Odd repeating low-energy distortion (?) in recorded LP

All,

I’m using Audacity 2.2.2 on MacOS (for recording using a MacBookPro) and Windows 10 for the audiofile editing (bigger monitor).

I have a digitized vinyl LP with odd low-energy (not low frequency) distortion (or whatever it is) in the audio file (this is the only audio file with this distortion of the six that were made from a boxed set of Isaac Stern playing violin concertos–so I don’t think the recording process was amiss). This “distortion” is clearly audible, but not visible in the waveform view of the audio. It is, however, unmistakable in the spectrum view of the audio:

This uploaded image is from the beginning of the LP and there is little music playing as yet. But even when the music volume increases, these transient distortions remain audible, though, of course, less intrusive. These distortion spikes no longer happen by 8 minutes into the audio file. Something was on the vinyl (not visible, I might add, I cleaned the LPs quite thoroughly, but not well enough to remove whatever caused this distortion).

Question then is, how can I diminish this distortion? I have used ClickRepair on this file, then in Audacity Click Removal and Noise Removal. All did something, but none did anything to this distortion.

Typically the distortion extends, as one can see in the image, across the frequency spectrum from 10Hz to 20kHz, though not always. Because of this my attempts to use band pass filters have produced no positive results (because this is not isolated to low-frequency or high-frequency audio). Knowing that the spectrum view maps frequency against time with the energy/amplitude of the signal color-coded, I was thinking that if I could reduce the energy/amplitude (?) of the audio at these moments the distortion might fall to inaudible levels. But how might I do that? What other kinds of filters or manipulation can one do?

I tried using Amplify with a negative value, e.g., -12, but for some reason this introduces artifacts - after applying the Effect the right or left edge of the selection has a short temporal region with an Increased amplitude. Very odd behavior. This occurs whether I select a little or a lot of the audio, so I conclude it must be the effect and that it should not be used as I’m attempting to use it. Using various types of EQ effects do not uniformly affect the entire selected region.

All comments and suggestions much appreciated.

Regards.

Can you post a short sound sample in WAV format?

WC

I tried using Amplify with a negative value, e.g., -12, but for some reason this introduces artifacts - after applying the Effect the right or left edge of the selection has a short temporal region with an Increased amplitude. Very odd behavior.

You can get a discontinuity if you don’t start/stop at the [u]zero-crossings[/u]. You might have to edit the left & right channels independently (split stereo tracks) if the zero-crossings don’t line-up. But make sure you are listening with your ears, not your eyes.

You might try [u]Wave Repair[/u] ($30 USD after free trial). It has several different repair options such as copying the just-preceding sound or the surrounding spectrum. These only work for very-short selections so it’s hard to tell if it will work for you. If you try Wave Repair, it also has a spectral view which is usually the best way to “find” the defect. Then, you can optionally zoom-in and switch to the waveform view to find the zero-crossings. And, you will have to find/select the defects one at a time so it can be very time consuming.

…so I don’t think the recording process was amiss).

It’s impossible to know unless you’ve got another copy of the same recording. Lots of things can go wrong during the recording & production process. Or, the record could have been damaged. If I’m seeing what I think I’m seeing, it’s happening at the same time in both channels which usually means something during recording/production.

Here you go – 10 seconds as the music begins (this is just after the graphic I posted earlier). This is Isaac Stern playing the 2nd movement Adagio of Brahms’ Violin Concerto with Eugene Ormandy and the Philadelphia Orchestra. If you import this in Audacity and shift to Spectrogram you should clearly see the distortion, or whatever it is.

Here you go. My second attempt at doing this. This is a 10-second clip right as the movement begins (just following the screenshot I posted earlier). Isaac Stern with Eugene Ormandy and the Philadephia Orchestra playing the Adagio 2nd movement of Brahms’ Violin Concerto.

After import into Audacity and shifting to the Spectrogram view, the distortion should be visible as you play the audio.

