What the dBFS?
Various online audio hubs and Waves Inc. recommend -18 dBFS as the nominal input gain level for analog modeled plugins. dBFS standing for “decibels Full Scale” where 0 dB is the upper peak limit before digital clipping occurs and -18 is 18 decibels under 0 dBFS.
In short 0 dBFS is a digital reference level equal to “full scale”. It is used in specifying A/D – D/A audio data converters. Since all signals can reach a maximum of 0 dBFS we use a minus (-) to specify the negative level in relation to the maximum 0 (e.g. -18 dBFS).
A common mistake mixers do is to equate analog signal levels with dBFS. Quoting sengpielaudio (acclaimed industry engineer):
“That’s why there is no relation between dBFS and dBVU or dBu, whatsoever. Analog meter (ppm): attack time 10 to 300 ms – reading rms values. Digital meter: attack time < 1 ms – reading peak values. That is really some difference.”
The biggest confusion factor hereby lies in the fact that we have technical limitations and on the other hand human perception. The decision to use the analog meter (VU) with an attack time of 300 ms averaged value was seen as perceptually logical because it resembled the way we hear an audio signal. A digital meter, however, reads peaks drastically, which may be less intuitive compared to the VU or PPM meters.
As a rough guide to be taken with a grain of salt:
“European & UK calibration for Post & Film is −18 dBFS = 0 VU = +4 dBu“
As with many recommendations, we should think of them as conclusions of trial and error of experienced engineers recording/mixing/mastering audio and music using analog systems. Also, earlier audio hardware manufacturers will have liaised with recording/mix/mastering engineers and set ideal input levels for analog circuits to exhibit musically advantageous characteristics when processing incoming music signals. After all the 1176 was made with musicality in mind 🙂
Fast forward the advent of digital systems pushed the “headroom” and the upper limit of clipping to today’s 0 dBFS while back then many VU (Volume Unit) meters were calibrated to “0 VU” around where approximately today’s -18 dBFS peak is (loosely!). Hereby it is important to emphasize “peak” as digital meters are drastic whereas VU meters average the signal over a window of 300 ms.
Musical relevance of “older” systems
So why do we even care about analog systems when we have so much headroom in the digital realm? It’s because imperfections can be beautiful in a musical context!
Although there isn’t much beauty in digital clipping, analog distortion (in moderation) can lend your music pleasant characteristics like “saturation”, “warmth” & “edge” among many popular terms.
To have a little peek at what’s happening to signals when processed by analog modeled plugins we can run a sine wave in this case arbitrarily at 1 kHz generated using Ableton Lives operator. We keep the signal peak at -18 dBFS as recommended input level (monitored using Waves PAZ meters as seen below):
To test with I will use various analog modeled plugins that I personally like and will start off with the API550a EQ. Running the sine into the input while the plugin is turned off results in no alteration as one might expect showcased by Pro Q2’s frequency analyzer at a 120db high res and fast response setting:
We see the sine unaffected as it should be, but what happens if we turn the API on as seen below?
Still, no changes occur! even if the API is turned on with the analog button off. That means the API550a plugin does not introduce any analog distortion by nature.
So if we turn on the button we should see some harmonics and noise being introduced as shown below:
After turning on the analog switch on the API550a we can see the spike at 2 kHz introduced (harmonic distortion = “HMD”) at about -100 dBFS whilst the analog noise ranges from 5k to 20k at about -115 dBFS.
So having the input gain at nominal operating levels -18 dBFS does not introduce any visible/audible harmonic distortion at least on the API550a unless the engineer turns on the analog switch! In the case of using a moderate amount of analog modeled plugins across your mix I personally would have the switch on as the API in itself is very pristine so you shouldn’t be scared of over-hyped HMD and noise, because both HMD and noise were below 100 dBFS (which I doubt anyone could reliable perceive even with monitors maxed).
To have a more realistic scenario it’s best to see what happens if we stack 10 API’s:
Having stacked 10 API’s with the “analog button” ON the HMD increases by almost 20 dB and the noise by about 10 dB. This results in about 2 dB of HMD and 1 dB of noise that’s increased with every API that we add across our multitracks or insert stack. The harmonic at 2 kHz now comes into the audible range (if not masked by the louder signal). Also, we need to consider equal loudness contours as we hear the 2 to 5 kHz frequency range with more sensitivity.
A pre-conclusion here is that having worked at -18 dBFS with 10 stacked API550a, in theory, raises the HMD into the audible range. However, we need to consider that the harmonics will be masked by the timbre of the musical sound. Even if we raise the output level we would not just increase the harmonic distortion but also the actual timbre of the musical signal.
