In which gain staging and VU meters make me cry

[tomdot]

Hello,

I am testing out using a VU meter while mixing to help with my level control to the mix bus after watching some interesting videos on the subject of gain staging. I know about the general concepts involved (peak/RMS/etc/that sort of thing), and the SOS article that oft gets linked to is something I understand, but I'm confused about something.

If the VU is calibrated to -18 dBFS, and this correlates to 0vu, why is my VU meter only tickling down in the -10–20 area when my master bus has levels no stronger than -12dBFS. In fact, the audio itself is peaking between -12 and around -18 dBFS.

https://imgur.com/a/mDao3lw

"Ah," you say, "Peaking – the music is probably transient-heavy. Normal behaviour, can't replicate". I understand this too, but I've threw together three Apple Loops for my test, a drum kit, organ, and bass. Okay, drums yes transienty, but the others not so much.

I have a sneaky feeling I'm being a thickard about something. I watch these videos and read the stuff, and you can see them all adjusting gain to 0vu for – in one case – a kick drum that had peaks of -12dBFS and an average of 0vu with no other processing other than a gain adjustment. It made me suspicious of there being some compression on there not declared, because I can't replicate that. I then wonder whether it's my problem and that I can't gain stage or use meters properly.

As an additional question, I got thinking about pre peak metering situations, where all you had were VU meters. How would you go about recording these transient instruments – for example, would you say, "get the kick showing -5vu" and know that the transients will sometimes get splatted by some analog distortion? Like, how would you set levels with VU only metering?

[Hugh Robjohns]

The fundamental issue here is different meter technologies and ballistics.

You can only realistically compare meter readings and calibrations when running steady tones. Comparing them with music signals is as pointless as it is misleading!

Its hard enough finding two brands of VU meter that give similar readings. Comparing VU and digital sample peak meters is a complete mindf**k!

But for what it's worth your comparison numbers for a sample-peak meter and a VU sound quite plausible to me.

My advice: pick one type of meter, align it as required, then gain experience in learning how to interpret it when mixing and relating what you see with what you hear... and stick with it.

As to 'how would you set levels with VU only metering?'... the answer is experience combined with analogue headroom! There's a reason audio equipment designers typically engineered 20dB or more of headroom into devices using VU meters!

The VU meter gives a crude approximation of perceived loudness but sustained sounds will always read higher than transient sounds of a similar volume so experience / practice is needed to recognise what its trying to show you.

Other types of meter -- q-PPMs, sample-peak, dorroughs, modulometer, R128, etc -- require different interpretations.

[ef37a]

Someone, many years ago (prob' a BBC tech!) gave the VU meter the sobriquet "Virtually Useless".

Many..MANY years ago I had a little DIY project published in the (then) hallowed pages of Hi Fi News, it was a simple circuit that gave a 'peak' indication of the mono sum of signals in a (very decent) Yamaha Dolby cassette deck. I calibrated it to jeee ust light at +3 Dolby level but was very nervous about the device because it would blink for a snare shot but the VU needle barely moved. Much checking of the calibration later showed this was in fact quite correct.

Dave.

[Mike Stranks]

Someone, many years ago (prob' a BBC tech!) gave the VU meter the sobriquet "Virtually Useless".

Many..MANY years ago I had a little DIY project published in the (then) hallowed pages of Hi Fi News, it was a simple circuit that gave a 'peak' indication of the mono sum of signals in a (very decent) Yamaha Dolby cassette deck. I calibrated it to jeee ust light at +3 Dolby level but was very nervous about the device because it would blink for a snare shot but the VU needle barely moved. Much checking of the calibration later showed this was in fact quite correct.

Dave.

Back in the day... when VU meters ruled the world (except at the BBC) one quickly learnt that VU meters couldn't be trusted when there were lots of transients about...

'Twas a revelation to me when I joined BBC local radio and was introduced to the PPM...

