Zaxcom's NeverClip explained

[Brian Liston]

Drastic? Nah. If the levels were peaking at -24 and -32RMS, it may be audible under lab conditions. Maybe. But peaking at -6, -18 RMS, not drastic. A fuzzy term, drastic. If we are going to argue this point, let's stick to the math.

Since people are throwing in busses and floats, and all kinds of other stuff I just want to say that there is a lot of misinformation flying around here. Convoluted concepts of bit depth, floating point, mix bus bit depth, file bit depth. This is just no way to understand digital audio.

Before anyone starts dissing PT as not having stuff that others did for years, let's clear this up. Their mix bus has been 64 bit (plugins at 56bit) for a long time. The other guys all did the 32 bit floating point to extend their mix bus depth and plugin processing depth which was the same 24 or 32 bit depth that the files were. PT could do 128 tracks at full range before their bus would clip. Gain changes in PT result in zero reduction in file bit depth, and thus hasn't required a dithered bus, unless of course you want to dither your bus. If your mix bus and file depth is the same, you have to dither any gain change, and a couple of tracks mixed at full scale can clip the bus as did Logic et all did. I don't know if any of them have changed their ways recently as I have't really paid attention to them. Until any of them can take a stereo track and make dual mono, or the inverse, I can't consider them. Dude was showing me Logic and was bouncing files out and re importing them. No way. Garbage.

So Pro Tools hasn't needed to have 32 bit, and still really doesn't, but it let's PT go past 256 tracks at full scale without clipping. Bring on the Wagner!

Mixers end up with 16 bit files all the time as so many folks are still mixing on FCP7, which has sync problems at 24bit. They just crush it down to 16 and up-res a 24 bit OMF later. Every day.

Then, we end up with low level 16-Bit recordings which when gain changed back to normal levels, will also drastically increase noise floor. Someone please correct me if I'm wrong on this.

Pro Tools couldn't even handle 32-Bit float files until version 10. Nuendo, Samplitude/Sequoia, Pyramix, Sadie etc. users have had this capability for years.

[Jeff Wexler]

Thanks for the feedback Glenn. I think what I would like to see/hear (and don't anticipate being able to see/hear at a demo) is a file with realistic type issues we deal with. If someone is recording dialogue and slamming 10-12db over full scale that is a whole other mountain of issue more related to the operators judgment. I would like to hear a same with an unexpected outburst from soft to loud suddenly. Any chance of recording an example like that? Thanks!

That's exactly the test I did on the floor at AES Convention. Started by recording normal speaking voice at normal levels on 3 separate tracks. Then had the speaker do a full yell, over the top, returning to normal speaking voice. Played back track 1 which was straight through, no NeverClip, just like we would always do, and the yelling part was clearly distorted, clipped, pretty much unusable. Played back track 2 which had typical limiter in play and this track also sounded pretty bad, not as bad as track 1 but still obvious the limiter was working pretty hard, clamping down on the yelling... still not a track I would want to turn in. Listening to track 3 with NeverClip, you guessed it, clean as a whistle --- no clipping, no distortion, no nasty limiting. I'm a believer now that I have heard it for myself. It just works.

[soundslikejustin]

Jeff, was that in conjunction with a NeverClip ISO, or did you just have a fader turned down? If it was a NeverClip ISO, what was your new headroom setting?

[Philip Perkins]

That's exactly the test I did on the floor at AES Convention. Started by recording normal speaking voice at normal levels on 3 separate tracks. Then had the speaker do a full yell, over the top, returning to normal speaking voice. Played back track 1 which was straight through, no NeverClip, just like we would always do, and the yelling part was clearly distorted, clipped, pretty much unusable. Played back track 2 which had typical limiter in play and this track also sounded pretty bad, not as bad as track 1 but still obvious the limiter was working pretty hard, clamping down on the yelling... still not a track I would want to turn in. Listening to track 3 with NeverClip, you guessed it, clean as a whistle --- no clipping, no distortion, no nasty limiting. I'm a believer now that I have heard it for myself. It just works.

I did the same thing (today, @ AES SF) with the same result. Pretty convincing I must say. I also finally had a personal demo of the Nomad trim procedure from GS himself and now I get it--well done.

philp

[Brian Liston]

Can anyone hit me with the setting used for each of those three channels so that I, and other Nomad wielding soundies, may simulate the magic at home. With no great explanation of what is going on, I find it difficult to begin a loosly scientific understanding of how to use this with some empirical-based strategy. I want to perform that test, and then also see what this looks/sounds like in PT. Rubber meets road and what have you.

