Clipping Comverters During Mastering

[user 133392]

How does it sound?

Thanks for asking. There are numerous sound files embedded within the thread. The later sound files are the current version.

 

[thor.zmt]

One way to get a little more loudness is to intentionally clip an A/D converter.

While it is, it is a bit of crap shooting in the dark.

Some converters had (have?) build intentional "soft clipping" circuitry. Others clip cleanly by design. Still others get a bit nasty when overdriven significantly.

I think the practice of overdriving for "more loud" comes from the days of tape. Tape saturates gradually and usually the "peak" indicators will be lighting at 1% THD in the midband, which is usually the old needle meters "pegging" on peaks.

In reality, you can usually put a lot of extra level on tape. It will compress / distort but not clip hard until you are way, way, way into the red zone.

Some converters handle this much better than others. For the ones that excel at this, what is the internal electronics mechanism used?

The ones that clip "cleanly are all what I call "direct conversion" types, essentially "multibit" converters. The PM M2 clips cleanly.

Most modern converters are based around single or multi-bit delta sigma modulators. These have limited stability and use extensive digital post processing.

Digital filters are subject to inter sample "overs" when operated near clipping. behaviour can be unpredictable. So it is generally better to not use the last top 3dB or so of a converter.

A diode "compressor / soft-clipper" could be build external (I feel KA's projects while very interesting is unnecessarily complex and still lacks the instant zero time constant reaction of tape or diodes).

Maybe make it so that anything below -9dBFS is "clean" and the next 6dB are progressively compressed so that a -3dBFs input sees 2:1 Compression and is actually reduced to -6dBFS and the next 6dB are compressed 3:1 so that a nominally +3dBFS becomes -3.5dBFS and then 4:1 for the next 6dB so a nominally +9dBFS input would end up somewhere below clipping.

This would probably be best implemented in the ADC Driver circuit. Most current ADC IC's use 2V RMS differential 0dBFS input and the driver circuit uses two inverting Op-Amp's as "active attenuator" (in the better designs), so applying these directly in the ADC would means a small extra PCB with suitable (zener) diodes and resistors, to provide the limiting/clipping.

A +20dBu capable line level version could of course also be made.

Thor

 

[cyrano]

Especially with the dawn of 32 bit ADCs, a +20 dB line out would be interesting...

That would transfer the clipping point to the power amp, the speaker, or your ears.

 

[user 133392]

A diode "compressor / soft-clipper" could be build external (I feel KA's projects while very interesting is unnecessarily complex and still lacks the instant zero time constant reaction of tape or diodes).

Thor

Wrong.

Just to make sure we're reading from the same page tell us which capacitor is providing averaging. The Waveulator: Two VCA-based Clipper Saturators - Page 7 - Pro Audio Design Forum

Maybe some imaginary C across R23?

 

[Tubetec]

In mastering your dealing with known maximum levels off tape/daw , while tracking a record its very much more difficult to predict when clipping will occur .


My only experience of diode clippers is in distortion stomp boxes and and guitar amps ,but to be honest its not something I like very much at all , imparting a kind of spitty sizzle to the notes edge .
I guess if threshold voltages/diode curves are adjusted correctly ,the effect can be made more subtle .

The effect of a movil coil meter on audio circuits is well established , old comps like the Federal drive a meter directly in the anode of the output , it adds distortion/compression in the audio path , its arranged as switchable in/out of circuit . The idea that spings to mind is a speaker simulator type effect, need not be tube at all .

Its true to say tape was a very much more forgiving format than the modern program chain in terms of overload , Typically with rock drums you could have the levels so high the meters are pinging off the ends stops on every hit of the snare in record and it would come back at you with off tape with tons of punch but still really smooth . Cymbals and overhead mics are generally given much more headroom when going to tape also because the higher frequency content can generate unpleasentness otherwise .

Theres something about 15 and 30 ips , theres so much tape material passing the heads you really get a lot of margin and abillty to soak up short term overload fast with minimal overhang . Dont underestimate 15ips non Dolby for punk/rock for one minute , the physics involved in the tape path and heads give it superb low end performance , its a bit like the bass end you hear on a an early album by Maddness or the Police .

I wasnt much into the neo synth pop rock revolution of the 80 , but theres the odd few artist and producers whos records hold up to the very best modern can achieve in terms of sound and in terms of the interaction between the musicians which today seems to have been ironed out by the click track and quantise and some dude with his head stuck in a computer sceen all day long , theres very little interaction at all going on in the 'studio' these days . Remote working kept many recording artists going through covid , but its never the same as face to face because of the extra technicalities involved .

