Zaxcom's NeverClip explained

[PTA]

I am a little confused about setting the ISO gain for NeverClip. Will 0 dBFS on my Nomad meter on the ISO be -20dBFS when I open the file on a DAW?

I was wondering this too, how is this represented on the metering screens?

[VASI]

It's different with NeverClip and limiter.

NeverClip clean the background noise and self noise.

[FredS]

Is this a re-purposing of dithering bits so they carry content payload?

[Matt Martin]

I was wondering this too, how is this represented on the metering screens?

I just installed 4.00 and tested it out. The Card meters DO display the extra head room. For example, if I send tone at -20db from my QRX, and I have say, 6db headroom selected as my "Iso Attenuation" then the card meters for those ISO tracks will show the tone at the -26db mark.

Input meters and Output meters behave the same as before (since the new neverclip Iso record option only affects the card tracks you select)

[RPSharman]

I still don't get it.

Is it just a seperate file with lower gain? Does it record all the time, or is it a buffering system in that it only records a "never clip file" when it peaks? If so, how does it work in a transmitter?

My brain doesn't comprehend.

[studiomprd]

next week we'll be discussing exactly how a superhetrodyne radio circuit operates, and compares to a superregenerative radio circuit.

[MT Groove]

It seems to me that the Neverclip ISO is simply attenuating the gain so as not to hit 0dBfs?? I feel the only way to have a true "Neverclip" file is to record to a 32-bit float file. On a 32-bit file, you can go over 0dBfs. Anyone with a 32-bit float DAW can try this. Take an audio recording that peaks close to 0dBfs. Increase gain by say 20dB and write/bounce/export it to a 24-bit file. The resulting file will be distorted on playback. The waveform will look chopped. If you import that file back into your DAW, and reduce gain by 20dB, it will still be distorted. Now if you would have bounced that recording to a 32-bit float file. It will also sound distorted on playback because it is clipping whatever D/A converter you are playing back with. However, if you import that file back into the DAW, and drop the gain by 20dB, you will once again regain the original recording's dynamics.

I really think the Neverclip ISO is nothing more than writing the file at a lower gain because it is still only a 24 bit file. Just my $.02.

[glenn]

I will be at AES with a live Neverclip demo. I hope to see you all there.

Neverclip files do take advantage of the 4-5 least significant bits of an audio file and uses them to provide up to an additional 24 dB of headroom. Nomad owners that now have version 4.0 can set their units up to demonstrate the NeverClip advantage and you can of course start using it right away.

"I really think the Neverclip ISO is nothing more than writing the file at a lower gain because it is still only a 24 bit file. Just my $.02."

A 24 bit file has 144dB of dynamic range and is perfect for our 137dB of dynamic range. Floating point files are unnecessary. Neverclip files record the extra headroom up to the 137dB dynamic range of the Nomad. The demo file clearly demonstrates the difference.

"Someone added background noise"

Nothing was added to either track. Our live Demo at AES will have the same result. Please stop by if you can.

Glenn

[RadoStefanov]

Glenn,

Can you write a short guide that we can forward to post.

It will be very helpful...

I will be at AES with a live Neverclip demo. I hope to see you all there.

Neverclip files do take advantage of the 4-5 least significant bits of an audio file and uses them to provide up to an additional 24 dB of headroom. Nomad owners that now have version 4.0 can set their units up to demonstrate the NeverClip advantage and you can of course start using it right away.

"I really think the Neverclip ISO is nothing more than writing the file at a lower gain because it is still only a 24 bit file. Just my $.02."

A 24 bit file has 144dB of dynamic range and is perfect for our 137dB of dynamic range. Floating point files are unnecessary. Neverclip files record the extra headroom up to the 137dB dynamic range of the Nomad. The demo file clearly demonstrates the difference.

"Someone added background noise"

Nothing was added to either track. Our live Demo at AES will have the same result. Please stop by if you can.

Glenn

[Chris Campion]

Someone added background noise

It's different with NeverClip and limiter.

NeverClip clean the background noise and self noise.

I think the question is being avoided,

The background noise difference between the two files are WAY beyond what one could term as noise floor, (and therefore in the 'non-significant bits').

