Do ME's purposely clip?

[kylen]

i.e. they don't do it in software. they let the ADC do it for them. most go through this process for format-conversion anyway (instead of running a sample-rate converter, for example) and to run outboard effects.

yes, I'm starting to see that bleyrad. I should say hearing that !

I'm beginning to move in the outboard hardware direction for a few items and have started my roadmap of upgrades with the dbx386 and the behringer DEQ2496. It's my first step outside of the box so I kindof went for what I could afford being a garage DIY guy.

I'll probably end up using a hybrid setup - part CEP or Sonar using hardware plugins. Ha Ha ! Sounds better already I reamped a program thru the dbx386.

kylen

http://www.tapeop.com/magazine/bonus/dbx386.html
http://www.behringer.com/02_products/prodindex.cfm?id=DEQ2496&lang=eng

 

[nessbass]

I'm totally confused by that picture

Sorry, I'm afraid I somewhat hijacked this thread. What is shown in the picture has admittedly only little to do with the original question fenix has asked. So, there are basically two different discussions going on. One about clipping as a means to master even hotter and one about a strange waveform.

I was trying to find examples for the clipping that fenix was talking about and had a look at the Audioslave record, as I recalled it being pretty hot-mastered. So I looked at one of the kick drum transients (as these are usually the most squashed) in SoundForge and then posted what I saw, because I've never seen something like this before.

Originally posted by bleyrad
... i'm pretty sure that clip only represents an extremely small fraction of a second.

What we're dealing with here is only 31ms long! So it's really just the transient being fucked up.

I already PMed some of the pro guys here about that, and I think I will get my answers there, so let's get back to the original discussion, Do ME's purposely clip?

Much unlike the other picture I have yet another one for you, this time really just clipped. It's from RHCPs Around the World. As you can see in the lower right corner of the picture, the highlighted part is 70 (!!!) samples long.

Here you go:

 

[Outlaws]

Man.....you guys wanna talk about some fucked up recording.........

The ending of Johnny Cash's cover of Hurt was massive digital distortion!

I bet if you guys put a picture of that up it will trump even the RHCP one

 

[sjoko2]

Interesting thread.
A couple of points:

Do mastering engineers clip things on purpose? Nope, definitely not. However, there are SOME occasions where clipping might occur in a number of consecutive samples, to low a number to affect the audio. In that case it will depend on the mastering engineer. He either lets it ride or he won't.

Second - and I believe this is a very important point. I have, on many occasions, mastered a track (and, as we commonly do here, used analyzing software to detect any possible clipping), and, even tough the track was perfect, found that once the CD is burned - it clips. This would be a master disk, burned at high accuracy. Guess what happens if such a track goes through a duplication plant? The clipping would be worse.
I am dealing with one such thing at the moment - the opening of a track clips out, only when transferred to a CD - it does not happen at its original 24/96 state, it does not happen when converted to 16/44.1, but always on a CD, even with the part eq'd and reduced in volume. The reason? Its not clipping.......it a harmonics issue the CD format cannot deal with. That simple. A number of tones / frequencies and waves which, when put together on a low-resolution format, cause distortion. As it’s not possible to pull the elements of the mix apart in mastering, the only solution is to return to the mix and fix it there.
Furthermore – I am incurring more and more examples of a duplication facility basically fucking up. Some lunatics seem hell bend on “upping” the levels at that stage and boosting them. Even a 0.1dB boost in levels might lead to clipping. I always ask for a sample of the glass master for approval and have returned 4 in the last 2 months for “inaccuracies”. In other words, its not always the ME’s fault, there can be many reasons.
For instance, I receive a whole bunch of tracks with tops so square you could stack them 50 high – for mastering.

Next one. The “flat tops” as seen in the examples. Don’t forget this does NOT have to be clipping, it simply can be a limiter. I’m doing some dance/trance stuff at the moment, and I can guarantee you that the tops are all flat due to the huge amount of processing on the kick mainly, but it does NOT clip out on a disk, its simply the waveform.

