What is non-linear processing, and why does it matter? - Episode #79 | The Mastering Show Podcast

What is non-linear processing, and why does it matter?
In this episode we dive deep into the topic with our guest Dan Worrall, and talk about: what non-linear processing is, why it can sound great and when it can sound terrible, and how to know the difference. Along the way, we also discuss
* Is reverb non-linear ? And why it’s confusing
* How linear processing makes top-down mixing a good idea
* The difference between saturation and distortion
* …and how to use them creatively
* The difference between a perfect 5th and a power chord
* ...and how to play chords on a trombone
Full show notes on our website
themasteringshow.com/episode-79

[@DanWorrall]

Thanks for having me, I enjoyed our little chat :)

[@RonnieVaiArovo]

Always good listening to your clean, un-saturated voice ;) & Wise knowledge 👌

[@ProductionAdvice]

Thanks for joining us, it was great to have you!

[@sundamusik]

This one is Classic!!!

[@zwsh89]

@Dan Worall Please do more public interviews! These and your videos are the closest thing we get to taking lessons from you!

[@dantig]

Would you consider ring modulation a distortion, Dan? Thanks

[@LS-ud7do]

I'm a subscriber to yourself and Dan, I'm also a Reaper user. It's difficult to express my gratitude over this podcast (and the channels you guys run). KZbin has some amazing content for begginers but for those who strive to go further; you're making a genuine, tangible, positive impact on our lives. I've dealt with addictions and problems that pulled me so far away from audio I was nearly completely lost. The information you pass down isn't just techniques and ideas, it's a roadmap to the life I'm working towards. Thank you Ian and all involved.

[@ProductionAdvice]

Wow, high praise indeed! I’m really sorry to hear you’ve struggled like this, and so pleased if it helped in some way 🙂

[@Dani-El.]

Wow, Dan is really clever isn't he. Good at explaining complicated ideas and processes.

[@zwsh89]

I hope fabfilter appreciates what they have, and is compensating you properly for the huge boosts in sales over the years that Dan has been single handedly responsible for thanks to his ingeniously creative and original lessons and teaching style.

[@danielweirdone]

Thanks to Dan and his 'I won the loudness war' track which I've tried to find on Spotify yesterday, I've found this podcast. Couldn't be happier. Great work!

[@leonkrimov6955]

Ian, Dan and John, thank you very much.

[@returnofthemilk]

This is brilliant. I had to pause and look up loads of terminology. Lots of learning to get stuck into. Thanks very much.

[@ChemaMrua]

I think you talk about multiphonics. A technique that wind players performance almost always in a contemporary classical music context. By changing the embouchure and air preassure you are able to push up certain harmonics to even the fundamental's level, obtaining clear and loud intervals (not full chords however). Not easy, and very often with unpleasant results, depending of the timbre of the instrument (the partials each one emphasise the most).

[@rockstarjazzcat]

Hey there Ian! Super fun to hear Dan Worrall interviewed! Best, Daniel

[@ephjaymusic]

AYYYY!!!! So cool to hear you two on the podcast! Thank you!

[@fatalitydnb]

Wow that was incredibly thorough!! 👍🏼

[@kaori-3882]

Great talk, please consider inviting Dan again! I think that TC reverb is non linear in the sense it never sounds the same.. rendering the same part never nulls.. can be a bit annoying actually :)

[@goldan_tv]

Algorithmic reverb = non linear Convolution reverb = linear (usually)

[@DanWorrall]

I think that's likely to be modulation rather than non linearity. I've heard this described as "linear, time variant".

[@ProductionAdvice]

They used to have a specific reverb called "Nonlin" but I don't know if that actually meant non-linear in the sense we're discussing here...

[@billsybainbridge3362]

Good chat! Many useful points, including the similarity between Limiting and Clipping. DM'd a response to Ian about Compressors that echoes, but further examines particulars of Dan's comments regarding Sidechain's Time Constant (Low Pass Filter) effects on a Compressor's results. Cheers!

[@ProductionAdvice]

Thanks Billsy!

[@klekrfaqer2114]

AFAIK the results of "singing into" a played wind instrument are actually called multiphonics (but I might be of course wrong on this), and one contemporary mage of the technique is Colin Stetson on sax. Not only is he an inhuman instrumentalist, but also an extremely good and unusual songwriter, while being very tech-savvy. He uses a mic setup that practically renders his 100% acoustic performance a one man band show and some of his tunes rely heavily on distorting the recorded signal before it would hit the console. Can't but recommend any of his live performances or the last two of his New History Warfare albums (which to my knowledge have been also recorded live).