I don't think that background has been added, but what it sounds like to me is that the demo track with the limiter is recorded hotter, so that it does bounce off the limiter in this ugly way. Otherwise whats the reason for the jump in ambient noise?

[soundslikejustin]

I think the question is being avoided.

The background noise difference between the two files are WAY beyond what one could term as noise floor, (and therefore in the 'non-significant bits').

I don't think that background has been added, but what it sounds like to me is that the demo track with the limiter is recorded hotter, so that it does bounce off the limiter in this ugly way. Otherwise whats the reason for the jump in ambient noise?

I think this is a bad example file, as you are not dealing with instantaneous peaks (the whole reason input limiters and NeverClip exists), but a recording consistently made too hot. Sure, NeverClip might be great, but you could have had the same effect as the NeverClip Iso (in this case) by turning the input gain down...

[bassmagnetic]

I'm still a bit confused with the settings and metering.... for iso attenuation there are 6db increments down to 24db... How do I know how to set this, should I always leave it on -24 to give the most headroom, what is the trade off?....

and at each of these 6db increments what is the reference level?

[Mark LeBlanc]

The specs of most of the mics/lavs I use show max SPL in the 120-130 range. Cos-11, MKH 416 and Mkh50. So in theory, they will start to have issues before neverclip.,A better demonstration is an actor talking at normal voice then go into a raging scream.. That's the situation I normally have issues with.. Those sudden spikes.. I'd like to hear that and see how the waveform looks in Protools...

Mark L

[Dan Ostroff]

Would seem more helpful to hear a comparison of clips with loud unexpected transients, where Neverclip seems it would pay off big time.. The non-Neverclip clip just sounds like its slammed against the limiter. I don't anticipate folks running their analog mixers in this fashion. Unless they're deaf.

[Jeff Wexler]

The background noise difference between the two files are WAY beyond what one could term as noise floor, (and therefore in the 'non-significant bits').

I don't think that background has been added, but what it sounds like to me is that the demo track with the limiter is recorded hotter, so that it does bounce off the limiter in this ugly way. Otherwise whats the reason for the jump in ambient noise?

Do not make the mistake of confusing noise floor with background noise --- two different things. Glenn's mention of the noise in the least significant bits is part of the explanation of how NeverClip works to extend the dynamic range --- it does not refer to the ambient real world background sound on the demo track.

[Jeff Wexler]

I think this is a bad example file, as you are not dealing with instantaneous peaks (the whole reason input limiters and NeverClip exists), but a recording consistently made too hot. Sure, NeverClip might be great, but you could have had the same effect as the NeverClip Iso (in this case) by turning the input gain down...

I agree that the demo could be more useful if it employed the recording situations where NeverClip is going to save you and the traditional limiting and compression will not, but the methodology used in this first little demo still demonstrates the principle quite well. In some ways it is easier to evaluate the benefit of NeverClip with the steady, even voice test, spoken at normal levels but brought into the input at very high level, not normal levels. I think in a short demo like this, it might be more difficult while listening to evaluate what the limiter is doing to a fast transient peak that NeverClip does not do. The recording in the demo "consistently made too hot" was just a choice for the methodology of the demo.

[Jeff Wexler]

Would seem more helpful to hear a comparison of clips with loud unexpected transients, where Neverclip seems it would pay off big time.. The non-Neverclip clip just sounds like its slammed against the limiter. I don't anticipate folks running their analog mixers in this fashion. Unless they're deaf.

True, no one is planning on these source levels getting out of control and slamming into the limiter, but we all know it happens and that's why we have limiters. Now, we have NeverClip so that when the source levels DO go beyond the traditional range that we have typically set our input trims, the high level source doesn't slam into anything (except at the very extreme, just shy of the maximum 144db dynamic range that a true 24 bit file can achieve). As it has been explained to me, input trim levels should be set for optimum gain setting for the microphone in use and the anticipated level of the source that is being recorded, this is always the way we have done it. When something un-expected happens and there is sound that would normally fall outside the upper range of your input trim setting, NeverClip allows that sound to fit into a larger dynamic range beyond your input trim setting without hitting a limiter or clipping.

[Dan Ostroff]

Thanks Jeff - makes sense. Hoping to hear another example where it will be more likely to save mixers' tails in unexpected situations.

