I almost feel guilty posting so many questions about my vinyl LP recording project. But so far everyone has been nice and helpful, so I guess I'll try to glean some more information from the pros.

I have recorded about twenty albums, each as a single WAV file. I chose to engage a "subsonic" filter on my preamp which supposedly removes frequencies below 17 Hz. The recordings sound great, but I can hear rare pops. On a couple albums, when there is a single instrument playing such as electric guitar, I can faintly hear low level "crackling" with headphones. I have not tried editing any of these WAV files yet, and I'm sure I'll gradually learn by trial and error. But here are some questions.

1. Since I recorded with the subsonic filter on, I thought I might skip doing a rumble filtering. Do people approve or should I not use the subsonic filter and instead derumble in Audition?

2. When declicking should I just select a small segment of the WAV file that I know contains an audible click/pop? I thought I'd first try the auto declicking tool in Audition, but I don't see a reason to declick the entire WAV file. Is this thinking correct?

3. Are decrackling and reducing noise the same thing? The Audition manual mentions briefly that the auto click/pop eliminator can remove vinyl record crackle. I'm not sure how I would change settings to remove crackle as opposed to large clicks.

4. Audition also has a hiss reduction effect and a noise reduction effect. So far, I don't think I'm hearing any hiss. But I wonder if the noise reduction effect would actually be the appropriate tool for the low level crackle that I've heard?

One of my problems is simply recognizing what the spurious sound is called. I know the names are discriptive, but it would be nice to be able to hear examples of all the various unwanted sounds and know exactly what "broadband noise" sounds like.

5. After I normalize the WAV file, I plan on changing the bit depth from 32 bit float to 16 bit. The Audition manual suggests that I enable "dithering" and use a dither bit depth of 0.2 to 0.7. It says lower dither depth values work best with noise shaping. A chart lists eleven different noise shaping curves, but very little guidance how to choose one. The manual also suggests using a p.d.f. (probability distribution function) and states that "triangular p.d. f." is usually a good choice. Does anyone have advice for me regarding these many options? Dither, dither depth, p.d.f., noise shaping oh my!

6. When I finally split my WAV file into tracks, I thought I would try to keep the original silent interval between songs. But sometimes this interval isn't completely silent because of crackles, etc. Is there any reason why I shouldn't just create digital silence between songs or do people prefer to decrackle, denoise, declick these silent areas instead?

7. Finally, in general when you apply a particular sound restoration tool to a file, do you apply it to the entire file or try to only select short segments where you can definitely hear the problem?

Thanks again.

Using a hardware rumble filter should present no problem. You can look at a Frequency Analysis graph of between tracks and see if it is adequate or you want more done in the software.

Declicking is for you, or your audience. If you don't care about the hundred to thousands of small clicks that are surely there, you can search out and treat only the larger ones. There is always some downside to doing automatic declicking, it always removes more material than it should (i.e. removes energy from some transients of the music). Done well, the deficiency is unnoticeable to noticeable only in a close A/B comparison. Doing it well takes a bit of experimentation and practice.

Noise reduction will have some impact on crackle, but it isn't really the same thing. The "Younglove" Decrackling technique (generally referred to as the Y. D. Script) is the best way to go about it. You will find adequate information on the Audiomaster site. I can answer questions if anything isn't clear; I use it almost every day.

Few people use Hiss Reduction for anything. I only recall one person saying he got better results on hiss than with Noise Reduction, and he did not give any details about how.

Listen to the unmodulated groove between two tracks for an example of broad surface noise. In general the term just means noise at many or all frequencies, such as white noise, which has equal energies at all frequencies. You don't get a white noise spectrum off vinyl, it is much more concentrated in the lower frequencies.

I rarely make any effort to remove hiss. Some quieter older recordings have a great deal of tape hiss and can be improved with noise reduction, but the majority of what I've done doesn't have enough hiss to bother with. It is always a narrow line to walk, getting good noise reduction and not losing too much musical energy along with the noise.

