Download fault

[Phileas]

I recently downloaded a Hyperion album, Martinu:Complete music for violin & orchestra, Vol. 3 and discovered a glitch (2 clicks) on one of the tracks. I tried downloading again but the clicks were still there so I contacted Hyperion and received this reply:

I've just checked this by listening to the Rhapsody-Concerto on one of our CD library copies and I'm afraid the two clicks are also on that, in exactly the same place. This means they must be on the master and are therefore unavoidable

I found this surprising.

[johnb]

Very strange. Depending on the nature of the clicks it should be very simple indeed to remove them with decent audio editing software. For Hyperion it would be a trivial matter to correct those glitches (with their skilled audio engineers and professional software). They must have somehow slipped through their quality control.

[amateur51]

How irritating!

Has anyone checked the score?

[Phileas]

Some more info from Hyperion:

"All the Martinů violin concertos are Czech recordings and weren't actually engineered by Hyperion"

[Phileas]

A couple of users kindly offered to fix the clicks on the faulty track. I declined the first offer because I wanted to have a go at fixing it myself using Audacity but soon gave up and accepted the offer from the second user who happens to own the CD. It's not perfect but it's a worthwhile improvement.

However, I couldn't resist having another play with Audacity. Eventually I managed to find the discontinuities causing the clicks. It looks like two very short sections may be missing. I smoothed them out using the repair tool.

Just for fun, I compared the before and after spectra for the 0.05secs encompassing the first click. The pre-mod spectrum has a lot more higher frequency content than the post mod spectrum, which is sort of what I expected from my understanding of Fourier analysis, but it was still quite surprising (before is on the right):

[Dave2002]

re msg 5

I wonder if there was an editing issue when the track was being mastered for CD. I have tried myself to edit audio files, and it can be surprisingly hard to insert or remove sections without introducing audible clicks at the transition point. I think professionals can do this more easily - others around here may have found ways to do it smoothly.

One I tried was of a recording which somehow I'd managed to record in two versions, and part of one was missing. I did get them patched together, but it was very hard to avoid a transitional glitch - and in that case there may have been an additional issue as I think one patch was mono while the overall piece was in stereo.

Another way to ameliorate the clicks could be to apply a strong low pass filter for the short fragment containing the clicks - but it would probably not be perfect.
Also the clicks might be more prominent on one side, so could reduce to mono, or simply use one channel for that fragment.

The noise removal tool in audacity might just work, if you can use the clicks themselves to make a noise profile. You would not then wish to apply that noise profile to the whole of the recording, as that would reduce quality, but perhaps you could apply the profile (say) to a 5-10 second section surrounding the clicks. You can make a noise profile with as little as 0.5 sec, maybe even less. I've not tried that myself, though I have significantly reduced low frequency noises.

You will indeed notice differences in the spectra if you apply various different techniques using audacity. Depending on your hearing and equipment it may be a good idea to only limit the HF loss to sections around the problem points, though in honesty many of us who are older and not asthmatic will probably not be able to hear the HF anyway.

Unless you are really interested in experimenting, it might be easiest to just listen to the music, and use some other means of "filtering" out the clicks, such as drinking beer, wine or champagne during your listening session, and gaining possibly heightened enjoyment that way.

[Phileas]

I'm not that bothered about the clicks, I just can't resist playing around with the software. I'll get bored soon.

Looking at the step in the graph, it almost looks as though one side has been shifted vertically relative to the other.

[johnb]

It looks to me like the result of a really careless edit.

There are standard methods of avoiding such things, e.g. editing at zero crossings, using cross fades (the digital equivalent of an angled tape join), etc.