By the early 90's when I started up in the studio game Dolby SR had come in , tape noise was no longer an issue , you didnt need to punch the levels anymore .
I remember at the time comparing the recordings we got from different sessions at different tape speeds with /wo Dolby etc , the non SR session they typically drove levels to tape to the point where distortion on the replay became objectionable , then wound it back a notch , you werent even looking at the meter ,because it tells nothing about the point where that distortion becomes intollerable on a given source, only the ear could be the judge of that .

 

[user 133392]

I went with a VCA/Modulator approach to provide adjustability of Threshold, Ratio (hardness) and Make-Up gain using familiar controls and in a way that provides simple stereo control coupling.

Thor - If you only looked at some of my earlier drawings and experiments you may have seen a small 2200 pF "2n2" at the Ct output of the THAT4316 for an MI version and a 1 nF on the 2252. Those Cs are small enough they don't provide any signal history as the toneburst waveforms confirm 0 attack 0 release. Those caps are only there to LPF HF edges from glitchy (out of zero crossing) FW log detectors that were never designed to operate without a Ct.

That C doesn't exist once I used a cleaner rectifier. Otherwise there's no detector averaging anywhere in any of the versions so your statement about it not being 0 attack 0 release is completely inaccurate.

 

[Viitalahde]

I don't clip converters much these days. Clipping can be done very well in digital if needed. I don't sweat it if a peak or two clips, but IMO you're painting yourself in the corner if you clip a lot. It affects cutting the master lacquers too if the sound's too distorted.

The only times I do clip in A/D are the times where "that sound" is clearly needed. Some mixes like to be pushed. But it's getting rare.

 

[user 133392]

I agree.

I did some experiments where I reduced the crest factor of a song to various degrees with brick wall limiting, inverse-RIAA'd the results and then normalized the peak modulation of the I-RIAA'd files for them to be equal. In all cases the clipped samples, when RIAA de-emphasis was applied, returned progressively lower overall RMS levels.

Clipping/brick wall limiting worked against final RMS level. For vinyl it should be avoided and if you use it you sure are painting yourself in a corner.

I do however want to run an experiment where I first apply I-RIAA, mildly clip pre-emphasized transients and then apply RIAA to see if it can control sibilance.

What I'm doing can be done in DSP for sure.

 

[thor.zmt]

Wrong.

My bad. You are correct.

I saw a complex sidechain and some capacitors in there and did not look more deeply.

I really dislike adding any VCA in the signal path and it's unnecessary complex for what a simple diode circuit could do as well (or better?) when operating and does worse when the diodes are reverse biased.

I went with a VCA/Modulator approach to provide adjustability of Threshold, Ratio (hardness) and Make-Up gain using familiar controls and in a way that provides simple stereo control coupling.

I do not see how this is any issue with a diode based circuit. But who am I to stop you from using an overly complicated circuit to do a job less well than a more simple one could do...

That C doesn't exist once I used a cleaner rectifier. Otherwise there's no detector averaging anywhere in any of the versions so your statement about it not being 0 attack 0 release is completely inaccurate.

Yes, agreed, my bad.

What I'm doing can be done in DSP for sure.

EXCEPT if we want to (for example) protect an ADC or analogue input power amplifier from being clipped hard, which is precisely where diode based circuits come into their own.

My only experience of diode clippers is in distortion stomp boxes and and guitar amps ,but to be honest its not something I like very much at all , imparting a kind of spitty sizzle to the notes edge .
I guess if threshold voltages/diode curves are adjusted correctly ,the effect can be made more subtle .

Correct. A correctly designed diode soft clipper is not "grungy", here how it was done by Bob Cordell:



Op-Amp A compensates the diode voltage and thermal drift. Op-Amp B creates the opposite threshold voltage. The series resistor in front of the circuit determines how "steep" the compression is.

The higher the resistor value the greater the compression and the harger the clipping, the lower the resistor, the lower the compression and the softer the clipping.

TBH, the Op-Amp's BC used are a bit superfluous, the same circuit could be done just with resistors, capacitors and diodes. Threshold and aggressiveness of compression/clipping could be made adjustable to match a given ADC and could range from "no soft clipping" to "ultra-soft" tube tape recorder like.

Here how the KK fixes nasty clipping:


From this page:

Review: Super GainClone & Super GainClone w/ Klever Klipper by Cordell

The issue with a soft clipper of any kind is that it introduces relatively distortion at levels that would normally be super clean.

To Measurebaters that is clearly a no-no...

The effect of a movil coil meter on audio circuits is well established , old comps like the Federal drive a meter directly in the anode of the output , it adds distortion/compression in the audio path , its arranged as switchable in/out of circuit .

In the 80's when I was using mixers and tape machines with mechanical meters these long had gained meter drivers... In fact, many had VFD or LED based meters.

Thor

 

[JohnRoberts]

I agree.

I did some experiments where I reduced the crest factor of a song to various degrees with brick wall limiting, inverse-RIAA'd the results and then normalized the peak modulation of the I-RIAA'd files for them to be equal. In all cases the clipped samples, when RIAA de-emphasis was applied, returned progressively lower overall RMS levels.