[Hugh Robjohns]

Someone, many years ago (prob' a BBC tech!) gave the VU meter the sobriquet "Virtually Useless".

They did -- I was taught that too... but it was an ignorant and unfair claim.

I spent several years trying to correct that injustice while I worked as a lecturer in BBC Training. The VU is virtually useless when incorrect calibrated... which it routinely was at the BBC.

However, a properly engineered and aligned VU meter is actually perfectly usable for its intended purpose, and usefully informative as far as it goes!

Unfortunately, a properly engineered VU meter is relatively expensive (although less so than PPM), and most equipment meters with VU written on the scale really aren't!

But there is no 'perfect' meter. They all have different strengths and weaknesses that make them useful for some functions and applications, but not others.

[tomdot]

This is all very informative, and also affirming. I have gone from recording at -0.1dbFS on 16-bit because of half a rizla of dynamic range, and weaned myself down to -12 per track peaks, and -6 peaks for busses. I suspect I should go softer than this and level at -18 (-12 on the master bus) judging by the information I read.

A lot of the gain staging advice I read made no sense, and left me wondering, “What’s wrong with setting channel peaks to -12db and carrying on?” It seems like the answer is “Not much”

Seeing as we’re on the subject, I’ve heard it said that peak meters are “useless” if you leave enough headroom, and VU and PPM are antiquated. Two questions on this: 1) What is the alternative that every DAW should include instead of peak meters, and 2) Why are digital VU meters a product if they don’t serve a lot of purpose? Is it that ‘vintage’ thing?

[Hugh Robjohns]

...and weaned myself down to -12 per track peaks, and -6 peaks for busses. I suspect I should go softer than this and level at -18 (-12 on the master bus) judging by the information I read.

Mix bus true peaks of -6dBFS wouldn't upset me, but you could run a little lower if you wanted.

A lot of the gain staging advice I read made no sense, and left me wondering, “What’s wrong with setting channel peaks to -12db and carrying on?” It seems like the answer is “Not much”

I'd concur!

Seeing as we’re on the subject, I’ve heard it said that peak meters are “useless” if you leave enough headroom, and VU and PPM are antiquated.

Analogue meters are antiquated in a digital world, and if you work with sufficient headroom knowing the (sample) peak level is irrelevant.

Two questions on this: 1) What is the alternative that every DAW should include instead of peak meters...

Some variation based on BS.1770 loudness metering with true peak measurement.

2) Why are digital VU meters a product if they don’t serve a lot of purpose? Is it that ‘vintage’ thing?

Yes, let's go with a vintage thing! Few people understand the subtleties of metering, and most just want stuff that looks like what the pros used in the good old days, with no idea of why or how it was used, or its modern day limitations.

[shufflebeat]

2) Why are digital VU meters a product if they don’t serve a lot of purpose? Is it that ‘vintage’ thing?

[Steps gingerly outside of comfort zone]

The way I understand it the place for VU meters in a digital world is to approximate the listening experience rather than the numbers behind the details, which are less apparent when listening to music rather than data points. A digi-meter will "over" when a single sample jumps the bar, which might prompt the incurious to simply drop the level where someone with both meters might recognise a discrepancy between the red lights on the screen and the unremarkable pointers on the VU and recognise the need to de-peak instead.

Phew!
[Retreats into guitar section]

[Mixedup]

I’ve heard it said that peak meters are “useless” if you leave enough headroom...

Well, they are One problem with the peak meters being next to the faders is that people who haven't yet properly got their head around metering (which, arguably, means a majority of DAW users today...) seem to assume it indicates how *loud* a signal is but it does nothing of the sort. What if the peak is too high? All it tells you is how high the peak is (and doesn't even do that accurately haha), but nothing about how loud the sound will be perceived — ie it tells you nothing about how the sound on one channel compares to sounds on other channels. So why take up all that screen real estate with these pretty bar-graph meters?