[@ananonynom8474]

As there was some confusion regarding the concept of linear systems: For me, the most useful is just thinking of the resulting properties: - additivity: basically meaning you can add together two signals and put them through a filter, or put the same two original signals through a filter individually, and you'll get the same output - homogeneity: scale your input and your output will scale the same way - order of filters does not matter Even with a background in controls and systems engineering, the mathsy definition of what a linear operation is seems ... kind of arbitrary - but it makes sense if you think of it as just "it is defined that way, so that we can work with these properties, which are nice to work with". What operations are linear operations you can then think of as just resulting from that. For example integrals and derivatives are linear operations, which may not be so obvious to some. For example the first explanation in this pod was "input proportional to output" which is a linear function (line graph), but that is far from all the behaviours you can have using linear operations!

[@musahenderson]

I am so happy. The reason I am happy is Ian Shepherd and Dan Worrall are my favourite two audio engineering content creators. In fact my dissertation was based off of optimising loudness which I formulated from Ian‘s AES paper. What has made me so happy is that Dan Worrall is still just a squiggly line on my screen I was terrified I was going to see his face. Oh it was a great chat by the way.

[@ProductionAdvice]

😂 Glad you enjoyed it !

[@juanchis.investigadorsonoro]

Well this is a hard hitting video. Thx for sharing.

[@CharlesFerraro]

I like that Jon Tidey guy. He brought up some interesting points, for example mentioning FM with EQ creating a non-linearity. I think for that it’s actually the volume of the modulator that is nonlinear and not the volume of the input to the EQ. For example the volume of an operator carrier does not effect timbre in an FM synthesizer. I also liked when Jon added the word “detail” while Dan was describing aspects of saturation. I agree btw that it’s nice to have a term for subtle distortion. We have more words for other distortion types too like decimation, commonly downsampling and sample rate reduction. There’s also fuzz and overdrive which are similar to each other. Not to mention tube and diode with the latter being asymmetrical. Also there are common specific amp emulations that everyone seems to use. I guess it’s technically fine to EQ pre or post reverb since both the EQ and reverb are linear. I typically always cut sibilance out of the input to the reverb. That actually SHOULD be the normal practice since the tonal aspect of the voice can have harmonics in sibilant territory. You’ll want those harmonics to be reverberated without being drowned out by reverberated sibilance. So yes, dynamically de-ess before your reverb. Or do what I (and Dave Pensado) do and chop all the sibilance out of the lead vocal track by hand. For anyone interested here’s an example of the trombone technique Ian mentions: kzbin.info/www/bejne/hXa7dHWPqq6hpLM

[@ProductionAdvice]

Ooh, LOVE that example! Amazing...

[@zwsh89]

I would say Dan should do an online course, but then no one would ever go anywhere else to learn and all the other educational platforms would dry up and the audio industry would implode. Plus, how scary would it be if the bulk of audio engineers and producers ACTUALLY KNEW THEIR STUFF and were legitimate competition?!

[@paisleepunk]

this has aged very well

[@grimmmstudios2367]

Re, digital clipping discovered at mastering stage, can this be resolved, or improved by addition of some form of gentle saturation? may not be a subtle fix, but could it help to alleviate the problem?

[@ProductionAdvice]

It might mask it slightly but it's a bit like painting over rusty metal without rubbing it down first. I'd be more inclined to try some of the de-clipping software first, or just live with the digital distortion if a mix alteration isn't possible

[@neville_bengali]

Good stuff

[@bpmlab7938]

Hi guys, fascinating discussion for a beginner like me, many thanks. Please may I ask a question? I'm looking to try recreate the thick sounding detail that you can hear on productions like Miley Cyrus 'Midnight Sky' or the new St Vincent album. I know they're warm sounding by rolling off hi end but I think they're examples of pushing saturation in modern production. Again, I'm a newb and I know my place so might be barking up the wrong tree, but my question is.. Is it generally considered best to make a decision on saturation right at the start when recording each track, rather than later in the buss? I like sampled complex string arrangements with many pieces so this is gonna make a big difference I think. Many thanks

[@CharlesFerraro]

Yah there’s a lot to unpack here. Two unrelated productions probably aren’t going to use the same techniques unless they were produced by the same person. And even then that producer might do things differently for each artist. So your question is a bit moot right out if the gate. “Thickness” is a nondescript adjective that is subjective from person to person. All that said, creating harmonic content via distortion is one way to “thicken” up a sound. You can also add a sub oscillator, or an overtone oscillator. Chorus is another useful tool as well as double tracking etc. Even delay or reverb will cake up a mix to make it sound “thicker”. Now you’re asking if you should use distortion early on in your signal chain? Truth is it doesn’t matter how you get to the sound you like so you should experiment with what works for you. Me personally, since distortion is a non-linear process, I like to put it before compression so I get more shift and shuffle in harmonic content. Won’t do too much for sampled strings though since there aren’t a lot of dynamics there to work with. Nonetheless try using distortion on the oscillator level if your sampler has per-voice distortion rather than dealing with distortion post as that can drastically change the sound in unexpected (and sometimes really cool) ways depending on your chord voicing and number of notes. Good luck and don’t be afraid to ask questions.