[Jeff Wexler]

I don't think that background has been added, but what it sounds like to me is that the demo track with the limiter is recorded hotter, so that it does bounce off the limiter in this ugly way. Otherwise whats the reason for the jump in ambient noise?

The ambient background noise is louder because the limiting (which is actually a form compression) makes the perception of the difference in level between the voice and the background smaller --- by holding down the voice it moves it closer to the background sound. The limiter actually limits (or diminishes) the dynamic range thereby making the level of the voice and the level of the background closer together.

[Chris Campion]

Do not make the mistake of confusing noise floor with background noise --- two different things. Glenn's mention of the noise in the least significant bits is part of the explanation of how NeverClip works to extend the dynamic range --- it does not refer to the ambient real world background sound on the demo track.

Thanks Jeff, I'm not making that mistake. You misunderstand - I'm saying the ambience in the clip is of much greater amplitude than the 'bits' that are to do with noise floor - therefore reassignment of the least significant bits does NOT account for the reduction in ambient noise in the demo.

The ambient background noise is louder because the limiting (which is actually a form compression) makes the perception of the difference in level between the voice and the background smaller --- by holding down the voice it moves it closer to the background sound. The limiter actually limits (or diminishes) the dynamic range thereby making the level of the voice and the level of the background closer together.

Nice try, but we can assume the limiter is not active in the brief moment when the track switches and the ambience jumps, before any dialogue.

In any case - reassigning the lowest significant bits and putting them as headroom bits is surely exactly the same as shifting the very same dynamic range up.

[Srgtfury]

True, no one is planning on these source levels getting out of control and slamming into the limiter, but we all know it happens and that's why we have limiters. Now, we have NeverClip so that when the source levels DO go beyond the traditional range that we have typically set our input trims, the high level source doesn't slam into anything (except at the very extreme, just shy of the maximum 144db dynamic range that a true 24 bit file can achieve). As it has been explained to me, input trim levels should be set for optimum gain setting for the microphone in use and the anticipated level of the source that is being recorded, this is always the way we have done it. When something un-expected happens and there is sound that would normally fall outside the upper range of your input trim setting, NeverClip allows that sound to fit into a larger dynamic range beyond your input trim setting without hitting a limiter or clipping.

Hi Jeff,

Don't certain sound capturing venues, which it would seem is the majority of venues, at least, emulate and likely surpass the dynamic range of the Zaxcom example? FWIW, nearly every; reality show, sports and news talking head shows and ENG venues, for example, would fall here. So, OTOH, we commonly choose to risk (and, in some cases, accept) clipping or limiter influenced recording, unless the Rider can Flo the trims, by design, while, OTOH, we now have an apparently unique method to record, in Nomad, the aforementioned, erstwhile offending peaks, with near impunity. Doesn't seem that the transmitter, Neverclip feature set offers the same Ad conversion process, for transmitter recording, that Nomad does, however.

Thank you very much

Fury

[studiomprd]

Now that we have NeverClip thoroughly and completely explained, lets move on to Phase Locked Loops...

[deep owl]

I hope Zaxcom makes another video that shows some settings and the effect that they have.

[Jeff Wexler]

I have removed a post that MT Groove made that took the demo clip from Zaxcom's website and posted his edit of that sound file. This was done, as MT Groove had stated, to prove his own theory of how NeverClip works or doesn't work. MT Groove has already demonstrated a lack of understanding how NeverClip works (see the post that states that the only way it can work is to use 32 bit floating point) and manipulating the Zaxcom demo to suit his own theories is irresponsible at the very least. Questioning the methodology, doing your own test demo, looking for clarification of how this technology works is all fine --- altering a demo posted on Zaxcom's website and putting it up here is not fine in my book.

[MT Groove]

Jeff, this is your site and I respect the decision you made concerning deleting my demo. I originally created the demo to satisfy my own curiosity. After analizing the results, I felt I should share my findings here. However, if you didn't think that was appropriate, then you have every right to remove it. I also raised a few questions on my post that you deleted and I felt I made a valid point. Maybe you should have removed the demo but leave my post with valid questions. But instead, you claimed I "demonstrated a lack of understanding how NeverClip works." Glenn claimed the Zax demo clearly demonstrated how NeverClip works. Quite a few people here didn't think so and questioned it as well. Nevertheless, I will respect the way you handled this.