Again, looking at the shapes I can see the effects. Adding the 'fix' just puts in a plausible smooth interpolation anyway. To deal with it in the way you suggest would require an accurate 'de-riaa' that also precisely deals with the stylus and arm responses over the full range down to almost dc. i.e. much lower than 10Hz or so. Even measuring that isn't trivial. And it differs in the vertical and horizontal planes anyway. So you'd also have to convert the L and R to V and H first. So simply applying a reverse riaa preamp curve won't in practice be much better than a simple integrator if your concern is LF spread. Agreed that the curve is problematic below 50Hz. But even at 200Hz the sound is spread out over more than 200 time pixels (400 for 96K sampling). That's a lot. Given that the mends I've made so far are generally inaudible except for severe events that clearly lose the waveform anyway. I'm happy enough despite the nice theory for preferring de-riaa. Life's too short. :-) It is cheap enough to try it. I agree that it may not be an improvement. Even just a differentiator would be a help (differentiate, fix, integrate) except you really have to make sure you have enough dynamic range. Since that is 10 octaves or 60dB emphasis of highs over lows, which is even larger than the 48dB of the RIAA curve ( which would fit in another 8 bits that sox stores stuff at.) Jim |
| All times are GMT. The time now is 07:00 AM. |
|
Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.
SEO by vBSEO 3.0.0
Copyright ©2004-2006 AudioBanter.co.uk