This guest post comes from Barry Gardner, mastering engineer at SafeandSound online mastering
The 24 bit advantage
These days almost all digital audio workstations have the option to record and operate at 24 bit resolution. In fact it is highly likely that the vast majority of musicians and engineers are already setting their DAW’s up to work at 24 bit. 24 has a number of advantages over operating at 16 bit. I am going to explain a very important advantage which may not be as obvious. It relates to mixing reference levels on your stereo master output bus. Firstly we have to look at record levels because this is where the initial advantages occur. When you record at 24 bit resolution you have a much greater theoretical dynamic range. This means that the noise floor of the recording in the digital realm is much lower than recording at 16 bit. It is theoretically 48 dB lower than at 16 bit. So as a consequence there is no need to record anywhere as hot as when using 16 bit. In fact, a suggested recording level for 24 bit would be to have an average signal level of -18dBFS. You could even have peak signals at this level without detriment to audio fidelity. With peaks at -18dBFS at 24 bit your noise floor is still theoretically 30dB better than that at 16bit. An additional bonus effect is that to obtain peak levels of -18dBFS your mic preamps, compressors and mixer channels will be operating at a lower electrical level which means a cleaner and clearer recording.
Ideal operating levels
-18dBFS = 0dBVU = 1.23V
This is line level and the technical sweet spot for all gear.
I always find this analogy of great value when speaking about digital signal levels and their relationship to analogue metering and electrical values (voltage). A large format console such as a NEVE, DDA or SSL has rows of VU meters present. A nominal channel operating level for many signal types is peaks at 0VU. An exception to this rule would have been instrument with lots of high frequency transient information such as hi hats and cymbals, these would have been recorded at -7VU. This is because the VU meter ballistics under read the level present on these types of signal and they used to more easily distort tape. On these mixing consoles (and in fact most 8 bus mixers of semi pro origins) 0VU would equate to -18dBFS on a digital meter. That might come as a bit of a surprise in part due to the logarithmic scaling on digital peak metering systems and -18dBFS looking very low as a reading relative to the size of the meter itself. However 0VU and -18dBFS referenced to a +4dBU operating level both equate to 1.23 Volts as an electrical potential. So this suggests that at 24 bit an optimal working level would be -18dBFS on your channel signals. It seemed to work for those working with a large format recording console so why not you in your DAW?
Headroom in mixing
All of this follows on nicely to mixing gain structure, the L/R bus output metering on an SSL console would often see something between 0Vu and +6Vu. This is perfect as the console’s electronics would have headroom to spare and so the monitoring electronics would be nice and clear. When you mix in your DAW I suggest an optimal working method is to peak an audio signal with lots of transient information such as a snare or kick drum to -18dBFS on the stereo master output. Then balance all other instrumentation against this reference instrument. Try not to be tempted to adjust the reference signals level too much or you will lose your reference. The beauty of doing this is you have now created ample headroom for the mix so you will not find yourself clipping or having to pull the master fader back at a later stage as you run out of headroom. Another substantial bonus is the analogue electronics in your digital to analogue convertor (audio interface outputs) will be operating at a less high signal level meaning clearer and cleaner monitoring. This means your mix decisions should be enhanced as a positive side effect. You may find you will need to turn your monitoring up to compensate for the lower operating levels. This is a small price to pay for the benefits from the point of recording all the way through to mixing. Don’t worry about maximized loudness at this point, whether you’re self mastering, or sending it to a local or online mastering engineer, you will be happy to have this extra headroom.
Your advice hinges on the idea that, “0VU would equate to -18dBFS on a digital meter”. This would only be true if you were metering RMS level in a dBFS context, which is almost certainly not the case in most DAW environments. If your peak amplitude is metering -18dBFS (which would be the product of your advice), your RMS level would certainly be well below the +4dBu = 0dBVU reference.
I would challenge anyone to come up with a theory-based justification for this in either an analog channel path, or a digital recorder (at any bit depth).
In reality “headroom” – i.e. reserved amplitude – in a digital audio recording environment metering peak amplitude (dBFS) is only useful to prevent inter-sample peaks. There are a lot of proposed ‘safe peak readings’ that would accommodate real levels, but -18dBFS is well below any theory-based recommendation I’ve read.
Hi Rob, thanks for the reply.
Firstly I invite you to look at the meter comparison diagram on this page:
http://media.soundonsound.com/sos/jun00/images/meteringfaq.l.gif
Theoretically it may be marginally low, agreed, however there is real world reasoning, in a recording situation there are a number of reasons to have the peak level even lower than 0VU. It will be fair to assume most readers of this blog are using semi pro kit where 0VU is referenced to +4dBu. (despite it’s professional signal level leanings, small mixers such as Mackie, Behringer, A+H, Yamaha etc. are still semi pro devices as we know.) In numerous cases I have personally heard distortion in such mixers when PFL’ing a channel to 0VU. As someone who has done a lot of recording I also know that once you have taken level you will find signal peaks up to 6dB hotter in many channels when an actual performance gets underway. That in itself is practical justification. Whilst the VU meter specification is meant to average out a signal level, those conforming to spec are rare. (and as such not commensurate with RMS metering) Being a meter with incredibly dubious ballistics, meter to meter and frequency dependent, this gives some additional margin for safety. (it did not receive it’s nickname “Virtually Useless” without good reason) That is without mentioning the ballistics of LED based metering which is much more common on low end equipment. Slightly lower level recording in todays incredibly low noise equipment is by far preferable than unintentional distortion caused by hot signal levels.The theory should not be misidentified as being of greater value or importance than the practical results achieved, and if you follow what has been written you will find repeatable, low distortion recordings across semi pro and professional equipment.
“In reality “headroom” – i.e. reserved amplitude – in a digital audio recording environment metering peak amplitude (dBFS) is only useful to prevent inter-sample peaks.”
As you suggest here, reality is somewhat more important than the theory which would be that digital headroom at input (at 24 bit) provides a great opportunity to ensure analogue distortion does not occur by over driving semi pro electronics.
-Barry
The soundonsound graphic is useful for relating various metering scales to a dBu reference, but that reference doesn’t reconcile the fact that the various scales are used to meter dissimilar signal attributes (particularly VU and any PPM scale). Those differences can’t be disregarded simply because there is a margin of error in execution.
If the gear sucks, let’s write about that. This just seems like a psych-out approach to something that could be addressed at face value.
Hi again Rob, from practical experience, I do not think it is possible to separate, analogue gain structure, digital gain structure, pro vs semi pro, theoretical and practical results as they are almost always interdependent on each other in any given recording /mix session. I think this is a valid way to set gain structure for the safest, most repeatable, low distortion recordings. For mix down the benefits are also clearly explained.(i.e. clearer monitoring in budget equipment)
Once you have distorted audio during recording you have just caused the worst problem you possibly can, it is the most difficult (often impossible) to disguise and correct. It is to be avoided at all costs.
I always think end result is most important, one can theorize until the cows come home it does not always have bearing on the end results (the above advice will) when it comes to art, audio, creativity and most importantly the music in this case.
It is quite easy to pick holes in things when taken at face value, understanding practical implications takes real world experience.