I did the same thing (today, @ AES SF) with the same result. Pretty convincing I must say. I also finally had a personal demo of the Nomad trim procedure from GS himself and now I get it--well done.

philp

[seth]

Can anyone hit me with the setting used for each of those three channels so that I, and other Nomad wielding soundies, may simulate the magic at home. With no great explanation of what is going on, I find it difficult to begin a loosly scientific understanding of how to use this with some empirical-based strategy. I want to perform that test, and then also see what this looks/sounds like in PT. Rubber meets road and what have you.

Brian I believe you have track 1 with out limiter/neverclip,

2nd with limiter but with out never clip,

3rd with out limiter but with never clip. All same trim settings.

Did the same test Jeff did today as well. Sounded pretty good!

I think a word that is missing from the discussion is DSP. Glen mentioned it today in our conversation. That's the magic that mixes the two "channels" a/d converters. I think its a really cool tool. I could see myself getting lazy with it.

Was thinking how it could work well with drums for live recordings. I wonder how it would deliver a musical performance, or change the dynamic range of someone screaming in a scene.

No matter how you look at it there is a 0 in pro tools(or -3 mabye), ect and tools like never clip and limiting effect the natural dynamics of the source. though after today I think never clip defiantly sounds more natural between the two. oh no I'm voicing opinions.

[Brian Liston]

There is a zero on every output bus, output or "Card Track" file in any recorder or DAW. The output bus can't go past zero. Fine, use the bits that you have more wisely, but there is a zero. The card track can't go past zero.

There is a very different zero on a bus. A 64 bit bus will allow one to Turn files up or down, without clipping the bus or changing any single bit of the file. The output bus can be attenuated to suit the need of the program material. All without changing a single bit of the program material. For this type of DSP, zero becomes relative. Let's not breech floating vs. fixed, unless ya'll want to.

If there is DSP on the track to alter it in some way dynamically...wouldn't that be a compressor/limiter/expander? DSP or not, turning things up or down is just that. You said never clip "effect the natural dynamics of the source." That is a compressor/limiter/expander is it not? So you are saying that never clip is a DSP compressor inside the Nomad? A natural sounding compressor, but a compressor.

So then what were the limiter settings used on 2? What were the settings for never clip on 3?

[MT Groove]

First of all, I really enjoy these discussions and I'm glad we have so many points of view on this topic. That's why I love this forum.

But peaking at -6, -18 RMS, not drastic. A fuzzy term, drastic. If we are going to argue this point, let's stick to the math.

Do you normally record at -6 RMS? I hope not. That's Pop Music CD levels. I don't even like math! In the event that your recordings get truncated to 16 bits, recording at -44dB can have quite an effect on quality and no type nor amount of dithering can save that.

Before anyone starts dissing PT as not having stuff that others did for years, let's clear this up. Their mix bus has been 64 bit (plugins at 56bit) for a long time.

Actually, Pro Tools HD's mixer depth is 48-Bit fixed. Pro Tools HDX is 64-Bit Floating point. HDX wasn't introduced until very recently, maybe a few months ago? I'm not sure what you consider a long time. Pro Tool HD's plugins still outputs at 24-Bits at the time I think. I think all that changed with HDX though. Most other DAWs have been 64-Bit Floating Point for some time as well. Besides, I wasn't "dissing" PT. I was merely stating that 32-Bit float files is not a standard in the post world probably because PT couldn't even handle them. We all know PT is the de facto standard for post. I brought up the other DAWs because it's a fact that they have the ability to handle 32-Bit float files for years. "Dissing" PT would be another 100+ pages worth of discussion and there have been plenty of that already. No need for us to go there.

Dude was showing me Logic and was bouncing files out and re importing them. No way. Garbage.

No one brought up Logic, however, I do agree with you that Logic is not very intuitive when it comes to editing and mixing audio. Nice platform for music production though.

So Pro Tools hasn't needed to have 32 bit, and still really doesn't, but it let's PT go past 256 tracks at full scale without clipping. Bring on the Wagner!

There is a reason why PT10 started supporting 32-Bit float files. On the 256 track projects, chances are only part of those tracks are active simulatneously. Plus, I'll be scared to hear any mix that has all 256 tracks blaring at full scale!!

Mixers end up with 16 bit files all the time as so many folks are still mixing on FCP7, which has sync problems at 24bit. They just crush it down to 16 and up-res a 24 bit OMF later. Every day.

When audio gets truncated to 16-bits and gets "up-res" to 24-bits, guess what, you still have 16-bit audio with a bunch of empty data way below. Think of what could happen if you -44dB NeverClip audio lands on one of these FCP7 editors.

