Mix Challenge

So, I've been watching some videos on gain staging using a VU meter. I guess you know the drill, you calibrate the VU meter to -18 so when the needle hits 0 you're on -18 DB RMS. Then you clip gain each element until it peaks around 0 on the VU meter. That'd mean that its RMS level should be around -18.

While that's okay for instruments like bass and overdriven guitars, whenever I do this with drums, the peaks usually go over 0 DBFS. That's the first problem. Does it just mean that the recorded drums have a huge amount of dynamic range and should be reduced, or is there something else going on?
When I unsolo everything, the levels on the mix bus are hot and I end up lowering everything anyway when doing some rough balancing. I get this, but I don't see the point of bringing everything up to the same RMS levels if you'll have to rebalance anyway.

So I'm missing something here. I either don't do it well, or I don't get what's the whole purpose of it. I know most plugins modelling analog gear are calibrated to work on -18 DB RMS, but still.


So, what's your approach on this, do you even do it, is there any benefit?

Up until now, I was doing the gain staging starting with the drums, making them peak around -8, then build the rest of the mix around it. never cared too much about RMS until the final stages of the mix, on the whole thing, not element by element.

Ooooh... that sure is a highly controversial topic (actually not, but people turn it into that). In fact, I've been talking about this over on KVR Audio for a couple of years at this point. People think that this topic is super complicated, but once you wrap your head around it, it's actually quite simple (hence part of the rule set for the Mix Challenge with the reference level).


I might do a dedicated post about this at some point down the road (you were definitely faster on this). But for the time being...

The benefits of proper gain staging
Simplified, it's the following

• it makes it easier to combine both hardware and software
• "analog emulation" type plugins and hosts (think Harrison Mixbus) are not overloaded
• really old (bitrate locked) plugins are not overloaded (think 20bit integer math, or 32bit integer math - they still exist!)
• better fader resolution
• the infamous "you don't have to touch the master fader"

You can use "clip gain" to setup your signals, though if your host has an input gain knob (Cubase does, Reaper does with a mod, Logic needs a dedicated plugin as the first insert, etc) it's way less work.




"So what am I missing?"

Let's start with the basics. You have an average signal strength, and a maximum signal strength.

Average signal strength is usually either +- xyz VU (relative) or - xyz dBFS RMS avg/realtime (absolute).
Maximum signal strength can be +- xyz PPM (relative) or - xyz dBFS max peak (absolute) or even - xyz dBTP max (absolute, ISP safe)

There are other metering types, but for the sake of this topic (mixing, not mastering or measuring environmental noises), let's stick to VU, RMS and dBFS max peak. We will also completely ignore how to calibrate a VU with a sine signal (this is an advanced topic). Most VU's (there are exceptions, those with single-needle views that do some weird summing of the stereo signal) are accurate enough so if you setup -18dBFS = 0VU, then your reference work level is now -18dBFS (EBU convention R68, the SMPTE with their RP 0155 do recommend -20dBFS).

Wikipedia is your friend:
https://en.wikipedia.org/wiki/Peak_programme_meter
https://en.wikipedia.org/wiki/VU_meter




So how to "ideally" perform with this

First, you have to determine what type your source is. Is it bass intensive, or is it transient heavy?

• If it's bass intensive, you setup your material according to +- xyz VU (example 0VU = -18dBFS) or - xyz dBFS RMS avg (example: -18dBFS).
• If it's transient intensive, you set it up to either +- xyz dB PPM (in case of the IEC 60268-10 Type I/DIN meter, that would be -9 PPM) or - xyz dBFS (that can also be -9dBFS, though personally I go up to -6dBFS per channel).

So if you have a snare that is fairly thin, or high hats, don't let the signal peak up to your desired maximum level (again, example: -9dBFS). In turn, the VU would barely respond as it is not only slow in response, but the VU has a stronger response on more lowend frequencies. If you happen to have vocals, a bass or a deep kick drum however, then you set up your channel to ideally not exceed 0VU.

You have to find a balance of these two values, then you're basically "in the green" in terms of setting up your project. In turn you should now have a way better fader resolution (which makes fine tuning way easier). And if you use a reference for the VU on the summing bus as well, you can perform the same way (0VU for average signal strength and personally I set my dBFS max ceiling to -3dBFS, in reality I never reach -6dBFS though) and barely even have to touch the master fader.




Why not use a RMS meter for individual channels?

Digital Meter, PPM's and VU's are standardized with their values. Digital Meters are usually sample accurate, PPMs can either be 5ms or 10ms in terms of "rise" (inertia/response time), and VU's are usually around 300ms rise/fall (inertia/response time). So if the tool in question is calibrated / programmed right, and has no weird summing mechanics (do not trust any VU that uses one needle for a stereo signal!), you can easily recreate your measurements with whatever is at your disposal.

"Basic" RMS realtime meters on the other hand can be all over the place. One uses 300ms as time window, another one 400ms, the next one 600ms, and yet another one uses 250ms. Then one uses bar graph smoothing, while the other one doesn't - it can be a huge mess sometimes.

The first considered and patented "RMS Meter" was the Dorrough 40A. And the patent stated, it's working twice as slow as a regular VU, while the peak measurement is as fast as possible. Making it somewhat a digital meter and a 600ms VU bargraph. The Dorrough 40A also has a relative scale, and the reference point is around the -20dB mark (-ish! You can't compare hardware behavior with software math easily). The K-System meter is actually based upon these specs, only with a shifted reference level. You can use it for "individual channel use", but you have to keep certain offsets in mind (due to the slower response, average signal strength readout it is lower than a VU at 300ms!).




"So what is your suggestion, Mister Fox?"

Set clear lines between mixing and mastering.

For recording and mixing, use a combination of a Digital Meter and a VU
For Mastering, either stick to the old K-System (v1) specs (Dorrough), or the current ITU-R BS.1770-x specs (LUFS meter)




I hope this is a good starting point.

Fox man, when you write a comment it's like if you were writing a book! I'm adding this to my to-read collection, since I don't usually pay much attention to gain staging and sometimes I get in a bit of trouble.

Thanks! So many things here to expand upon... This is the kind of stuff I mean when I say "put some exclusive content for patrons on the vault".

So, what I take from this without getting into all the technical, would be to use Klanghelm VU meter since it has RMS and PPM (calibrated to -18 and -9 automátically, btw), and use PPM for percussive instruments and RMS for practially everything else.

Correct.

If you happen to use VUMT2 (either standard or deluxe), take a look at the meter settings. The VU is 300ms (rise/fall), the PPM is DIN type with 5ms (rise) and the RMS meter uses 600ms (Dorrough patent) - though all modes offer a "dBFS max" readout in numeric form. I'd only use the VU and PPM mode - to keep things "simple".


Of you want to work "oldschool", use a VU (reference at -18dBFS) and a PPM (reference being -9dBFS, +- 5% overshot)
The PAR (peak-to-average range) will be 9dB in this case

If you want to be more "modern", use a VU (reference at -18dBFS) and Digital Meter (sample accurate, up to -6dFS is suitable IMO)
The PAR (peak-to-average range) will be 12dB in this case



On the summing bus, you can raise your personal dBFS peak maximum value to -3dBFS, while still running your average signal strength between 0VU to +3VU (or -18dBFS to -15dBFS avg), which would be a PAR (peak-to-average range) of 15dB. Can't get any more dynamic than that.

And that is using reference levels "in a nutshell".

This was an interesting read!

Thank you!