Controlling Transients for a Punchy, Loud Mix (without excessive final limiting)

Controlling Transients for a Punchy, Loud Mix (without excessive final limiting)
0

So, I know the title sounds like an oxymoron, but I deliberately titled it that way, because it is kind of counter-intuitive.

In this thread, the discussion on my mix lead to both @ncls and @Lophophora asking about getting punchy, loud mixes without the need for excessive limiting on the final master buss (and the associated negative artefacts associated with heavy limiting and/or compression).

To give a short answer, I said this:

To which Phil and Jean-Marc replied:

So rather than sidetrack that thread, I thought I would start a new post. Here is my answer:

BTW, I’m pretty sure some won’t like my answer…because this method is very “non-purist”… but here goes anyway…

It’s very simple actually. It’s all about peak control at a track level. Especially with transient-rich sources (eg. drums, piano, clean guitars, acoustic guitars) you simply need to find a way to reduce the peak level of the transients with a method that doesn’t soften them (if your goal is to retain transient punch, that is).

This is why old blokes go on about how great analogue is. Analogue equipment does this for you automatically, even when you’re not trying to do it. Tape (for example) is not capable of capturing drum fast transients - it basically destroys them. So when you slam drums into tape, the transients saturate and distort so much, it basically acts as a clipper/limiter. Of course if you slam it too hard into tape, you will get audible distortion, and sometimes that is desirable with drums, because the impact is so fast and noisy, it isn’t perceived as distortion, but rather “fatness”, “warmth”, “punch” or whatever other buzzword applies.

The “problem” with modern digital capture is that it retains complete fidelity of transient information, meaning that in any purely digital capture, the peak to RMS ratio will be very high. This, of course is not really a “problem”, but it becomes such if you expect to treat the audio in the same way as analogue capture and get the same results.

It stands to reason, that if your peak to RMS ratio remains high across 30, 40 or more individual tracks in your project, once all those transients sum together at your mix buss, you’ll have one very “spikey” mix… and if you then slam that into a poor limiter - matter how great and transparent that limiter is - your mix will be distorting long before it even gets close to hitting even modest LUFs levels, like say -15.

So basically, it comes down to this: If you tightly control the transients all the way through your mixing chain - from the track level, to the group level, the buss level, and finally the mix buss, your mix should be pretty much hitting -16 to -12LUFs (sometimes even louder!) even before you do any final limiting.

“…and how do you do that?” Well this is the part of the answer you’re probably not going to like, because it flies in the face of “accepted internet wisdom”…

The “authentic” (and hence probably more “acceptable”) way to do it is to use analogue modelling summing plugins, transformer emulation, tape simulation plugins and the like to simulate what happens in an analogue recording and mixing chain - multiple layers of analogue “sponge” to soak up all those pesky transients.

The problem with using this method strictly, is that you run into the same problem that old analogue guys used to complain about: The transients become “soft” and the mix loses “punch” and immediacy… And of course, the problem with trying to be “authentic” is that nothing is actually “authentic” and it is all digital emulation anyway.

So how do you do it while retaining apparent transient “punch”? In the most “transparent” way we have access to: Clippers and Limiters… yes, that’s right… clippers and limiters at a track, group, buss and master buss level - All the way through the chain.

Before you recoil in horror at this heresy remember, we’re talking about only limiting or clipping 1, 2, maybe 3dB at the most at each stage… and because you’re using it on transient rich sources, it is only touching those very brief individual transients, not the “body” of the sound… Think about it, if you can “buy” 1-2 dB of headroom across multiple tracks and busses in your mix, that translates to a whole heap of peaks that aren’t multiplying on each downbeat when they are summed together across your entire mix at the mixbuss stage.

So that’s it. If you don’t believe it could work, try this: Grab a mix project that is nice and rich in transients, and try just a limiter on each track where the transients need taming. Use it first in the chain, just to limit by maybe 1 or 2 dB at the most. Don’t be terrified of using multiple limiters on your mix - it’s not illegal, and the mix police won’t arrest you. :grimacing: You don’t have to use super-fancy ones that use tonnes of CP-U - just the bog-standard limiter that comes with your DAW will be fine and probably won’t chew up any computing power.

If you do it right, you’ll actually find things much easier to mix, and by the time you get to the final stage of the mix, you’ll be hitting much higher LUFs levels than usual, and your mix will be punchy and immediate.

Once you get confidence in using that method, incorporate clippers into your workflow, so you can use either hard or soft clipping to suit the source material much better.