I keep an eye on the peak meters on my converters when recording — and if you don't have those meters onboard, there's a certain sense in having peak meters on input channels in your DAW (or at the input of channels on to which are recording, depending on the DAW).

But during playback — eg when mixing — they tell me little of any value, and certainly nothing more useful than a pair of virtual LEDs could tell me. I'd prefer the bar graph meters on each channel to indicate loudness (so a VU or momentary loudness meter of some sort). That way I can compare tracks at a glance. If I hear something too loud in a big, busy mix, I can look at the screen and instantly see which track it is. Add a peak-hold on top of it and it's better still.

Personally, I reckon a more informative lightshow — which would give back huge amounts of the screen for other things — would be to have a pair of 'clip' LEDs on every DAW channel, one pair at the input, another at every insert slot, and another pre and post fader, which tells you if the signal is exceeding a user-defined threshold at any stage. One LED for an average level of some sort and one for peak (or true peak if you must), and changing colours (eg traffic light). It would provide a simple warning at every stage in the signal path that you might want to pay closer attention to something, and it might also encourage some newbies to pay attention to gain staging (which, as I've said elsewhere is still a bloody sensible way to work, even if the DAWs themselves can cope with huge dynamic range... not least because you need to understand that to be able to use, say, preamps and hardware EQs/compressors, or guitar pedal chains...).

On the master buss/output channels and for mastering, I'd still want a standard loudness meter.

[Hugh Robjohns]

That all makes sense to me!

At the end of the day, meters serve only two purposes.

1). They allow the user to optimise the signal levels passed between equipment to optimised the SNR and headroom margins. This can only be done with steady sine tones.

2). They provide visual confirmation that programme levels are within acceptable parameters...

But what your ears are telling you is all that really matters! It's interesting that
I've had the great pleasure of working with some blind radio studio managers in the BBC who could reliably mix extremely high quality programmes without ever seeing a meter!

Set the system up with headroom, set your monitoring level appropriately (and leave it alone thereafter), and you need never view a meter again!

[tomdot]

Part of the problem with peak metering that doesn’t get talked about is that it’s quite easy to normalise the bar after looking at it.

What I mean is that, yes you can see bouncy bits turning orange and red at the top, but when the signal drops to the lower end of the range of the input, your eyes adjust to almost dismiss it. I’m not sure I’m articulating this very well

I wish that you could turn off the meters in Logic, so you don’t see them, but you can’t. In modern Formula One cars, drivers realised that watching the rev limiter to change gear didn’t work out well, so instead, you now have little blips in an ear piece to indicate the top of the rev limit for a gear.

That would be weird to implement in a DAW of course, but it would be really effective to highlight peaks without looking at a bar jumping about. The clip lights mentioned would be good as well, and I’d be excited to see that.

[Hugh Robjohns]

The major problem with clip lights is that when they illuminate you've already screwed things up! (The other problem is that they often lie, lighting up when there's no problem, or not lighting up when there is!)

What is needed is advance warning of an impending problem, not something that effectively says 'I told you so'!

Peak meters with sensible colour coding can be very effective. My preference is green up to the reference level (-20 or -18dBFS), yellow up to the permitted peak level (-9dBFS), and red above that.

Normal material will be green all the time with some orange depending on what it is. Any flicks into the red provide the advanced warning that you need to optimise levels, and the more red the more desperate the situation.

This replicates the useful feedback that analogue meters traditionally provided.

But this kind of full-blown metering isn't necessary on every DAW channel when mixing, and what I'd prefer to see on replay DAW channels is just a single LED that stays green when levels are acceptable, fading through to red when they're not. It wouldn't be difficult to create an algorithm that can analyse the signal taking into account its characteristics and indicate when it's within acceptable parameters and when it's not.

[Aled Hughes]

I've changed my channel meters in Reaper to show LUFS-M, but unfortunately they default back to peak meters if a track is record enabled.