We know the input stage of NeverClip cannot be clipped. As explained by Glenn, the demo was created by recording at 10-12dB peaking over full scale on the Nomad's input. Once the audio is within Nomad's mix engine, it is in 32-Bit Float form. Yes, writing to a 24-Bit file can reproduce the full dynamics by attenuating the signal. However, wouldn't it be great to write the file to 32-Bit Float without any attenuation? To "record" the Nomad's mixer output in all it's full 32-Bit float glory?

At the risk of being ridiculed for improperly trying to demonstrate the benefits of a 32-Bit Float file, I will attach a few screenshots.

First screenshot is the original recording. It peaks at near 0dB Full Scale. I normally don't record that hot but for this demo, it is suitable.

Second screenshot shows the audio boosted by 12dB, exported as a 24-Bit file and reimported back into the DAW. Obviously the waveform is chopped. Also notice the level meters peaks at 0dB. The Bit Scope shows 24 Bits and no gain change is applied within the mixer.

3rd Screenshot shows what happens when you take that same boosted audio and attenuate it by 12dB (Notice the fader on channel 1 is at -12). On a 24-Bit file, anything that was originally in the internal mixer over 0dBFS would get chopped and you can never recover. That file will still sound distorted.

4th Screenshot shows the original recording boosted by 12dB and exported as a 32-Bit float file and reimported back into the DAW. The waveform is still chopped but levels actually exceeds 0dBFS as shown in the meters.

5th Screenshot shows that boosted 32-Bit float file attenuated by 12dB. Fader on ch 1 is at -12. The waveform is restored to the original recording's. The audio file does not sound distorted and sounds identical to the original.

I hope I have given a visual example of the benefits of a 32-Bit float recording. I think this perfectly complements NeverClip!!! In fact, we should call this process NeverEverClip. (Pre-Pat. Pending) By going this route, you can truely never ever clip the input nor the ISO or even the Mix track if implemented. I know 32-Bit Float files are not commonly used these days. However, since PT10 now supports them, it may well become a standard in the near future. Don't be surprised if Zaxcom implement this feature in the new line of Deva/Fusion. (Deva 32, Fusion 32 maybe??) I'll definitely trade in my current Fusion 12 for one.

[Dan Ostroff]

That's exactly the test I did on the floor at AES Convention. Started by recording normal speaking voice at normal levels on 3 separate tracks. Then had the speaker do a full yell, over the top, returning to normal speaking voice. Played back track 1 which was straight through, no NeverClip, just like we would always do, and the yelling part was clearly distorted, clipped, pretty much unusable. Played back track 2 which had typical limiter in play and this track also sounded pretty bad, not as bad as track 1 but still obvious the limiter was working pretty hard, clamping down on the yelling... still not a track I would want to turn in. Listening to track 3 with NeverClip, you guessed it, clean as a whistle --- no clipping, no distortion, no nasty limiting. I'm a believer now that I have heard it for myself. It just works.

Thanks Jeff, that example makes more sense.

[Wandering Ear]

MT

32 bit float files are not necessary to take advantage of neverclip. 24 bit files have a dynamic range of 144db, and the nomad with neverclip has a dynamic range of 137db. Therefore a 24bit file can easily store the entire dynamic range available with the nomad.

Brian,

The dsp refered to is not a compressor, but signal processig that puts together the output of the two ad's and attenuates the entire signal (not just loud peaks) until all the audio is below 0dbfs. In this way the original dynamics are maintained.

Wandering Ear

[Jeff Wexler]

Hey, Dan, no one is expecting to routinely over modulate and slam into anything... Limiters and now NeverClip are there to protect those unexpected excessive levels and attempt to record those sounds as best as they can. Good limiters can do a decent job but NeverClip allows for the full use of extended dynamic range without any of the affects or artifacts of limiting.

[Dan Ostroff]

Hey, Dan, no one is expecting to routinely over modulate and slam into anything... Limiters and now NeverClip are there to protect those unexpected excessive levels and attempt to record those sounds as best as they can. Good limiters can do a decent job but NeverClip allows for the full use of extended dynamic range without any of the affects or artifacts of limiting.

Definitely understand. The example provided seemed to be doing just that (providing an example of DIA slamming the limiter and comparing that to the NC file), thus hoping for a different more subtle example or should I say a more realistic use for NC. It would seem that in most situations you would hope that running your ISO's significantly lower than the mix track would avoid limiting, but the added security of NC would certainly be welcomed.