If you’re at a loss for material to practice on, here is a choice of 3 multitrack projects from my last album:

http://www.multitracks.cambridge-mt.com/FytaKyte_Bitter_Full.zip
http://www.mtkdata.cambridgemusictechnology.co.uk/FytaKyte/LifeGetsInTheWay_Multitracks.zip
http://www.mtkdata.cambridgemusictechnology.co.uk/FytaKyte/LifeGetsInTheWay_Stems.zip

@ncls and @Lophophora I hope that explains it clear enough.

6 Likes

I’m trying to digest all this, you’ve given us lots of content to ponder, but isn’t this all pretty much under the headline “gain staging”? And if so, I would ask how to incorporate attack/release settings into it. That’s the key IMO. It’s not just about knocking down the level, and as you mentioned it can soften the transients if not done right. The attack/release have a lot to do with that, individual to each source. Generally, a long attack and short release might be the formula, but again that’s very specific to the source. Long Attack + Short Release = “punch”, and Short Attack + Long Release = “crunch”. It’s a delicate balance. Of course, you also have Threshold and Ratio to contend with, assuming you’re going for that shaving off the top of the transients only. Per the above, High Threshold + High Ratio can trim, Low Threshold + Low Ratio can control. Name your poison. :wink:

Hey Andrew, thank you so much for taking the time to explain your process here, much appreciated.

It makes a lot of sense indeed. I will definitely practice this. Although I know (some of ) the theory behind hard and soft clipping (odd and even order harmonics and all that), I have never really used it in a purposeful way while mixing or mastering so I really learned something from you today.

There is just one thing that I feel needs to be mentioned here, but it’s getting back to our old debate about “how much louder is loud enough” or rather “how much louder is too loud”. Now that most releases are digital, the challenge is not so much to get the best ratio between sounding good and sounding loud, it is more about finding the right balance between inherent loudness vs retaining a good dynamic range. I know that might sound like the same thing but there actually is a difference.

As you surely know already (I’m mentioning in case someone else doesn’t), how much a loud master will be turned down by streaming platform algorithms isn’t just calculated by the integrated LUFS, it also vastly depends on the dynamic range, specifically on two measurements and the relationship between them, namely PLR (peak to loudness ratio) and PSR (peak to short-term loudness ratio). As far as I know, for Spotify at least there is a threshold that triggers more “turning down”, and it is around a PSR value of 8. I don’t know exactly how this is calculated but I learned it from reading about how the “loudness penalty” plugin by MeterPlugs was designed, and it is easy to verify after having uploaded a number of songs to Spotify, or even by recording the audio from Spotify into your DAW: with a premium account you can turn normalisation off so you can record both the normalised and original version of the song (of course the mp3 compression is slightly messing up the accuracy of the results, but it gives a good idea of how this whole thing works). I also played around with a 3-months TIDAL trial (TIDAL is one of the few that lets you stream original masters in their wav, hi-res formats).

So even if you process makes a lot of sense when pursuing the goal of “retaining punch while making the whole mix louder”, it obviously become counter-productive when pushed to a certain limit, because after a while you just start squashing everything and killing the dynamics, just in a more accurate and controlled way at the individual track stage rather than the mix bus (or mastering) stage.

I understand that it is precisely this level of accuracy you’re after, by taking advantage of the individual track dynamics control for a final result that can sound both louder and punchier, and this has positive side effects that you didn’t mention but seem obvious (like having control over the listener’s perception of individual instruments dynamics and loudness within the whole mix, or between sections of the song). But I think it is worth mentioning that just like everything else, you need to find the right balance in doing so.

I know the question is for Andrew but I’m taking the liberty of chiming in while I’m here: gain staging is different because it is only about setting the peak level on a scale, you are not controlling any of the dynamics with gain staging.

1 Like

Oh, very much opposed to my way of thinking, and I thought that was Andrew’s whole point is to control the dynamics through “limiting” on each track. The question then becomes “how” to control the dynamics. Limiting might help some tracks, compression may help others, but these are essential parts of gain staging to my way of thinking. Keep that horse from getting out of the barn, rather than trying to catch him once he’s escaped. :wink:

2 Likes

I see, just a question of vernacular then.

Gain staging evolved from the analog days (when it was all about getting the best signal to noise ratio), and nowadays converters and DAWs allow for a huge dynamic range (mine, Reaper, can work at a bit depth of 64 bits FP). But this “usable” dynamic range has nothing to do with the dynamics in th music you’re making.

Setting the gain input level at each stage of the signal amplification in the chain (gain staging) doesn’t affect the music dynamics, but makes a better use of the available dynamic range in your chain.

The phrasing can get a bit confusing, I suppose. It looks like you are describing gain and gain staging exclusively by the channel/track gain/fader controls. And yes, that is more of the classic analog definition. But the use of compression and limiting is frequently combined with this “in the box” these days, on individual tracks and instrument busses, and was also used to some degree on some of the analog mixing desks. As I believe Andrew was describing, if you address the “peak control at a track level”, there is less need for it when it comes to the mix/Master buss. It is combining the aspects of gain staging, both the gain/fader levels you are describing, and the use of “dynamic control” (compressors or limiters) to reduce the peaks or dynamic range of individual tracks or instrument busses. A hybrid approach, you might say.