Finally the “saw” wave on the audioslave kick. Its virtually impossible to tell you exactly what caused it, as you took the sample of a finished CD. However, it is very likely the effect of wind on a dynamic microphone placed in a kick drum. Most dynamics do not have the capacity to withstand the SPL a kick generates, but despite this they are used to catch the attack. The sound will be captured but a microsecond later the sound will be followed by wind, which then creates a pressure level on the diaphragm which it is not capable of handling. This can lead to a saw-like waveform, which is not necessary detrimental to the overall sound (still……………I’d rather do without it). There are now many good solution available to avoid this, like for instance AT’s AE2500, the dual element kick mic, or, even better perhaps, the superb Russian build kick drum mics from Elation. (the latter, in a matched pair inside the kick, with the Yamaha sub-kick on the outside, produce the best sounding kick I’ve ever heard, without any kind of processing.

 

[TripleM]

The “flat tops” as seen in the examples. Don’t forget this does NOT have to be clipping, it simply can be a limiter.

Good post - I learned quite a bit from it.

I didn't realize there was a difference between "flat tops" and clipping. I thought they were one and the same thing. Isn't a flat top a situation where you have samples that extend the medium (16 bits) to its maximum? Whether or not they were put there by a limiter or anything else, isn't it still clipping? And lastly, does clipping have to be "audible distortion?"

Thanks in advance.

 

[Son of Mixerman]

Interesting thread.
A couple of points:

Do mastering engineers clip things on purpose? Nope, definitely not.

Im going to disagree. Since the L2 hit the market the number of ME's purposely clipping has increased significantly. Why? Because the L2 does it alot easier. See the Audio Slave CD.
There is a whole thread at PSW dedicated to Vlad purposeful clipping.

However, there are SOME occasions where clipping might occur in a number of consecutive samples, to low a number to affect the audio. In that case it will depend on the mastering engineer. He either lets it ride or he won't.

You mean the label not the ME, for a majority of the cases. Seems likes ME's are being subjected to some sorta of slavery or S&M thing.

Second - and I believe this is a very important point. I have, on many occasions, mastered a track (and, as we commonly do here, used analyzing software to detect any possible clipping), and, even tough the track was perfect, found that once the CD is burned - it clips. This would be a master disk, burned at high accuracy. Guess what happens if such a track goes through a duplication plant? The clipping would be worse.

F u d g e !!

If your getting clipping on the CD its because you put it there somehow What kinda QC are you doing on your discs? Gotta a problemo in your chain somewhere.?The clipping should not get any worse down stream unless there are some serious QC problems. Have you been to the Mastering webboard lately? I recently went through alot of the same problems. Glenn and the crew were very helpful in helping figure out the QC stuff. Which discs are you using? My problem was a 0dbfs intolerance. Basically anytime I got near zero the player I had didn't like it, it froze or went into an infinite spin mode. If you figure out a good anal retentive QC procedure you can track it down to the weakest link. Mine was the player, Brad Blackwood checked my disc and it was fine everywhere but in my player. I recently saw a guy with problems with a $5000 digi-eq, so you never know where little things can go wrong. System QC is one of the most overlooked yet aggrevating things to work out. IMO

I am dealing with one such thing at the moment - the opening of a track clips out, only when transferred to a CD - it does not happen at its original 24/96 state, it does not happen when converted to 16/44.1, but always on a CD, even with the part eq'd and reduced in volume. The reason? Its not clipping.......it a harmonics issue the CD format cannot deal with. That simple. A number of tones / frequencies and waves which, when put together on a low-resolution format, cause distortion. As it?s not possible to pull the elements of the mix apart in mastering, the only solution is to return to the mix and fix it there.
Furthermore - I am incurring more and more examples of a duplication facility basically fucking up. Some lunatics seem hell bend on ?upping? the levels at that stage and boosting them. Even a 0.1dB boost in levels might lead to clipping. I always ask for a sample of the glass master for approval and have returned 4 in the last 2 months for ?inaccuracies?. In other words, its not always the ME?s fault, there can be many reasons.
For instance, I receive a whole bunch of tracks with tops so square you could stack them 50 high - for mastering.

Heh heh... see above verbage. After years of learning from guys like Dave Collins, Brad Blackwwod and Bob Katz Im finding myself backing off to -3db peak, its sounds better to me and I can actually use this thing called a volume knob that came with my stereo Duplication facilites used to rely on their own QC dept to make sure discs were right,now they don't even look. Time and money issues means whatever they get they repro, if something goes wrong...the Band or the ME have to car enough. Its horror sheer horror. Everything is just falling apart.

Just hand on a little longer.

SoMm

Finally the ?saw? wave on the audioslave kick. Its virtually impossible to tell you exactly what caused it, as you took the sample of a finished CD. However, it is very likely the effect of wind on a dynamic microphone placed in a kick drum.

