What Happens in Noise Reduction for a Record (May 2017)
Disclaimer: There is ongoing discussion of vinyl vs. digital: their properties, qualities, advantages, disadvantages, and of course preferences. I acknowledge each perspective and do not wish to take sides in the debate. I do plan on addressing the topic in a future posting.
That noise accrues on vinyl LPs and 45s and on shellac 78s is a fact of life. Dragging a diamond or steel stylus through the grooves of softer material will eventually wear the grooves, thereby affecting the record’s fidelity. Dust, fingerprints, accidentally dropping or bumping the tonearm, heat, etc. can increase the damage. Converting a record to digital format will, unfortunately, faithfully capture such imperfections along with the music.
After one becomes accustomed to the clarity in digital formats, record noise can be unexpected, therefore more noticeable, and likely distracting. Unless one is emotionally attached to a record’s pops or lead-in “fry noise” (don’t laugh—it happens!), the converted recording will benefit significantly from noise reduction, as this yields a cleaner and more enjoyable sound. Here’s how I approach noise reduction.
My “prime directive” is to preserve as much music as possible while removing as much noise as possible. I strive to achieve a satisfactory balance between these goals, but if this isn’t possible, music preservation takes precedence. Aggressive noise removal can leave the music sounding dull, flat, tinny, muddy, metallic—the list of negative adjectives goes on and on.
Second, when all tracks on an album have similar acoustic properties, I can apply noise reduction algorithms to all the tracks at the same time. But if some stand out as different, I extract those tracks and perform noise reduction on them separately. For example, I worked on an album of a cappella choral music with a couple of tracks that included a small drum. As drum beats and pops share acoustical properties, what worked for most of the album removed too much of the drum is these tracks, and so I addressed them individually.
Third, I cannot allow myself to fall victim to ego involvement. If, after finishing noise reduction on an album or track, I don’t like what it sounds like, I must remain prepared to start over from scratch, varying the order in which I apply the noise reduction algorithms as well as changing their parameters. I’ve done so several times, and I’m happy to report that the second (or third or fourth!) effort proved successful in each case.
There are different algorithms for removing clicks, crackle, pervasive noise, and hum. In the rare occasions where I encounter hum (usually in recordings made on personal tape recorders), I address that first. Otherwise, I generally tackle clicks and crackle before pervasive noise, since, in my experience, removing pervasive noise first makes it more difficult for software to detect clicks and crackle. I say “generally” because the result is not always consistent, and I sometimes end up reversing the order.
Each of the noise reduction algorithms has a collection of knobs, sliders, and check boxes for setting processing parameters. I won’t dwell on them except for the two important ones that all of the algorithms share. The “Bypass” check box allows me to turn the noise reduction on and off while I’m previewing what it will do to the recording. Toggling this check box on and off provides a good sense of what the music will sound like with and without the noise reduction, as well as how much and what type of noise will be eliminated.
The second allows me to preview just the noise to be removed. This provides a better indication of the amount and type of noise to be removed, but it also reveals how much music has been included in the noise. Detecting any music among the noise is an indication that I need to back off on one or more of the reduction parameters.
There’s an undocumented feature of this control that I discovered by accident when I first started using my current software. I inadvertently left the “listen to the noise” box checked when I clicked the button to process the file, and the algorithm dutifully produced a file by throwing away the music and retaining just the noise. (I have never been so grateful for the Undo button!) This turned out to be a blessing, however. I can listen to the noise-only file, jumping around to different places in the file, especially where the waveform in one area is visually different from the rest of the file. This review of the noise allows me to ensure that I’ve got the right balance between music preservation and noise removal.
I sometimes apply equalization, pretty much exclusively when dealing with 78s or tapes made on a personal recorder. This is addressed in depth in the article on Equalization.
​
Back to Paul's Blog & Contents
​