What if we didn’t really care about headroom and have signals incoming at -6 dBFS? Let’s consider a client sending us hot and peaky multitrack stems:
As you can see above the 1 kHz sine at -6 dBFS does not introduce audible/visible HMD with the analog switch off, meaning the manufacturers completely offed analog errors in terms of HMD and noise.
So, turning on the analog switch (as seen above) immediately reveals HMD at 2, 3 & 4 kHz and the noise from 5k to 20k. Having the input signal earlier at -18 dBFS introduced the 2 kHz harmonic at about -100 dBFS, however, the -6 dBFS input introduced a 2 kHz harmonic that’s about -80 dBFS. That’s more than 20 dBFS higher than working at recommended operating levels!
So naturally, stacking analog modeled plugins without taking care of conservative input levels can easily result in lethal distortion as seen in below screenshot where 10 API’s are stacked with input being -6 dBFS:
The 10 stacks now introduced another raise of 20 dBFS at the 2 kHz harmonic now being at -60 dBFS, which in isolation will be definitely audible, although in context with a mix would probably not be heard but could rather be felt as something being “harsh”.
In conclusion the API on its own delivers pretty pristine sound out of the box, yet I personally would still follow the nominal operating levels for a reason: musical signals are more complex than a single frequency sine wave, meaning that the “smear” and “spread” of HMD will be more complex across the frequency spectrum and in turn more drastically “felt” as being unpleasant or “harsh” if not used with a defined purpose or if overdone.
API550a VS. 1176
But do all analog units behave similarly? How would the API compare to an analog compressor in terms of -18 dBFS vs -6 dBFS input gain?
Unlike the API the 1176 blacky (shown above) immediately introduces various harmonics with the majority being odd at 3, 5, 7, 9, 11, 13, 15 kHz etc. with the even 2 kHz being fairly present at about -70 dBFS although we are at -18 dBFS input gain! Now consider how many of those babies we stack! I personally use around 4 of them in my mixes! And I’m not conservative with my trap beats! Now imagine a mix with 20 Multis and let’s assume you use 80% of various analog plugins, that’s 16 plugs, chances are you stacked plugs on a couple of them (vox, drums, HMX etc.) let’s say 5 of them that’s 21 units each adding harmonic distortion and noise! To keep it short it’s better to stay on the safe side and stick to the input level around -18 dBFS average. Luckily most analog modeled plugins come equipped with a reference VU meter which should make it fairly easy to follow conservative levels.
What about the 1176 bluey? Chris Lord Alge mentioned he used the bluey a lot on vocals because they are brighter in their sound. So how would it compare with the blacky when experimenting with the exact same sine wave frequency?
We see that the HMD is distributed fairly differently: The bluey also adds both odd and even harmonics with most being odd in relation to the 1 kHz sine wave. What differs though is the HMD from 9 – 16k being pronounced by the bluey, which may also explain why it is lighter (blue colored hardware) and overall brighter in its tone compared to the “beefier” blacky. Also interestingly the output level was reduced by 2 dB for the bluey, which I guess has to do with the true modeling WAVE’s engineers used when replicating the unit (for whichever reason).
That’s not to say one should be conservative in every mix: if you are using the plugin for that extra edge, let’s say you have sparse multitracks with very digital sounding character and you’d like to add some more edge or warmth to your mix then, of course, a higher input level could make sense. It all depends on the mixing vision!
Things to take away
- -18 dBFS indeed offers an “ideal” input level, especially when you utilize multiple analog modeled units and clients sending you “hot” and peaky multitracks.
- Some analog units allow turning off “noise/hmd” whereas others introduce them by nature
- Harmonic distortion increases with the number of plugins used (duh!)
- The 1176 blacky introduces more odd harmonics, hence may sound different in perceptual terms (beefier)
- The 1176 bluey introduces more “highs” and may sound perceptually “crispier”
- All plugins exhibit varying amounts of harmonic distortion and noise
- Comparing the API550a with the 1176 is like comparing apples with strawberries hence the rule of -18 dBFS input gain should be observed stricter when using “grittier” analog modeled units such as the 1176.
Limitations of my small experiment:
- Artifacts such as IMD/frequency response weren’t considered (using white noise or sweeps)
- Experiment was undertaken in standard 44.1 kHz / 16-bit
- Higher harmonics can’t be analysed therefore
- Tested only with 1 kHz sine wave although music is far more complex!
- Ideally need to experiment with more analog plugins (SSL, Pultec, Fairchild etc.)
Ultimate take away…
My experiment should just give you a rough idea of how analog plugins might affect your multis. Given the observable distortion with just a single sine wave, we should remember that musical signals with varying ADSR and complexity (timbre) introduce more complex distortion/noise. All these artifacts can result in either positive characteristics such as warmth, grit, beef etc. but also negative ones like harsh, distorted, piercing etc. when not utilized with defined goals.