Clipping/brick wall limiting worked against final RMS level. For vinyl it should be avoided and if you use it you sure are painting yourself in a corner.

I do however want to run an experiment where I first apply I-RIAA, mildly clip pre-emphasized transients and then apply RIAA to see if it can control sibilance.

What I'm doing can be done in DSP for sure.

Funny, when I was designing dynamic processors I made a gadget that could increase the crest factor of dynamic music. Basically a variant tone burst modulator that could use music instead of a sine wave for the source. It sounded like crap but was very useful for stressing dynamic processors with extremely dynamic musical signals. I killed a lot of brain cells last century trying to make transparent NR compressor/expanders for tape recording.

===

I went back to the 2007 beginning of his project over at Wayne's forum to try to determine the design goal. It sounds like it was to be a good (?) sounding distortion effect. Of course design targets evolve over years of melting solder and bouncing ideas off the world.
===

The loudness wars, perhaps still with us, precipitated a lot of serious design effort in the broadcast industry, while they received credible results using analog technology. It seems a rigorous application of DSP could be even more effective at concealing artifacts and perturbations from sudden level changes. This could include a very high quality look ahead delay.

Processing for vinyl mastering could also take advantage of non-real time processing.

JR

PS: I did not do a rigorous search of modern technology, I bet design work in this area (loudness enhancement) continues taking advantage of modern technology.

 

[user 133392]

The issue with a soft clipper of any kind is that it introduces relatively distortion at levels that would normally be super clean.

Are you talking about low-level signals in the middle of the waveform, below threshold, which should be untouched?

 

[kags]

Thanks for that input. I work a lot as an assistant to a mix and mastering engineer who is constantly insisting that they are behind us, and who generally masters to -14dB LUFS, but it doesn’t represent what either of us see when checking reference material.

I do think that some genres sound better pushed, but rarely do I think a track sounds best at -7dB LUFS. Obviously people having clipped has become part of a genre’s sound, as much as the hardware L2 was part of the sound of early 2000s masters.


The question for me is, what is it in terms of sound that you hear as beneficial about clipping as opposed to limiting? I’ve been mixing some pop/hip-hop of late, and I’m trying to understand why I should want to clip.
A decade ago, I knew why I did it on less transient sounds (to get a distorted bass sound, for example), but obviously clipping a whole mix or subgroup is a different thing.
Also, is anyone here then using clipping plugins (eg. SIR, Kazrog, that new acustica one), or is it “an analogue only trick”?

I use clipper plugins these days, instead of clipping converters, especially since I very rarely run an analog mastering chain anymore. I find that a clipper before my final maximizer helps transparently shave peaks before the maximizer. Maximizers tend to have some tonal effect on the music, and that tonal effect changes with the amount of limiting it is doing. 1 or 2 dB of clipping can be completely (almost) invisible. In cases where I need a very high level master, I'll use two maximizers in series instead of a clipper followed by a maximizer. I find the Kazrog K-clip useful, as well as the Stealth Limiter before my Maximizer. Again, I'm using the maximizer for level and tiny bit of tone and the preceding clipper is for an extra (transparent) couple dB.

While mixing I've started using more clippers (or just transparent limiters) on drum and percussion busses so that my stereo bus doesn't contain peaks that might push a maximizer too much. We don't have tape to absorb those pesky peaks anymore! I can usually clip off a few db of transients on a live drum bus that frees up some overhead in my mix bus.

In short, clipping can be a transparent and easy way to shave off a few dB of occasional transient peaks. Mastering limiters (maximizers) set your ceiling, add some tone and provide some dynamic limiting control with a lot more options.

Analog clippers (even those in high-end mastering converters) add a bit of harmonic content that may or may not be appropriate for a given project. It was a great way to go when analog signal paths were the norm...

 

[zonkers31]

Thanks for that input. I work a lot as an assistant to a mix and mastering engineer who is constantly insisting that they are behind us, and who generally masters to -14dB LUFS, but it doesn’t represent what either of us see when checking reference material.

I do think that some genres sound better pushed, but rarely do I think a track sounds best at -7dB LUFS. Obviously people having clipped has become part of a genre’s sound, as much as the hardware L2 was part of the sound of early 2000s masters.


The question for me is, what is it in terms of sound that you hear as beneficial about clipping as opposed to limiting? I’ve been mixing some pop/hip-hop of late, and I’m trying to understand why I should want to clip.
A decade ago, I knew why I did it on less transient sounds (to get a distorted bass sound, for example), but obviously clipping a whole mix or subgroup is a different thing.
Also, is anyone here then using clipping plugins (eg. SIR, Kazrog, that new acustica one), or is it “an analogue only trick”?