I like
Dither Depth 0.5
p.d.f. Shaped Triangular
Noise Shaping C1

I've used
D. D 0.7
p.d.f. Triangular
Noise Shaping C3
which makes for measurably more noise, but probably no audible difference.

Some people always do a fade out to silence, yea so many seconds of silence, fade into the next track. You probably would not like abrupt demarcations between treated and untreated if you just selected and applied silence. If you do general declicking & decrackling, with some overall NR, the between track intervals may not need anything else (in general). Sometimes I do special declicking and/or NR on the between track intervals, because they are extra noisy. Every LP presents a bill of decisions for itself.

Some people claim to always separately consider each track, and frequently different parts of each track, for the amount and type of treatment. Occasionally I will do an entire LP manually, as far as impulse noise goes (clicks, pops, crackle, etc.). Most often I treat the entire album the same, once I decide just how much declicking to apply (how aggressive). Then I listen to the entire thing and do manual fixes as necessary (which varies from almost none to many many hours worth.

When a recording requires so many hours of attention, the recording could probably have benefitted from more aggressive auto declicking. However, sometimes the nature of the music (or more likely the vocals part thereof) can't take more automated declicking. Of course, if I had managed to figure out better parm settings for that particular recording, I might have done with less effort, but sometimes there just doesn't seem to be an acceptable compromise.

I've already indicated something about selective use of NR. I almost always do some with the lower frequencies, up to 500Hz or so over the entire album. Sometimes a notch filter for hum is beneficial. I often do less NR on the higher frequencies. Sometimes I apply some, or more than the general amount, to quieter passages, but I don't often go to that much trouble and I don't think anyone could tell with very many recordings. The beginning and endings of tracks, often being at a lower level, more often require more attention than the main part of the track.

In addition to Andy's excellent advice, I'll offer a couple more points:


If you're prepared to put in the effort required to do declicking manually, then you'll get better results that way.

Since you're using Audition, the most efficient way of tracking down isolated clicks is to use spectral view - clicks are noticable as obvious vertical lines. And to fix individual clicks, the "Fix Single Click" option in the click/pop eliminator usually works well.


As Andy has pointed out, they are not the same, and the Younglove technique can work well. If you can't find the info you need on the Audiomasters site, it's also described on my website here. I have never found any automatic declicker which does decrackling as well as the Younglove technique.


I find the sudden disappearance of surface noise into digital silence between tracks to be extremely irritating. You can try using broadband noise reduction to reduce the level of inter-track surface noise, but my advice is not to eliminate it completely.

And finally, when splitting tracks, be careful to do so on CD block boundaries (ie. exact multiples of 588 samples).

cliveb, I have "snap to frames" enabled in Audition. I assume that means that whenever I select a portion of the wave form, or if I just place a track marker, that the boundaries or marker will adjust exactly to CD blocks. I expect this graphical adjustment to be so minute that I wouldn't perceive it even happening, unless I was really zoomed in on the waveform.

Here comes a dumb question, but does crackle sound just like that, low level crackling almost like crumpling paper? So far I've only been able to hear this when the music was softer. Should I just assume that if present in the quiet sections, it must be present throughout the song and apply this Younglove technique to the whole song?

I haven't read about Younglove yet, but is this a plugin that can be purchased or is this a manually intensive processing technique? Clicking a button would be nice.

AndyH-ha, thanks for the specific dither, etc. recommendations. I'll use those settings since my understanding of this process is pretty poor. I mean, how can you even have a fraction of a bit? I thought bits where 1 or 0. Now I'm going to use a "bit depth" of 1/2 a bit. Crazy! Of course, I guess dithering etc. can be completely avoided by simply doing 16 bit recording initially. I think cliveb prefers this. Hopefully, the advantages of processing in 32 bit and then converting to 16 bit outweigh the small distortions that get created. Otherwise, we should all follow cliveb's method.

I believe you will find having ‘snap to' on is extremely annoying, except when you need it to separate tracks, or for some other specific reason. To enter track boundary cues, the Time Display needs to be set to ‘Compact Disk 75 fps,' or whatever Audition 2 labels that condition.