Zax is certainly putting together some great features, when automix comes out maybe folks will be able to set it and forget it! Kidding.

Hope everyone in the path of Sandy is safe out there, heavy rain starting here in Maryland.

[Jeff Wexler]

I agree completely that the first demo with nice normal level speaking voice pushed to an excessive level was confusing and did not really demonstrate the true value of NeverClip. I mentioned this to Glenn Sanders right after I heard the demo, explaining the using a pleasant and reasonable program source (normal speaking voice) was at odds with the situations NeverClip was designed to deal with. That is why the demos they have been doing at AES have wide dynamics which truly demonstrates the distinct advantage of NeverClip over any sort of limiting. I'm pretty sure after AES, Zaxcom will put up a new demo recording.

[Dan Ostroff]

I agree completely that the first demo with nice normal level speaking voice pushed to an excessive level was confusing and did not really demonstrate the true value of NeverClip. I mentioned this to Glenn Sanders right after I heard the demo, explaining the using a pleasant and reasonable program source (normal speaking voice) was at odds with the situations NeverClip was designed to deal with. That is why the demos they have been doing at AES have wide dynamics which truly demonstrates the distinct advantage of NeverClip over any sort of limiting. I'm pretty sure after AES, Zaxcom will put up a new demo recording.

That'll be helpful for folks not being able to see/hear it in person. Thanks Jeff!

[RPSharman]

So neverclip records two files and then reassembles them as one file?

[Wandering Ear]

So neverclip records two files and then reassembles them as one file?

NC uses two ad converters in combination to bring the audio into the nomad with a large dynamic range. It's recorded to one file.

[Max Hirtenstein]

Glenn gave me a NeverClip demo at AES, and I'm a believer. It made perfectly useable production sound out of what would otherwise be garbage. 32-Bit makes sense for plugin processing, but I think 24-bit is still enough for production tracks.

[airport70]

Possibly, I think I'm starting to understand it, maybe perhaps. In practical terms, when handing over files to post, will the files appear to be recorded quietly. That wouldn't be a problem but a heads up would be certainly be useful, otherwise I could see post assuming incompetence.

[Philip Perkins]

No, the files will appear to be recorded correctly, full fi and no clipping.

philp

[pindrop]

NeverClip seems like one of the holy grails of sound recording along with Zynaptiq 'Unveil' reverb removal.

Are there ANY downsides to NeverClip?

[rich]

i think there is some confusion about neverclip being used in the real world. for example, just because the demo has the audio recorded 12dB to hot, doesnt mean that you would, as a matter of course set your gain to that high.

for me personally, i would use never clip as a better sounding alternative to a limiter, and have my iso's attenuated so that they dont have to be limited for the occasional peak, or at least have less limiting. which in turn means that my mix track is at the appropriate level, and there is enough headroom for the isos. though of course you may have to hit your limiters for you mix track and output tracks for unexpected peaks. but at least your isos will be free of limiting.

and this attenuation is something i would like to have seen on multitrack recorders previously.

[Christian Spaeth]

Just so we're clear, if we're talking about ISOs with NeverClip funcionality - I wonder who here records multiple ISOs hardwired? For any wireless on your inputs NeverClip is quite useless (even Zaxcom wireless until they implement it there too).

[Jeff Wexler]

Just so we're clear, if we're talking about ISOs with NeverClip funcionality - I wonder who here records multiple ISOs hardwired? For any wireless on your inputs NeverClip is quite useless (even Zaxcom wireless until they implement it there too).

This is true and NeverClip can only do it's magic with a microphone hard wired to the input. When the input sees the output of a wireless receiver, whatever has happened on the way to that receiver will not be helped by the extended range of the NeverClip input. The NeverClip input will be available, so I am told, on the transmitter input at some time in the future (maybe already on some models) and this will of course benefit those using wireless (Zaxcom) to the input on Nomad.

[Tom Duffy]

Just a reminder that Glenn has said that he is applying for a patent on the process. While the application is pending and he does not yet enjoy protection for the invention, he may be constrained as to how much he can explain. It sounds like he is restricted to explaining broad principles when people are seeking all the details.

David

Yamaha used dual ADC circuits in their converters since the early 1990s to overcome the limited dynamic range of the chips available at the time, and yes, they patented the heck out of those designs, e.g.

http://www.google.com/patents/US4999628

The chip manufacturers themselves have a huge stake in this style of processing as well:

http://www.google.com/patents/US6271780

I'm certainly interested in seeing what Zaxcom has come up with that improves on this process.

Tom.

[al mcguire]

" Seismic playback system "