It was caused by Rick Ruin and Vlad the Impaler. How does having the 2 buss ran through an L2 before mastering only to have Vlado L2 it again help, Im surprised Rich let Rick mix it to those levels. Lunatics as you say.

 

[sjoko2]

Good post - I learned quite a bit from it.

I didn't realize there was a difference between "flat tops" and clipping. I thought they were one and the same thing. Isn't a flat top a situation where you have samples that extend the medium (16 bits) to its maximum? Whether or not they were put there by a limiter or anything else, isn't it still clipping? And lastly, does clipping have to be "audible distortion?"

Thanks in advance.

I could create a track for you, using mastering limiter / compressors, where you'd see the waveform at 16/44.1 with 90% of all the tops completely flat, all topping out at 0.1dB below 0dB, without any distortion. Naturally such a track would not have any dynamics either. In the same way I could create such a track at any given dB figure under 0.
Clipping occurs when a sample exceeds 0dB. In a digital medium this means distortion - always. However, as the 16/44.1 CD medium is pretty low resolution, consecutive "overs" of between 1 and 3 samples are largely not audible.

 

[TripleM]

Aha. The "90% samples at -.1db" clears it up a bit for me.

But then you said something else I've heard before but didn't understand. That is, "clipping occurs when a sample exceeds 0dB..." How does something EXCEED 0dB? Isn't 0dB a sample with "all binary 1's"? I can see how you could EQUAL 0dB for an extended period of time (and therefore distort). But I've never understood how you could EXCEED 0dB. You've used up all your bits at 0dB right? Or wrong? Proly the latter huh?

Sorry for being an annoyance but I've always had this question and now is my opportunity to get it answered.

And again thanks.

 

[kylen]

Yes, nice post sjoko2 !

I think we, or me (Ha Ha!) , are running into a slight 'definition of terms' issue. What on-line audio dictionary or reference guides do you folks use on this forum ?

For me 'clipping' is an active state - something is causing a red 'overs' light to go on - present tense.

'Clipped' for me means it could've been shoved thru the digital ceiling at some point, flattened and may or may not have been readjusted to -0.1 dB or whatever. If I run the stats using Cool Edit Pro it shows me possibly clipped pieces of the wave. Sometimes a spectrum will show other forensic evidence especially if I know the 'signature' of my own limiters.

'Clipper' is something that the broadcast guys use to prevent 'clipping' - there's a couple of examples of equipment that do that in the links I posted. My thought on those are that they are pretty similiar to a very fast attack 'limiter' with no overshoot. My other thought is that the digital hardware ones can limit to sample level also with no overshoot.

In software T-Racks makes a 'Clipper' but it is not a fast attack device like its' so-named hardware brothers and allows overshoot. In fact it's more of a saturation device.

So, for me at least, there is a lot of noise surrounding the actual 'clipping' vs 'limiting' thing - Ha Ha big surprise ! I can see them being interchanged.

I'm leaning towards 'clipped' to mean something originally associated with a distribution medium such as CD or broadcast airwaves. It does creep into the home recording space from time to time though.

As far as audible distortion goes that TripleM mentioned there's another definition - what's 'distortion' ? Ha Ha In the case of 'overs' I think I know what you're thinking - I was too. Hmm...

I'll see what links or thoughts other folks have ...
kylen

 

[TripleM]

what's 'distortion' ?

True. It's kind of like defining pornography. You know it when you see it.

There's good distortion (like a Les Paul thru a Marshall) and bad distortion (like when you clip ).

 

[kylen]

There is a whole thread at PSW dedicated to Vlad purposeful clipping.

Thanks Son of Mixerman ! I'm over ther now reading - the topic is very interesting to me for some reason.
kylen

 

[bleyrad]

hold it a minute here!
i think there is some confused information going around.

- In the digital world, there is no such thing as an 'over.' A clip IS A FLAT TOP at 0db. they are one and the same thing! Put simply, if digital software sees a level greater than 0db, it shoves it back down to 0db. That's it.

- To the claim that a wave with 90% flat tops will have "no distortion": this is impossible. Flat-topped waves ARE distorted by their very nature. Think synth-generated square or saw waves. Although it is based on a pure root sine wave of a low frequency, its shape naturally contains the sum of many other higher-frequency harmonics. Those harmonics are 'distortion', and that is EXACTLY what a digital clip (flat-top) does - turns a round wave into a square wave, which introduces unwanted harmonics.