I find peak meters next to useless - a numerical display showing the max peak level above the meter is plenty (or a light perhaps)

The Reaper master meter is very configurable, and I've got that at LUFS-M also, with a -20dBFS sine wave reading '0' at the moment. The meters turns yellow above 0 until +10, at which point it turns red. There is a numerical peak display above it, and a numerical LUFS-I display underneath. I aim for my mixes to hover around the 0 mark, or just below. I find this works well.

I'd really like the channel faders to be this configurable, but in the meantime I've set up the included JS loudness metering plugin with it's embeddable GUI as a sort of traffic light system when setting input gains and mixing - I haven't quite settled on it yet, but at the moment it turns yellow at -30 LUFS-M, and red at -24 LUFS-M. When setting up initial gains when tracking or getting a rough balance in a mix, I try to aim for that 6dB yellow window. This seems to give me a mix at around the master fader's '0' as stated above. Really, it's a continuation on the 'pink noise mixing' approach to set initial levels.

I've recently set up this system, but it seems handy to me, although I might bring the target levels down somewhat for tracking as the peak levels were sometimes higher than I'd like. It's a work in progress...!

[Hugh Robjohns]

I've changed my channel meters in Reaper to show LUFS-M, but unfortunately they default back to peak meters if a track is record enabled.

I don't think that's unreasonable.

When tracking the peak level is the single most critical parameter, since converter clipping is unacceptable. However, when mixing you don't need to know the peak level -- it's already a safe value -- while the instantaneous loudness of the track is a critical factor in the mix... and the Momentary Loudness value indicates that quite well (it's actually very similar to a proper VU display!)

[Mixedup]

I wish that you could turn off the meters in Logic, so you don’t see them

Yes! This is a feature all DAWs should offer!

[tomdot]

If we’re going further, I think there should be a sample of Geoff Emerick tutting whenever your signals go above a specified threshold.

[Aled Hughes]

I've changed my channel meters in Reaper to show LUFS-M, but unfortunately they default back to peak meters if a track is record enabled.

I don't think that's unreasonable.

When tracking the peak level is the single most critical parameter, since converter clipping is unacceptable. However, when mixing you don't need to know the peak level -- it's already a safe value -- while the instantaneous loudness of the track is a critical factor in the mix... and the Momentary Loudness value indicates that quite well (it's actually very similar to a proper VU display!)

Certainly not unreasonable, but it would be nice to have the option now that they’ve incorporated it when tracks are not record enabled. I just don’t need that much screen space telling me how far I am from clipping!

[Richard Schiller]

Apologies for being a bit late to this discussion. Maybe some of this might be useful to someone reading this thread:

Metering operational sound levels is a complex topic and no one should feel bad for finding it so. Meters are needed for two very different reasons, to balance sound and to prevent over modulation. Originally, it was the over modulation of transmitters and then magnetic tape but now it is the finite capacity of digital coding that needs to be addressed.

The BBC peak programme meter is not, as is widely thought, a peak meter but instead a compromise to indicate programme level yet give some idea regarding over modulation. It's a clever piece of UI design. The way to ensure quality across a huge organisation with many different users was to standardise. In level terms, this meant lining-up every piece of equipment for zero gain and with similar headroom. In the analogue era this was solved by distributing reference level tone and aligning to that. Line-up level was defined as BBC PPM 4. This was vertical for the meter needle - both easy to remember and easy to read.

By the way, BBC PPMs used needles that swept an arc for a reason. The human eye is most critical when viewing angular displacement - we have to be good at this as we use it as part of our balance system.

When reading programme levels, the meter was designed such that typical speech would average around PPM 4. If the needle seemed to be as much above 4 as below, then your programme would sound similarly loud to another balanced in the same way. Controlling the music of the day to also average around PPM 4 worked too. Later music would be more compressed and anyway, with all sounds, the operator had to use their judgement to create the sound field required for the programme. In summary, '4' was both the target for line-up tone and the target for balanced content. More experienced operators knowing that this nominal level would be varied with specific content types.

