Crackle in Levelizer

[Cary B. Cornett]

I sometimes get asked to do mastering.
These days I upsample to master, so a 44k1 source file gets run/processed in an 88k2 session for better process quality.
It is possible that Levelizer doesn't like working this way.
I recently used SawStudio to construct my own multiband upward compression setup.
I copied the original region to five new tracks.
I used the hp/lp filters in the console eq channel so that each of the five tracks would handle one band of frequencies.
For each band, I set up the console compressor as a slow attack hard limiter, then inserted a Levelizer as a fast compressor. I did not use Levelizer's limiter in these bands.

As a soloed each band to make adjustments, I noticed an odd craclkling noise, not loud but definitely there. Bypassing the Levelizer stopped the crackling. Leaving Levelizer in, but setting its compression to OFF, also stopped the crackling. To clear the problem, I had to replace the Levelizer in each band with a different compressor plugin.

In this session, I also used a Levelizer at the very end, just the peak limiter, and this worked without trouble, as far as I could tell.

So, I have to think that under certain conditions (higher sample rates?) there is a bug in the compressor section of Levelizer.

[Bob L]

My guess is that the crackling could have been caused by using too fast of an attack time... the Levelizer is capable of absolute zero attack time... which on certain instrument waveforms (like low frequency bass guitars) can add a spike to the attack of the waveform... most compressors do not reach zero attack time and actually start at a few milliseconds... try adding a few milliseconds to the attack time and see if the crackling disappears.

Bob L

[Cary B. Cornett]

My guess is that the crackling could have been caused by using too fast of an attack time... the Levelizer is capable of absolute zero attack time... which on certain instrument waveforms (like low frequency bass guitars) can add a spike to the attack of the waveform... most compressors do not reach zero attack time and actually start at a few milliseconds... try adding a few milliseconds to the attack time and see if the crackling disappears.

Bob L

I did not use zero attack time on any band, and I used longer attack times on the lowest bands. When I switched to the Ultrafunk compressor, I used the exact same settings I had used with Levelizer's compressor, except that the Ultrafunk has a "knee" setting which I set to 3 dB (Levelizer, as we know, is strictly "hard knee", which I don't think is the cause of the problem). So, same settings with the replacement compressors, but no crackle.

Is a 2 or 3 mS attack setting too much for Levelizer to handle without crackling at midband (over 400 Hz) frequencies?

I suppose I could re-try with longer attack times as a way to narrow down, though.

[Bob L]

Never noticed a problem before. It should be fine.

Bob L

[Cary B. Cornett]

Never noticed a problem before. It should be fine.

Bob L

Neither did I before this, and I have used Levelizer for years. Then again, I only just recently started trying to work at 88k2 sample rate with it. More when I get to running that test (probably tomorrow).

[Naturally Digital]

Neither did I before this, and I have used Levelizer for years. Then again, I only just recently started trying to work at 88k2 sample rate with it. More when I get to running that test (probably tomorrow).

Cary, over the years I have done the same and similar to what you are doing here. Two things come to mind...

You will find that when processing low frequency content (as you would have in your lowest of the 5 bands), some compressors suffer from this when the attack time is set very fast. The channel compressor and Levelizer compressor (being the same algos) demonstrate this under these "extreme" conditions (a self constructed 5 band multiband compressor). This is very understandable IMO and requires a compressor designed to avoid these issues when/if using very fast or near zero attack times. Peter's (Sonoris) compressor plug in handles this well but at the cost of some very slight latency built in. You're best to try other compressors than the SS comp or else increase the attack time until it goes away.

I remember discussions in the forum about this same question of clicks. It was in the context of processing bass guitar IIRC. Again, not a bug IMO but a consequence of the fast (latency-free) and efficient performance of that compressor design. You can print (render) a LP filtered signal (that includes LF info) and inspect the waveform on the timeline, and this will indicate how long some of the low freq waveforms are in time.

The second point likely has nothing to do with the issue at hand but that is, the phase of your 5 band crossover... If you haven't already, you should verify via test and measurement the 5 bands will sum correctly to (re)produce a perfectly flat response when the 5 bands are recombined (without any compression processing). Again, very likely has nothing to do with audibility of the clicks, but there's also the possibility that no combination of HP/LP and phase settings in your crossover will create a perfectly flat response when summed back together.

I moved to using the Sonoris LP EQ to accomplish what you are doing with combinations of various compressors but especially the Sonoris Comp and Sonoris Leveller.

I don't feel the clicks have anything to do with the upsampling since I have years of experience doing that in SAWStudio, in real time, with clean results every time. You may want to verify you are using the highest quality settings for both real time and offline rendering, but you've likely already checked that.

Cheers,
Dave

[Cary B. Cornett]

OK, Bob suggested that attack times being too short caused the crackling problem. This morning I ran a test using the problem session with the levelizers. For each band, I had threshold at -30, ratio at 8, release time at 150 mS, and 6 dB makeup gain. On the low band, I used 12 mS attack time. I used 3 mS for lo mid and mid bands, 2 mS for hi mid, and 1 mS for top band. Soloing the lo mid band, I could hear crackles (kind of like the groove noise of a messed up LP, but at a low level). I could understand a sort of "snap" at an attack if the attack time is too short, but an ongoing crackling noise?

If I increase attack time to 12 mS, the crackle goes away.

If I use the fast attack times with the Ultrafunk R3 compressor, all other settings being the same as on Levelizer (except for a 3 dB "knee", which Levelizer does not offer), I do not get the crackling. What I want to know is, why the difference? And if fast attack times are problematic with Levelizer, why even offer them? I am confused...

[Bob L]

Forget the numbers... raise the attack time on the lower frequencies till the crackle goes away... the values are offered so you get the opportunity to set them where you want them... for certain frequency material, the very fast attack times work fine... so why would I not offer them. Use your ears... not your eyes... number values are for reference only.

Trying to compare completely different algorithms by setting the number values the same is not a meaningful concept. Each will produce their own results... they will sound different even though they are both labeled as compressors... after all... why have more than one if they all sound exactly the same.

Bob L

[Grekim]

I agree with Bob. It would be odd for two different compressors with the same parameters to produce the same results. I have actually designed a couple and believe me there is no one way to go about it. That really is what makes them fun...there is no hard rule on how to design it. For example, is the attack linear or following some sort of curve? Does the compresson "stick" a little even after the signal drops? Are we sampling at zero crossings of the waveform or over some interval? All that plays a part. So, perhaps make the levelizer sound as good as it can and then compare.

Are you sure this project needs such a radical treatment? I'm sort of cringing at the thought of breaking up the bands....usually it is a last resort to have to multiband compress a mix. That said, I have had some nice results with the Waves C4 multiband and I've always wanted to pick up the Sonoris multiband. I usually use the Sonoris (regular) compressor followed by either the Levelizer (as a limiter) or the Waves L2 if it really needs to be pushed. Good luck.

[Bob L]

I agree with Grekim here... you may be creating all kinds of extra artifacts that are cancelling any usefulness your complex concept may be introducing. Just the fact of upsampling, there is one set of artifacts... then the split into the multi-bands and the re-blending of those bands back together are most likely craeting many artifacts at the crosover points... then the downsampling adds more artifacts... in the end... you may have actually done more damage than good... way too complex for my taste.

I have done many mastering sessions without all this complexity that have made clients extremely happy with the final results... it really does not need to get this crazy, in my opinion.

Bob L