(there is a bit of an exception to this rule with very old CD players that 'blip' when presented with a 0db signal, but that doesn't even mean the signal has to be clipped! it could simply be the legitimate top of a perfectly fine sine wave causing the incompatibility with the player. this kind of blipping distortion is NOT the same thing as a digital clip.)

- Many legitimate ME's, even really high-end pro's, use clipping very often as a way to increase loudness. turning the top few samples of a waveform square CAN go completely unnoticed, if done right. once again: that's all that clipping is.

The interesting thing about digital clipping is that it's impossible to know what has really been 'clipped' because a 0db sample isn't necessarily one that's been forced down (clipped.) All that those clip-lights tell you is the software's "best guess" - usually around 3 samples in a row at 0db will show up as being clipped. But frankly, I could generate a sawtooth wave with its peaks at 0db, without clipping it at all, and the cliplights will turn on. That said, they're usually pretty accurate.

- The L2 limiter isn't a clipper. It's a limiter. It retains the natural undistorted shape of the waveform, so long as you use it with a long enough release time (or ARC.) You can run it very hard without getting ANY flat-topped samples. To sum up: look-ahead limiters like the L2, when used properly, are not what are causing clips in media. Generally, clipping is done by the ME on purpose.

 

[kylen]

bleyrad - thanks for the cold shower ! A vein was startin to pop out the side of my neck too ! Ha Ha...

I'm liking what you said - thanks for the concise info. Much what fenix' original point was. I'm also liking the flat-top square-wave harmonics connection - yes it would seem that it should change the sound.

I'll take a temporary exception to your comment about software pushing digital 'overs' down to the ceiling of 0dB until I check something out in Cool Edit Pro using 32 bit float. I think the majority of apps will do that in 24bit. It's of minor consequence though - like who wrote Macbeth, Shakespeare or Thomas Marlow kind of thing...

I'm going to think about this for a bit of time and post later if I have anything to add... I already have some cool experiments I want to run - thanks everybody !

kylen

 

[sjoko2]

F u d g e !!

If your getting clipping on the CD its because you put it there somehow What kinda QC are you doing on your discs? Gotta a problemo in your chain somewhere.?The clipping should not get any worse down stream unless there are some serious QC problems. Have you been to the Mastering webboard lately? I recently went through alot of the same problems. Glenn and the crew were very helpful in helping figure out the QC stuff. Which discs are you using? My problem was a 0dbfs intolerance. Basically anytime I got near zero the player I had didn't like it, it froze or went into an infinite spin mode. If you figure out a good anal retentive QC procedure you can track it down to the weakest link. Mine was the player, Brad Blackwood checked my disc and it was fine everywhere but in my player. I recently saw a guy with problems with a $5000 digi-eq, so you never know where little things can go wrong. System QC is one of the most overlooked yet aggrevating things to work out. IMO

S of M, you are a very brave man
Without meaning to talk down to you, your information comes from a lot of reading and asking. All very good things indeed, because that's one way of learning. However... a little practical experience would help as well. If you had that, you would know that its possible to create harmonic that, when transferred to the CD medium, sounds like clipping and popping, as the media cannot cope with the sound. An effect you can also recreate by, for instance, recording a particular note on a Grand, and dubbing it with the same note, detuned by an extremely very fine margin.
If you would burn the same files to a different medium, for instance encode it to DVD format, the clipping effect would not occur.
Regarding our QC
Sony's HD film demonstration program is going around the globe at the moment, showing movie production professionals what can be achieved by using high definition digital film and audio. I wonder why the recording, mixing, audio post production and mastering were done here? After all, Sony has some of the top facilities? This place was described in a pro mag as "one of the most advanced digital facilities in the US" recently. Perhaps that has something to do with it? Or maybe its that, while everyone uses routing and converter technology with at best between 106 and 112dB headroom (and that's just in their converters - NOT through the system), we manage and maintain well over 132dB throughout the system, and in some front-end parts of the system achieve over 150dB? One thing is for sure - with a system like that any faults are much easier to spot, be it with ears, eyes or through software.