1 Like

Wow, a lot of discussion here… And to think… I tried to make my post as easy to follow and unambiguous as possible! I already spent a tonne of time composing my original post, so I’m not going go through and reply to each individual point.

… But just in case the original point gets completely obsfucated and side-tracked, I’ll repeat the main points:

  • The goal is to avoid heavy limiting (or any limiting, if your final LUFS targets are lower) on the master buss. This will assist in avoiding distortion and in retaining a punchy, transient rich sound. If you mix music that isn’t concerned with these things, then this technique is probably not relevant to you. Feel free to ignore.

  • Anyone mixing modern, drum-centric music aiming at minimimum -14LUFs would probably benefit their music by experimenting with this technique.

  • The premise is simple: You basically use clippers/brick wall imiters on individual tracks/groups/busses upstream of your final master buss to play “whack-a-mole” with the very top 1-3dB of your transient-rich tracks on an individual basis.

  • When you control the transients upstream of your master buss in this way, these transient rich sources are stopped from multiplying as much when they are summed together in your final mix. (Remember that transients from multiple elements tend to hit all at the same time in rhythmic music).

  • The result is less limiting at the final stage, less distortion, and more perceived “punch.”

  • PS. Discussing the ethics, the whys and the wherefores, hearing the sober cautionings of apocalyptic, pandemic-level hearing destruction of the populous related to the so-called Loudness Wars is more tiring to me than listening to Death Magnetic. :wink: Whether you decide to crush your mix to a pulp with a limiter at the end of your Mixbus chain or not is your choice (or possibly the choice of the artist/client). Either way, IMO using this method upstream of that will still yield a better result.

I hope that answers any questions.

1 Like

Thanks for this great explanation Andrew. I found both your original post and the reply quite clear. You’ve confirmed something that I had been wondering about myself. And this leads to a couple of more detailed question on the how side of things:

  1. You advise to use either a limiter or a clipper for transient rich tracks, at the track level. I know they both have similar results, but is there any reason to prefer one over the other?

  2. And then there’s the transient designer type of plugin. If you decrease the transient using such a plugin, I suppose you don’t shave off the peak but bring the RMS level of the peak down? And get a less punchy result? Or is a transient designer just a limiter in disguise?

1 Like

Great question!

It depends on the source… limiters tend to be the most transparent on most sources, but they can soften transients too much.

“Hard” Clippers can be pretty brutal, but if the source transient is very brief and noisy, they can improve the sense of impact without any perceivable sense of distortion.

Soft clipping is more “gentle” and tape-like. It won’t create as harder sense of impact, and the onset of distortion happens earlier, but is more gradual. Soft clipping can work pretty well on sources like acoustic guitars with too much pick transient, if used moderately.

Definitely another option, but in the hierarchy of “transparency”, I’ve found Transient Designers come in 3rd behind limiters and clippers. Limiters and clippers are “brick wall” devices, whereas transient designers are not, so there is less ability to put an “end stop” on the transients, so the goal of “control” is harder… Although, Waves’ Snap Attack does have a built in limiter that can be activated.

Again- great questions- I hope that explains it :grinning:

PS. I know most are familiar with the Waves Scheps Omni Channel. One of the genius features is SOC is the final limiter right there in each plugin.

My current favourites for track processing
Limiter - Waves L1
Clipper - IK T-Racks
These are both probably almost of drinking age in the US, but they just work for me really well!

1 Like

Yes, those were great questions from Evert, and this thread has been very interesting. I hope I didn’t confuse anything with my responses, I was just trying to clarify, as well as explore my thinking and explanation of it.

I think you meant Scheps Omni Channel, just to clarify. :wink:

Ha! We sometimes use the phrase “long in the tooth”, but I like the drinking analogy. :beerbanger: Yes, those are tried and true good plugins, though I don’t seem to use the IK Clipper nearly as much as I probably could, so among a few other points that is something this thread reminded me to check back in with and explore further.

1 Like

Blame auto-correct :sweat_smile:

A friend of mine does this .I usually just use a compressor and up the attack .The ssl Duende stand alone comp is one of my favs for smoothing transients. There is something special about that compressor i find

1 Like

the loudness is measured in approximately half second intervals and the dynamic range changes over that half second. If you have a transient that goes over the trigger threshhold within that interval, then that sample is considered “loud” by streaming platforms. The more ‘such samples’ you have, the louder your song will be considered and turned down by streaming platforms. A way to avoid being squashed too much by streaming platforms is to ‘pump’ the loudness in 1-2 second intervals to avoid the half second detection over prolonged periods, hit hard and then soft (going over the threshhold and then under it). This can be done to high mids and drums.

punchiness is relative to the listener. A listener will consider a song punchy if they have some non punchy parts in the same song to compare and contrast.