In my experience, when clipping slightly in the analog domain, the added distortion makes things sound a little brighter. I have done a/b tests not explaning what is going on to the listeners, and almost all people pick the clipped music as sounding better. Just my .02

 

[thor.zmt]

Are you talking about low-level signals in the middle of the waveform, below threshold, which should be untouched?

No, below the actual threshold a "soft clipper" should be completely transparent.

I am talking about signals above the threshold but below the clipping level of the device behind the soft clipper.

For such levels distortion is increased.

Thor

 

[user 133392]

No, below the actual threshold a "soft clipper" should be completely transparent.

I am talking about signals above the threshold but below the clipping level of the device behind the soft clipper.

For such levels distortion is increased.

Thor

Agreed.

 

[iomegaman]

This discussion is way over my head so my input is rather questionable...

My Apogee Symphony MKII has a built in "Soft limit" which is obviously different from a clipper but may produce some of the same sonic opportunities the OP is looking for...

In a purely digital plugin world the Pulsar Modular P42 is rather good at this...lots of ways to drive the saturation and signal to clip.

I have experimented with 32 bit float files and clipping is fairly non-existent in the normalized files regardless of how clipped it was recorded...again a different beast.

Are we looking for just "louder" or perceived louder?

I imagine the mastering guys get all kinds of hot driven tracks that need to be tamed so there has to be some professional analog device and/or ADDA and/or algorithm to accomplish this...MAS Overstayer? Weiss DS1-MK?

What happens when you max out the sample rate and clip?

 

[Gold]

I have experimented with 32 bit float files and clipping is fairly non-existent in the normalized files regardless of how clipped it was recorded...again a different beast.

The advantage of floating point is that it becomes virtually impossible to clip. I don't use a plug in for clipping ITB. I just raise the level of the 24 bit file until it clips to my satisfaction. Then I bounce the file and the clipping is baked in.

One well known mastering engineer who gets credit/blame for pushing the level is known to just raise the level to where he wants it. Artifacts and all. Then his assistant goes through and un splats the splats with something like RX. No limiter used.

 

[living sounds]

I don't use a plug in for clipping ITB. I just raise the level of the 24 bit file until it clips to my satisfaction. Then I bounce the file and the clipping is baked in.

I use a plugin for clipping ITB because I can then back off the amplitude a little afterwards and then bounce the file. I don't like having actual overs (samples at 0 dB) in digital files.

Intersample peaks are another, different matter.

Also, converting those clipped/massively limited files to mp3 can cause very significant overs in the output.

 

[cpsmusic]

The question for me is, what is it in terms of sound that you hear as beneficial about clipping as opposed to limiting? I’ve been mixing some pop/hip-hop of late, and I’m trying to understand why I should want to clip.
A decade ago, I knew why I did it on less transient sounds (to get a distorted bass sound, for example), but obviously clipping a whole mix or subgroup is a different thing.
Also, is anyone here then using clipping plugins (eg. SIR, Kazrog, that new acustica one), or is it “an analogue only trick”?

I come at this from a background in music, some mixing and even less electronics! From what I can see, the main reason people are using clipping (during mixing) is to control pokey transients. Limiters tend to make them sound a bit soft and spongey while clippers can actually make them sound louder (due to the high harmonics and aliasing!) even when they're quieter. So anything that's fast and clicky like drum transients or acoustic guitars is a candidate for clipping.

With respect to entire mixes, clipping adds a "rough brightness" which I think most people perceptually equate to loudness and energy. And as has been mentioned, it gives the mix more "loudness potential" if that's what's desired.

For those interested, here's an analysis of some contemporary hard rock/metal



Obviously lots of clipping going on!

Cheers!

 

[user 37518]

Are people still doing this monstrosity? I thought that the extreme loudness of the 00's and 10's was no longer being pursued so fiercely. Back in the 00's and early-mid 10's, I used to see and sometimes do this; I thought that it was something better to be buried as a dark side of history.

 

[user 133392]

I start my thread on the subject with:

I'm not a big fan of over-hyped saturated mixes. They're fatiguing to listen to. I call it "ear burn."

Despite that I decided to experiment back in 2013.

What I found is that by "gently" rounding peaks the reduction in peak level, and the resulting reduction in crest factor, allows the unadulterated RMS energy in the "middle" of the waveform to be increased - to produce the same peak output level - by simply increasing gain. It sounds compressed but the "compression" is just gain and added density.

It may be an over-simplification but for complex material there's not much loudness or energy in the transients. The loudness is the band of signal in the middle region of the waveform. When the converter is driven hard it's the same process though perhaps not as rounded and graceful.

I started these experiments ten years ago trying to create color and what I got was loudness. My recommended application for it is converter protection limiting in moderation and when the project calls for it it could be used in greater amounts for a louder product. Again I don't consider clipping/rounding to be vinyl friendly.