Sometimes I Amplify by a moderate negative amount, say -30dB, between tracks (with appropriate fade out/fade in) to get rid of unpleasant noise but still remain mellow.

Crackle is impulse noise, not unlike clicks, but much lower level. It is generally widely present throughout the recording when it exist anywhere. There is a higher density per time unit when it is worse. The Younglove process is useful for more than just crackle, and I sometimes use it on limited selections for these more specialized purposes, but it would, in general, be extremely tedious to try to process very many small sections for crackle rather than just doing the entire recording.

Some people always do track separation as the first step, and put the tracks in separate files to be independently processed. I did that for awhile, but find it to be significantly more work without much benefit. As cliveb says, you can get superior results by giving individual attention to the individual problems (such as each individual click or smear) but I think, if your techniques are well developed, it gets to be like the improvements from spending $15k rather than $10k on your playback system -- smaller and smaller payoff for greater and greater investment. Since all value is personal, only you can decide how much is worth while.

Younglove is a process, named after its developer, who sometimes pops in on the Audiomaster site. Since it involves four steps, it is often run with a script, and thus referred to as the "Y ... Script," but it can easily be done one step at a time.

The advantages of using 32 bit float for processing are quite real in the same sense that gravity is real. If working in 16 bit, dither should be applied during transforms. However, doing very much of that adds noticeable noise to the recording: dither during transforms in Audition cannot be noise shaped. There is no downside to the audio from processing in floating point. There are no distortions created when downsampling to 16 bit (with dither), if done properly.

The only question is whether or not the differences between processing in floating point and integer are audible. The above statements definitely apply to good clean original recordings, but since there is so much background noise on LPs, transforms are self dithered (the same really applies to resampling from 32 bit to 16 bit). You are adding "much" more quantization noise when processing LPs in 16 bit, but it is all relative, and thus not so likely to be noticed.

As Andy has pointed out, having snapping turned on all the time is likely to be frustrating. If you are serious about doing the best possible cleanup you can, then there will be plenty of times when you'll want to be able to edit to very precise points - having your selections auto-snapped to slightly different places won't help.


Crackle is a constant background of little ticks. It sounds a bit like frying bacon or rice crispies.

As for whether you should apply the Younglove technique over the whole file, I'd say: it depends. In general the Younglove technique does not generate artifacts, and so one could argue that it's harmless to apply it in sections where you can't hear the crackle. And since it's quite an involved process, you probably don't want to manually go through each stage multiple times (once for each quiet section where crackle is audible). That said, if you're using Audition you can set up a script to do it with a single button push, and since it takes quite a long time to run, you might want to restrict it to just the sections with audible crackle in order to save waiting time.

One other comment: if you have an LP where the crackle is so bad that it's audible during the loud bits, I'd say it's probably beyond help. Getting rid of crackle at such high levels is next to impossible without very serious damage to the music signal. You always need to keep a sense of perspective and be realistic about what's achievable when doing LP cleanup.


As mentionded above, you can set up a script in Audition. There's an old Audioforums thread archived here which includes an example script for CoolEdit Pro - I don't know if it's forwards-compatible with Audition. The only other editor I know of which does it with a "single button press" is my own shareware Wave Repair: the decrackling function uses the Younglove technique. But Wave Repair is 16 bit only, so that's probably not on your radar.


The reason I work at 16 bit is not to avoid dither. Indeed, I fully agree with Andy that in principle working at 24 or 32 bit then doing one final conversion to 16 bit at the end will yield more accurate results. The question is, with LP source material, how accurate do you need to be? It is my strong belief, based on many years experimentation, that the noise level of vinyl LPs is high enough that the inaccuracies of working at 16 bit never become audible above the LP's noise floor. And since there are some very useful tools around that only work at 16 bit, I see no reason to deny myself access to them. (And I'm not only referring to my own program Wave Repair: I think that Wave Corrector (no relation) is one of the best auto-declickers around, and that only works at 16 bit).