The BBC PPM was not perfect and peaks would get through but only short ones and compressors and peak limiters mounted at the end of the chain would ensure they were taken care of. Remember, clipping just a short peak does not matter to the same degree as clipping the body of the sound would. When recording, the magnetic tape recorders would naturally provide compression of the occasional short peak.

BBC PPM meters were very costly. The later VU was introduced as a more affordable meter but was also designed for a different purpose. As the name implies, it is there to match volume (how loud the programme is) not indicate peak levels at all. As a volume indicator, the VU meter is sometimes superior to the BBC PPM. As a peak indicator, it is always worse. The ever-present comparison between PPM and VU is somewhat unfair - like comparing a measuring jug and a kitchen scale.

The VU meter's biggest drawback for programme level indication is that it is not linearly logarithmic. The BBC PPM had 20 and then 24dB of dynamic range of true logarithmic indication - plenty enough for the purpose for which it was made. The difference that being linearly logarithmic makes, in practice, is that it allows the operator to mix to different levels. You can mix some louder voices and then say a quieter conspiratorial whisper, hitting two different spots on the meter.

On a purely aesthetic level, this is what makes choosing between VU and PPM meters rather difficult. The ballistics of the VU are arguably more pleasing but only when the audio level is exactly at the sweet spot. For general use, the PPM therefore looks better.

In the digital domain, we don't have the luxury of tape compression or (at least usually) analogue delay line compressors and limiters and so the requirements for digital metering are different.

In digital, true peak meters have become common though not because they address its unique challenges but rather because they are easier to code. These have an instant rise time and so indicate absolutely any overload peaks. This instant response would seem like a good thing and in some ways it is. However, clipping very short peaks is less important than clipping the body of the sound and such meters do not discriminate. Remember, if you have silence with just a single sample peak and it is clipped, then programatically there is no distortion.

Some of these peak-reading meters have been given a slow fall-back like the BBC PPM. On the one hand, a peak reading has to be held or else you won't see it but on the other these ballistics suggest a programme capability which such meters don't have. A meter with an instant rise time and slow fall back is not a programme meter and so is not a PPM - though invariably that's how they're labelled.

As has already been suggested, when recording you will be more concerned by over modulation. When mixing, it is the programme level (getting the volume right) that matters most. In the analogue domain, the BBC PPM did a fair job of both. The VU was definitely a mix meter rather than a recording meter but was of course used for both anyway. Neither though are designed for the wide dynamic range we now have thanks to transistor pre-amplifiers and cheap high-performance analogue to digital conversion.

Many years ago I patented a programme meter for digital operations. It had two indicators, one for programme level and another for over modulation. I still think that this is the only way to solve the dichotomy. The level indicator was a bit like a linearly logarithmic VU meter but improved to be linked to audible volume (using a weighting curve to match the ears response). It was the foundation of later Loudness Unit measurement. The over modulation indicator was something like a peak meter. So as to not be distracting, it would only display when sample levels got near maximum and would gracefully fade away when not needed.

[Hugh Robjohns]

Apologies for being a bit late to this discussion. Maybe some of this might be useful to someone reading this thread:

A useful contribution indeed -- and thanks for that -- but there are a couple of points I'd like to pick up on to avoid potential confusion in other readers.

Meters are needed for two very different reasons, to balance sound and to prevent over modulation.

I'd say three key reasons. The first, and arguably the most important, is for equipment alignment. Meters allow us to optimise the reference signal level between the noise floor and clipping level, as it passes through different pieces of equipment in the signal chain, as well as to ensure unity gain throughout the signal path.

The second is to warn of over-modulation -- although that is much less critical now than back in the day as we generally work with more headroom now -- and the third is as an aid to mixing and maintaining consistent perceived loudness levels, as you say.

The BBC peak programme meter...

The analogue PPM was/is not exclusive to the BBC. Pretty much all European and Scandinavian broadcasters developed very similar PPM designs around the same time using the same core technologies.