[@bpmlab7938]

@@CharlesFerraro cheers man, appreciated. I seem to get more of what I envisage in my head when it goes on early

[@CharlesFerraro]

@@bpmlab7938 perfect! Keep doing that. Also yah Midnight Sky has that retro vibe for sure. If I were to redesign that snare for example I’d prob start by looking for a similar sample in the hundreds of snares at my disposal. Then I’d shape the timbre with compression>distortion>EQ. In that example I would compress before distorting since there’s not a whole lot of harmonic movement in the snare. I’d EQ last to finalize the spectral balance. And of course I’d be jumping between all three processors, tweaking this and that, until I’m satisfied with the result.

[@rjcmusicproduction]

Excellent as always but it has me wondering something, if linear digital eq’s don’t add any nonlinearities why do we have natural phase and linear phase options on some of the more expensive eq’s? Or do phase changes as a result of eq not equal a nonlinearity?

[@ProductionAdvice]

Correct, both "linear phase", "minimum phase" and the other options are all linear processes, assuming there's no saturation or dynamic processing going on. It's confusing, I know !

[@CharlesFerraro]

Yep, as Ian pointed out it’s a difference in vocabulary. “Linear“ phase is being used in a different sense with EQ versus the nonlinearity of distortion. In fact I don’t really like the term nonlinear when it comes to distortion. I think calling it “volume dependent“ is more accurate. One thing to note here is that changing the phase of something (with say an EQ) will also change the nonlinear response of the distortion you’re feeding the signal into. For example try using an allpass filter to change the phase of your sound while running it into a waveshaper.

[@rikyjacho9653]

amazing !!!!!!!!!

[@JimijaymesProductions]

Yay Dan!

[@nexusobserve]

love the "crunch" of an 1176 limiting

[@Oliver-ty7xu]

Ian Shepherd, Dan Worall and Jon Tidey... When gods collide :')

[@DomMcsweeney-DoomPlugs]

There's now a Dan Worral / Loudness War inspired plugin made by a dude called Florian - he shows it off here! kzbin.info/www/bejne/d2jcgpaPhN-dsJY

[@CharlesFerraro]

Something that confuses me… does the order of delay and reverb matter? Both are linear processes. But A delay after reverb is delaying a “smeared” (reverberated) signal vs a delay before is delaying a clean signal that’s then getting reverberated. Either way I guess you’re getting smeared echoes. I’d have to test it for myself.

[@ProductionAdvice]

If both are linear, there shouldn’t be a difference. In practise I would do a test to be sure, though 👍

[@AstronautDown]

Honestly, huge missed opportunity to start this podcast with the patented DW "hi".

[@ProductionAdvice]

😂 Damn, you're right !

[@dantig]

Black Metal is technically a style of music that relies 75% on intermodulation distortion,

[@dodgingrain3695]

As engineers we should all understand this at a basic level.

[@grafzhl]

Convolution is an LTI process, hence reverbs are linear: en.wikipedia.org/wiki/Linear_time-invariant_system?wprov=sfla1 Generally, with this discussion, I think you're a bit all over the place, often jumping between the mathematical foundation of LTI systems and the practicality of using gear, which seems to be grounds for confusion. Anything that can be modeled with an impulse response of a system is a linear process.

[@ProductionAdvice]

Sorry, I disagree with that logic. Convolution can perfectly reproduce any linear reverb, but not one that saturates (for example). Just because most reverbs can be reproduced using convolution doesn't mean they all are. (Eg. an overdriven plate) And yes, I agree the conversation wasn't completely disciplined - that happens when interesting people say interesting things! The show is intended to be fun and entertaining, not a text-book.

[@grafzhl]

@@ProductionAdvice But you're doing it again now, confusing the concept of reverberation, which is a linear process, with using plugins or gear, which might do all sorts of things. Sure, there might be reverb plugins or gear that *also* do other things. Non-linearity is then introduced not by the reverb component of that piece of gear, but by whatever saturation it applies *in addition* to reverberating the signal. The answer to the original question "is reverb linear" is still "yes".

[@ProductionAdvice]

@@grafzhl Fair enough, the answer to the question "is *the concept* of reverb linear" is yes. In practise the answer is "not always", but we concluded fairly clearly on the show that it usually is. As far as your point about convolution goes - sure, but you're just saying the same thing from the opposite direction. Convolution can reproduce linear processes, so only linear processes can be modelled fully using convolution. But convolution is an even more obscure mathematical concept than linearity in my opinion, and our goal was to try and help people who *don't* have a handle on the maths get an idea of what non-linearity means, and how it might affect their work. I guess we didn't fully succeed in your opinion. Fair enough.