Well isn’t there a lot of other methods to measure loudness? Average, short term, momentary, integrated, loudness range, true peak… What you are describing is the momentary loudness (measured over 400 milliseconds) right? This is one of the loudness measurement stages as recommended by the International Telecommunication Union (ITU-R BS.1770-4) but other recommendations exist (like EBU in Europe, ATSC in the US, etc) and they all add their own tweaks and specificities to this standard. As far as I know, Spotify and other streaming platforms don’t care about these norms. Some use LUFS (as described by ITU but not with the -24 LUKS target), some use ReplayGain (Spotify actually just switched from the latter to the former a month ago), some turn music down but not up (Youtube, Amazon…), and some don’t normalize at all (Bandcamp, SoundCloud…). As of December 2020 these are the main platforms algorithms targets (source):

### Platform ### Peak ### Loudness ### Dynamic Range
Spotify -1.0 dBTP -13 to -15 LUFS >9DR
Apple Music -1.0 dBTP -16 LUFS(±1.0 LU) >9DR
Apple Podcasts -1.0 dBTP -16 LUFS(±1.0 LU) >9DR
Amazon Music -2.0dBTP -9 to -13 LUFS >9DR
Spotify Loud -2.0 dBTP -11 LUFS >9DR
Youtube -1.0 dBTP -13 to -15 LUFS >9DR
Deezer -1.0 dBTP -14 to -16 LUFS >9DR
CD -0.1 dBTP > -9 LUFS >9DR
Club Play -0.1 dBTP -6 to -9 LUFS >8DR
Soundcloud -1.0 dBTP -8 to -13 LUFS >9DR

So all this leads to my question for you: why focus on the momentary loudness only? I am genuinely interested in the loudness topic so if you have reliable source material about this I’d love to see it.

1 Like

momentary lufs is a big factor in determining loudness by some streaming platforms. It is a factor we can control quite easily and it is a factor that also gets overlooked quite easily without taking a huge dip in overall loudness. Which is why my focus is more on momentary loudness. Maintaining good healthy momentary loudness levels is still very critical to what is perceived as punch. A listener wont really perceive deviations from the average loudness over a long interval ( for example a casual human listener isnt going to worry about parts louder than the parts 30 seconds ago in the song, unless ofc its obscenely loud suddenly) so DR becomes less of an issue. Our goal is to impress the the human not the machine … atleast not yet lol. I dont want to derail this thread into loudness war direction :slight_smile: but it does have some common elements to the topic.

1 Like

Well this is a topic about loudness and in a world where the majority of streamed music is normalized, I feel that this is an important one to understand for anyone who releases music digitally. We don’t have to make it about loudness wars though, personally I am not at war with anyone (but myself maybe!)

In 2017 only 17% of the music that was streamed was not normalized. I don’t have the recent figures but it is obviously less by now, the trend clearly is to have everything normalized. So IMO the loudness war is not even a relevant topic now. But when you’re involved or interested in mastering, you need to have a good understanding of the normalization algorithms in order to ensure the best possible playback for the end user.

I always read that PLR and PSR are the key values that streaming platform algorithms rely on to normalize, this is why I’d like to know where you got your info about momentary loudness being a big factor. I’m not trying to contradict, just seeking to learn and understand. It is quite a big deal because in the domain of short-term loudness we are looking at 3 second periods of time so we’re no longer in the transient realm.

Here are some sources quoting PLR/PSR being key factors (and not mentioning momentary loudness):

https://www.soundonsound.com/techniques/mastering-streaming-services


https://www.producenewmedia.com/podcast-dynamics-loudness-range-vs-psr-plr/

momentary loudness is factored into the lufs calculation overall. I wrote a blip about streaming here on a fine day.


I mostly mix for streaming platform and while some info in that thread maybe old, there are still some streaming platforms that use lufs as a main factor.

1 Like

Yes, and I also found out that the “Heavy” Saturation mode is a clipper! It doesn’t seem to spell this out in the manual, but the Schepinator said it himself in his “In-Depth” video for Waves on the plugin. I don’t know if it’s hard or soft clipping, though maybe that can be controlled with the “Saturation Drive” (?) control … the big round knob that goes from 0-100. Something to play with. So while this plugin is the Swiss Army Knife of audio, it looks like it also covers your criteria for either limiting or clipping, or both.

1 Like

That’s cool to know… that Andrew S is a smart fella!