The BBC's PPM has a unique simplistic scale (officially known as the type IIa PPM), but the underlying design and response is is also used in the otherwise identical European type IIb PPM, and is functionally very similar to Type I PPMs often known as the DIN meter, (and the Nordic meter is a variation of ot, too).

When reading programme levels, the meter was designed such that typical speech would average around PPM 4.

You won't find many networks that keep speech averaging around PPM4 today! Most tend to keep speech in the top half of the meter, around PPM5 or (if more compressed) PPM6.

The BBC PPM was not perfect and peaks would get through but only short ones and compressors and peak limiters mounted at the end of the chain would ensure they were taken care of.

This was an intentional element of the original design and was introduced specifically to avoid operators under-modulating the transmission networks. Occasional brief transient overloads were deemed acceptable (and generally unnoticeable), and a higher overall modulation level was more desirable. The Type 1 and Type2 PPMs really only differ slightly in the duration of transient peaks they ignore.

Remember, clipping just a short peak does not matter to the same degree as clipping the body of the sound would.

This is true for the analogue world, and that's why PPMs evolved the way they did, but brief peak clipping sounds considerably more obvious and intrusive in the digital environment than in the analogue one (mainly because it creates aliasing which results in unnatural anharmonic distortion).

BBC PPM meters were very costly. The later VU...

Not much later! The type 1 (DIN) PPM was defined in 1936 and the type II (BBC) PPM in 1938. The SVI (better known as the VU) was defined in 1939, but the work to develop it was a couple of years earlier...

That said, the PPM's conceptual origins can be traced back to the late 1920s in the telephone industry, and the BBC's development of those concepts into the 'Smith' meter in 1932.

The ever-present comparison between PPM and VU is somewhat unfair - like comparing a measuring jug and a kitchen scale.

We can certainly agree on that!

The ballistics of the VU are arguably more pleasing but only when the audio level is exactly at the sweet spot.

Yes, the criticality of correct VU calibration is overlooked by many.

In digital, true peak meters have become common though not because they address its unique challenges but rather because they are easier to code.

The term 'true peak meter' has acquired a significant and specific meaning that is different to the one I think you are describing here which is more accurately referred to as a 'sample peak meter'.

Most digital recorders and DAWs simply examine short blocks of digital samples to ascertain the loudest sample value, ad then display that on the meter -- hence 'sample peak meter'. It's very simple and fast to achieve, through both hardware and software implementations.

However, while an entirely acceptable solution for most purposes, it is not accurate when it comes to evaluating the actual peak audio levels as reconstructed from the digital data.

For example, most sample peak meters will indicate an overload ('over') if a single sample hits the top quantisation level (0dBFS) -- and yet this can be a perfectly legitimate situation which does not represent an overload at all. To overcome that issue, some early digital meters only warned of overload when a number of consecutive samples were at 0dBFS (Sony settled on four consecutive samples)... but again, this could be perfectly legitimate non-overload condition for a low frequency signal.

So sample peaks meters often warn of overloads that don't actually exist, and at least that's erring on the side of safety... However, they can also fail to warn of overloads that really do exist -- known as 'inter-sample peaks' and that's a much more serious problem when working intentionally close to the clipping level (as CD mastering stupidly did/does). These inter-sample peaks occur frequently in peak-normalised music as the reconstruct analogue waveform often peaks significantly higher (3-6dB) than the peak level of individual samples, resulting in an overload of the D-A circuitry.

For this reason, the BS1770 loudness metering system includes an oversampling peak metering element specifically to detect inter-sample peaks, and this is defined as the True Peak Meter. So it is important now to be clear in the use of the terms sample peak and true peak meters in the appropriate contexts.

Remember, if you have silence with just a single sample peak and it is clipped, then programatically there is no distortion.

In the digital realm clipping distortion results in aliasing which is much more obvious and unpleasant to the ear than analogue transient distortion, hence the far more stringent metering requirements to avoid digital clipping.