Creative Noise Reduction [part 2 • denoising]
After nailing down the new running order, I set about re-mastering the last tracks from vinyl. It was a not overwhelming task with most of the vinyl purchased new and rather pristine. With one exception. In my early 2000s universe where I had been buying up the few China Crisis record I needed for this project, the hardest thing to obtain was the “African + White Steve Proctor Remix” 12 inch. I had bought the CD single of the same title, but in 1990, format chart rigging was still rampant, so this led to the 12″ remix only being on the vinyl, with the more pedestrian 7″ remix on the CD. Since I probably bought this in 2002-4, I was not yet in the Discogs.com world, so that meant that I had to haunt eBay or GEMM to finally buy this record.
When the day came to digitize it, I was aghast to hear how much noise was inherent in the disc. This was compounded by the decision made by Mr. Proctor, to have the long, three minute buildup for the track be at a much lower threshold of loudness. Let’s take a look at that wave now.
As the wave shows, the delicate music is at a low level of dB until at roughly the 2:30 mark [see third group of red lines], where the volume fades up over a 0:30 period, after which the song fairly erupts into bold sonic life. There’s nothing artistically wrong with this, in fact, it’s what makes this remix so magnificent. You may listen to the original here. Proctor thinks highly of this mix and so do I. In today’s modern-a-go-go world, you may now listen to this on Soundcloud, but a decade ago, it was vinyl or nothing. And my vinyl had… problems.
Every second of that long intro had crackle and pops far beyond the ability of the interpolate filter in Sound Studio to cope. I would still be denoising this manually in 2013, and when the filter is applied in rapid succession, the negative effects on the sound file are palpable. It wasn’t designed for this level of abuse.
Listen to a typical second of this raw track here.
The dreaded Rice Krispies® of audio restoration! How could I cope with this? It looked like time to use the much dreaded broadband noise reduction capabilities of Sound Soap 2. With Sound Soap 2, I could finesse the NR until I literally no longer heard the massive amount of crackle inherent in the record. But at a cost. Always at a cost. So I fired up Peak and loaded the raw file, and opened the Sound Soap filter and began to try different settings to remove the noise.
In Sound Soap [the small window at upper left], the round dials control the NR attack on the wave. If you look at the center portal, you will see the noise in red at left be wiped off of the scrolling wave as it crisses the blue line in the center of the portal. As you monitor, you will hear the live results in your headphones. It was possible to really remove 95% of the noise but the effect on the high end of the remaining music was profound. Whenever I use NR like this it makes me die a little inside [unlike Jon Astley, but that’s another post!]. The rest of the song after the fadeup now sounded very compromised. The effect of the NR on the long intro was not terrible in and of itself, since there was very light sound there to begin with. Just synth flutes, a touch of percussion and some airy Fairlight patches. After the NR was applied, it really cleaned up the music and any artifacts imparted by the NR were congruent with the sound that Proctor was aiming for in any case.
Listen to the same second of the remix after NR was applied in Sound Soap 2.
I couldn’t just splice the clean intro into the rest of the file. It would sound abrupt and crude. That’s when I realized that my salvation lay in Apple’s free-with-every-Mac Garage Band DAW that allowed multitrack mixing. So I applied NR to the file and saved it as a separate file. I fired up Garage Band and discerned how to load files into it. I then quickly saw how to apply volume fades to the files and with both the raw and clean version of the file loaded, applied a cross fade at the 2:30 to 3:00 mark in the song. The delicate intro was unmolested by sound, and the full bore track that erupted at the 3:00 mark, was not beset by much in the way of noise compared to the quiet intro where the surface noise was battling the music for sonic dominance.
Due to the nature of this arrangement, I was able to have my cake and eat it too by applying heavy NR to the intro, where it did the least amount of damage, and leaving the much louder portion of the record untouched by Sound Soap 2. Fortunately for me, the louder part of the record either had much less noise, or the noise it had was fully masked by the [much busier] music; it’s been many years since doing this and I can’t quite remember any manual denoising in Sound Studio for the bulk of the song, but there may have been some. This track was an eye opener in learning the ways that clever mixing may enhance remastering efforts in the hobbyist realm where The Monk plies his trade.
This was the last of the vinyl tracks to be remastered, and it gave me insights as to how I may enhance my efforts going forward. I’ve used this technique several times hence to reclaim many records that were “too far gone” as I am forced to buy much of my collection in the sometimes heartbreaking secondhand market, where not everyone takes such good care of their records as I do.
Next: …Living on video…