Well, as I was not using the waveform view, I certainly was not doing zero-crossing selections. I was using the Spectral Selection options for applying an effect on the audio.

You might try > [u]Wave Repair[/u] > ($30 USD after free trial). It has several different repair options such as copying the just-preceding sound or the surrounding spectrum. These only work for very-short selections so it’s hard to tell if it will work for you. If you try Wave Repair, it also has a spectral view which is usually the best way to “find” the defect. Then, you can optionally zoom-in and switch to the waveform view to find the zero-crossings. And, you will have to find/select the defects one at a time so it can be very time consuming.

I have downloaded the app. Will see what it can do. Thanks for the recommend.

I imported your file and amplified it with Audacity

a) the first thing to note is that you have a big spike at 0.938 seconds - could be dust, scratch or hair

b) I removed that and amplified again which revealed another scratch - which I also removed

c) further amplification lets you listen to the noise - it sounds somewgat mechanical - but possibly not regular enough to be from the deck - it could be an indication of a cold-pressing of the vinyl (it was getting too many cold pressings that was one of the things that drove me to CDs in 1984, much to the pleasure of my record shop in Charin Cross Road, as he no linger suffered a series of returns)

Let us know how you get on with DVDdoug’s advice (which I think is sound).

WC

Looking at the amplified form again - this time in Spectrogram view - I note that the noises appear to be fairly evenly spaced - which would suggest mechanical noise of some sort.

WC

Welp – it’s Click Repair!! Or so I think. On a lark I opened the original .wav from the digitization from the vinyl (using a Teac USB turntable directly into the MacBook USB port).

While the incessant pops and clicks are there, since they have not been “repaired,” the low-amplitude distortion is neither audible or visible in the spectrum view.

I’m going to experiment with Click Repair on this .wav files, and Audacit’s Click Remove, and Wave Repair and compare the results of three apps.

For funzies.

Gentlemen, thanks for your help. I believe I have identified the source of that strange “distortion” – Click Repair! I have the original .wav as digitized from the LP. I have two .wav run through Click Repair with default and slightly more aggressive settings. Both of these files exhibit that “sound”. I purchased Wave Repair and using various options to de-click, etc. have files that do not exhibit that “sound”. I can only conclude that this particular .wav file (which I add was created in exactly the same way as the other five from the three LPs) poses problems for the algorithm employed by the Click Repair app to remove clicks.

Interestingly, Wave Repair’s default click removal also produces a very, very bad result. I had to use non-standard options which do not identify and remove all clicks and pops. In fact I am unable to do this using Wave Repair or Audacity’s Click Remove effect. Neither app identifies and removes all clicks (Audacity settings were very aggressive, but many, many clicks remain).

So, I think, short answer, the LP itself has so many issues that the digitized audio is enough “non-standard” that is poses problems for “default” settings. But I have a decent, if lesser quality, .wav file and I can listen to it and enjoy it (after three days of playing with it in three different audio processing applications).

Thanks again for your help.

Fascinating - I used ClickRepair when I transcribed my (and my wife’s) LP collection - and I never encountered anything like this.

The settings I (mainly) used can be seen a the foot of the first post in this sticky thread: https://forum.audacityteam.org/t/click-pop-removal-clickrepair-software/1933/1

Many thanks for the feedback, much appreciated.

WC

One of the options in Click Repair allows you to listen to the difference between the processed and unprocessed audio while it is processing. This is useful to detect that the “repair” work really is removing noise and not music. You can also accomplish the same end by bringing the original and processed files into Audacity and subtracting the two. My guess is that Click Repair’s algorithm is getting tripped by something in the recording (or possibly in-audible noise from the turntable) and over-correcting. Listening to the difference signal might give clues as to what.

The third option that ClickRepair offers is to hear just the “noise” that is being removed (I suspect that this is what flynwill is referring to as the “difference signal”)

WC