 

[bleyrad]

I'll take a temporary exception to your comment about software pushing digital 'overs' down to the ceiling of 0dB until I check something out in Cool Edit Pro using 32 bit float. I think the majority of apps will do that in 24bit. It's of minor consequence though - like who wrote Macbeth, Shakespeare or Thomas Marlow kind of thing...

sorry, i should have mentioned the floating-point exception.

basically, floating-point formats contain an exponent that says what range the bits currently represent.

for this reason, it's almost impossible to clip while in 32-bits. that's why it doesn't matter a whit if you're seeing the individual track clip lights go on while you're mixing.
however, pay attention to the master fader's clip light, because if you're clipping there, it's gonna be sent as flat-topped 0db samples to the soundcard in 24-bit integer - which won't prevent the clipping.


in any case, i was talking about final-format integer clipping. consider that, after all, we're talking about mastering, where it's the only format that really applies.

 

[bleyrad]

as an example:

Song with clipped samples


This is a (somewhat shitty) mix I did a couple weeks ago.

I applied some pretty relaxed L2 limiting with ARC then checked the resulting waveform. There were no detected clipped samples over 2 in a row.
Then, for a little bit of an extra volume boost, I boosted the resulting file by 0.6 db.

Now there's clipped samples all over the place. I don't remember for sure, but when I checked, I think there were quite a few snare hits that had a HUNDERED samples in a row clipped.

but guess what?
i couldn't notice any difference between this and the original, non-clipped version. even those snare hits sound the same.

 

[sjoko2]

hold it a minute here!
i think there is some confused information going around.

- In the digital world, there is no such thing as an 'over.' A clip IS A FLAT TOP at 0db. they are one and the same thing! Put simply, if digital software sees a level greater than 0db, it shoves it back down to 0db. That's it.

- To the claim that a wave with 90% flat tops will have "no distortion": this is impossible. Flat-topped waves ARE distorted by their very nature. Think synth-generated square or saw waves. Although it is based on a pure root sine wave of a low frequency, its shape naturally contains the sum of many other higher-frequency harmonics. Those harmonics are 'distortion', and that is EXACTLY what a digital clip (flat-top) does - turns a round wave into a square wave, which introduces unwanted harmonics.

There definately is some confusion here
You are correct that a clip is a flat top at 0dB.
You are however incorrect in stating that all flat top waves are distorted by nature. It is perfectly possible to limit material at a certain setting under odB, but at the same time compress/push it to such an extend that the total creates a flat top. With the right mastering gear you can do that, without creating distortion, but not forgetting that you create a huge blob of sound without any dynamic range whatever.
I am at this moment in the middle of mastering a trance/dance project on the dsd8, a projec that arrived at 24/48. I converted it to analogue and put it through the Pultec / optical mastering compressor chain and returned it into the dsd. This material is intended purely for danceclub use, it needs to be loud as hell, no dynamics required. tops flat - overs, or clips, whatever you'd like to call them - nil.
In other words, there is a difference between a track "clipping" / distorting, and the sum of many tracks saturating the available space up to OdB to the maximum.

 

[kylen]

Cool posts bleyrad - this thing is beginning to make more sense now.

I'm getting more interesting reading and independent confirmation from reading this stuff that SOM recommended:

http://recpit.prosoundweb.com/viewtopic.php?t=8193&postdays=0&postorder=&start=30

Looks very interesting as long as it sounds good ! Especially the comment one of the MEs made about the bass component sounding a little better clipped than limited. I seem to have a bit of trouble with that using my software limiters and multi-bands. We'll see though - experiments in progress (following my room treatments upgrade!).

OK fenix - great thread you started here ! I owe you a chili dog next time I'm in Cincy ! Ha Ha - or are you a 4-way guy ?
kylen

 

[nessbass]

Sjoko, thanks for taking the time and having a look, thanks.

Originally posted by kylen
like who wrote Macbeth, Shakespeare or Thomas Marlow kind of thing...

I guess you're talking about Christopher Marlowe, right

David

 

[bleyrad]

how close are you zooming in to look at these 'flat tops'?

i have this feeling that what you're calling flat topped is the overall view of the whole song, or a portion of it, that looks in the waveform display like it's totally maxed out and flat.

if you zoom in to the sample level though, i think you will find that the samples are not all flat.
they cannot possibly be without creating harmonics.

look-ahead limiters and compressors do not create flat-tops unless their release time is too short.
and in any case, if you're dealing with non-look-ahead jobs, or your release is too short, it is distortion, regardless of whether or not it's particularily noticable.


flat tops are clips are distortion. period. that's all there is to it.