I’m going through my notes today and disposing of a few questions.
What levels do you record at and how do you allow for headroom? I’ve heard competing ideas on this. One, a video explained that recording should be done at about -20dB. But doesn’t this question revolve around how many tracks are ultimately mixed? Because we are adding dB together with each additional track. A sound engineer once implied that it’s important to get the strongest signal possible. This contradicts the -20dB rule.
What do you think? Are there any other contingencies to consider?
You need to pay attention to your environment and your performance. If you have a nice quiet room and you’re announcing an audiobook, I like occasional voice peaks somewhere in the -10dB to -6dB range. This is visually handy because it’s half-way up the size of the blue waves on the timeline. You should be wearing good quality, sealed-against-the-head headphones feeding yourself back in real time which can do a terrific job of keeping your volume from wandering during a chapter read.
If you have no live monitoring and a completely wild performance, then yes, peaks in the -20 range are probably justified. We can help you with high background noise from recording too low, but nobody can help you with peak overload or clipping. That sounds terrible and it’s forever.
A sound engineer once implied that it’s important to get the strongest signal possible.
And he’s right. You have to fit the performance into the space between too loud clipping/overload and too quiet lost in the ever-present noise. I say “you” because I know it’s popular to YouTube a standard practice and go with it, but when you press the record button, you are the recording engineer. The video poster isn’t going to be there to tell you what you’re doing wrong.
There totally is technology that can handle all of these jobs unattended. Cellphones do this. Do you want your theatrical or musical performance to sound like a cellphone?
One of the YouTube shows I like featured a character who turned out to be a broadcast techie. They got so many comments about how they got the videos to sound so good, they posted videos with the equipment listings. One of the items that caught my attention was a lavalier (chest) microphone which, in addition to being very well behaved, also made two recordings at different volumes.
Mixing considerations are different. Audacity doesn’t overload when you’re mixing. It uses 32-bit floating sound format internally which has effectively no top end.
The bouncing sound meters monitor everything all the time unless you stop them and you can use those to set mix volume while you’re working.
We should remember that tracks don’t just “add up” even if the show is several players playing the same tune on different instruments. The blue waves are always going to be slightly different and the differences prevent addition.
If you’re doing pure electronic music and you intentionally arrange multiple perfect copies of a single song, then you may get addition. And even if you do, there will be no audible effect. It will sound just like one instrument, no matter how many you add.
When you get your mix so it sounds exactly right, Mix and Render to a single track and final adjust the volume so it’s compatible with your export standard.
Digital recording levels are not that critical as long as you avoid [u]clipping[/u].
Your ADC (analog-to-digital converter for recording) and DAC (digital to analog converter for playback), regular (integer) WAV files, and CDs will hard-clip if you “try” to go over 0dB.
Nothing bad happens when you get CLOSE to 0dB and nothing bad happens unless the level is VERY-VERY low to the point where you can hear quantization noise. (You might hear quantization noise if you record at -40dB and at 16-bits but normally it’s nothing to worry about.)
Headroom is a funny thing… If you don’t use the headroom you didn’t need it and if you do use it it’s no-longer headroom! We leave headroom because analog levels are unpredictable, especially when recording live.
You generally do want a strong acoustic & analog signal for the best-possible signal-to-noise ratio and a (unexpected) low level is often an indication of an analog problem. There is no problem with reducing the signal (together with the noise) just-before it goes into the ADC.
It’s common for pros to record at -12 to -18dB. I don’t know if that’s just to make darn-sure to avoid clipping during recording or if it’s “tradition” from the early digital-days when Pro Tools used integer data and they wanted to leave room for mixing/processing.
You are correct that Mixing is done by summation (analog mixers are built-around summing amplifiers) and many effects/processes can also push-up the levels. Audacity (and most DAWs/audio editors) uses floating-point for " internal" processing so it can go over 0dB with virtually no upper (or lower) limit. You just have to be careful to avoid clipping when recording or exporting (and sometimes when listening to what you’re editing).
A sound engineer once implied that it’s important to get the strongest signal possible.
That was true in the days of analog tape because you needed to overcome the tape noise. And, analog begins to soft-clip as you go over 0dB (and it can go over 0dB) so it was common to go occasionally “into the red”. …There also seems to be a myth in pro circles that ADCs have a “sweet spot” and “sound best” when you leave a certain amount headroom.
You need to pay attention to your environment and your performance.
A lot of those “ideal” rules apply to studio recording. If you’re recording at home, you have a lot less elbow room to make mistakes.
The longest forum message ever, 39 chapters and over a year was not somebody trying to cut a track at Glen Glenn Sound Studios. It was Ian who just wanted to record an audiobook from his apartment on La Brea in Hollywood.
We recommend setting your recording levels for a peak level of - 6 dB. This allows 6 dB of “headroom”, which should be sufficient so long as you are careful with your settings.
Most consumer level sound cards are 16-bit, which provides a theoretical maximum dynamic range of 96 dB, and usually significantly less in practice, so you need to avoid allowing too much headroom in the original recording.
Professional recording gear is generally 24-bit, with an actual dynamic range greater than 100 dB, so more headroom may be safely allowed. It is quite common for recording engineers to allow 20 dB headroom for live recordings.
When mixing tracks, it is usually necessary to reduce the level of each track so that the mixed level does not exceed 0 dB.
OK. I am using a Tascam DR40 set for 24 bit. I also set the backup recording at around -06 to -10 dB. I probably should move up to external mics to get a clear signal? The little onboard mics are pretty dern good, though!
Someone said -20 and someone said -06. Does the total number of tracks in the project bear relevance? For the -20 people and the -06, how many tracks are you typically using?
The important thing about headroom is that the peak level never reaches 0 dB.
The suggested peak level of -6 dB (6 dB headroom) is usually sufficient to ensure that the peak level never reaches 0 dB, though it depends what you are recording.
Live music recordings can be problematic, especially with amateur performers, who frequently get much louder once they get going than they do during the sound check. In such cases it can be wise to allow more headroom.
When recording things like records and tapes, the peak level is usually fairly predictable, so it’s easier to guestimate the recording levels required to ensure that the level never reaches 0 dB.
I select Show Clipping and do some premixing along the way to see how all the tracks peak when mixed. Is there some reliable math? I understand that 10dB + 10 dB = 13 dB, an addition of 3 dB. Is that correct? And how would the math be for 3 tracks of 10 dB? 16 dB?
Except for special cases where the tracks being mixed are related to each other, there is no way to accurately calculate what the peak level of the mix of multiple tracks will be.
An example to illustrate the problem:
Generate a tone
Duplicate the track
Select both tracks and “Mix and Render”.
Note that the mix track is about 6.02 dB higher than the original track
Undo the Mix and Render step
Select only the second track
Apply the “Invert” effect
Select both tracks and “Mix and Render”.
Note that the mix is silent.
