Proper tracking levels

  • Thread starter Thread starter BBlack
  • Start date Start date
B

BBlack

New member
I've been recording my first few songs lately, and have quickly learned that you can't track too hot. But here's my question: Should every instrument peak around the same level? Or does the nature of the instrument and the type of sound it creates determine where it should peak?
 
You want to get a decent signal...but how close the levels are to each other when going in isn't that critical....especially recording to digital, because you will adjust your levels as needed when you mix, and in the DAW the level adjustments are simply mathematical moves.

If you are tracking to tape...then your levels going in might need to all be at least hot enough to get a good S/N ratio, but again, they don't all need to peak exactly the same...it's not critical.

I shoot for “ballpark” same.

Now….if you are recording a full band/ensemble of musicians, then each instrument will tend to be at the level that is needed for the right room balance…though again, if everyone has headphones, you create whatever cue mix you desire.

Probably only in a simple stereo-pair kind of recording of an ensemble will each instrument be at exactly the level it needs to be going in, because you are creating the mix live.

Sorry if that was more info than you were looking for. :)
 
BBlack said:
Should every instrument peak around the same level?
Click to expand...
Only in that they usually should all peak somewhere less than 0dBFS. Other than that, peak levels are fairly meaningless, both by themselves and compared to each other.

No they do not have to be at the same level; each track should be optimized and tracked based upon it's own gain structure needs; what the tracks around it look like is irrelevant.

Wait until the mixing phase to worry about getting the relative levels between tracks right, and even then worrying about matching peak level is barking up the wrong tree.

G.
 
Thanks everyone, it all makes sense. Just as long as it doesn't go above 0dB, really?
 
The only level that means anything is line level, which is an average reading. Peak is meaningless as long as it's under clipping.

What you do is make the sustain of an instrument sit around -18dbfs. Then just let the peaks fall where they do, as long as they don't clip. If they are too close to 0dbfs, just turn it down until it's safe.

Percussive instruments like drums don't have much sustain, so shoot for a peak around -6dbfs or so.

If you have an instrument that doesn't have any dynamics, like a synth on a sine wave patch, the peak and the rms are about the same. So the recording level should peak around -18dbfs or so.

It all depends on the crest of the sound you are recording.

Setting recording levels is like playing horseshoes, getting close counts.
 
back when i went to audio engineering school it was all analog, Tascam stuff was just breaking out, and we were instructed to use the stereo VU's to 'add up' the mix; we would always get the kick and drums first, then the bass guitar, and there were different VU values that we strived for-iwish i could find some of those old notes-but kick was like -7, guitars were like -5, vocals and keys were different, and when you turned all the channels on after setting each up separately, it would 'add up' to what looked like a great level on the stereo buss, which naturally was connected up to a 2T mixdown machine of some variety. (preferable a 30ips Revox 1/2track reel to reel in those days). so today with digital multitracks storing our raw tracks, and mixing perhaps even done in the digital realm, those value-oriented tips per each sound source would not apply. today a calibrated PFL buss is all you use to set your gain, so it seems the individual separate VU values per each channe'ls signal does not seem as critical; and certainly the mixdown is not as labor-intensive and performance-critical as it used to be.

if only back then there was such a thing as "Save as"...
 
When tracking I try to keep all my peaks around -10dBfs

There is no real advantage to tracking with hotter levels than this, and I have seen plenty of tracks ruined by tracking close to 0 even when they do not actually hit zero.

There is no downside to conservative levels, and many potential problems with hot levels.
 
BBlack said:
I've been recording my first few songs lately, and have quickly learned that you can't track too hot. But here's my question: Should every instrument peak around the same level? Or does the nature of the instrument and the type of sound it creates determine where it should peak?
Click to expand...

No, you dont need all meters to read the same but, ideally, they should be close.

Understand a critical difference...there is the level you record, and then there is the level you set for playback/mixdown. If you try to track a signal and adjust the input level for what sounds like a good mix, you could end up tracking at -58 db.

The correct way to record is to adjust the source to output a level that's well below it's clipping point, but well above it's noise floor. Then adjust your input gain such that you see average meter levels of around -8, maybe -12 db. This will give you enough additional input headroom that, if you get a level spike from the source, you're not likely to hit 0db on the input level meter.

The objective is to get a nice. clean (ie.. not overdriven at the source) input signal and to record that signal at a good robust level but never hit 0db. In the digital world, 0db is all bits on. Where overdriving/clipping analog tape would saturate fairly smoothly (and much of the distortion this might create would be even harmonics...ie...reasonably pleasant sounding), clipping in the digital world sounds very ugly (lot's of odd harmonics).

Best to have a hot enough recorded signal that allows you to turn it way down in the mix, if you need to, than to have to low a recorded signal that forces you to have to crank the fader all the way up.

Here's why...if you record at too low a level and find that you need to "increase the gain" by normalizing or some other digital gain increase...because you recorded at a very low level the signal to noise differential (there will always be some noise from the analog stages) will be smaller. In other words, your signal level will not be much above the noise floor of the input analog stages.

As a result, if you normalize the track, the signal level will be increased but the noise floor (which you also recorded) will be increased by the same amount.

In short, if you can look at the levels of all the recorded tracks and they're all floating at around -8 maybe -12 db, with no peaks that reach 0db, then you've got good levels to work with.
 
Thank you, that last post was fantastic. I really appreciate it.

One thing - are you all referring to a peak meter or an RMS meter when you're suggesting levels? Not sure if this is dumb question or not, as I'm not sure which is more common. I'm using Logic, and I'm pretty sure it uses peak meters.
 
WE are for the most part talking about average levels. Most of the distortion and problems of tracking too hot are caused by the analog signal chain leading up to the conversion process. When you over-stress that, you are giving the converters a wonky signal to convert.

This is a decent way to get your levels correctly. Remember, it doesn't matter what the peak level is as long as it doesn't clip.

The meters in Logic are peak meters.
Farview said:
The only level that means anything is line level, which is an average reading. Peak is meaningless as long as it's under clipping.

What you do is make the sustain of an instrument sit around -18dbfs. Then just let the peaks fall where they do, as long as they don't clip. If they are too close to 0dbfs, just turn it down until it's safe.

Percussive instruments like drums don't have much sustain, so shoot for a peak around -6dbfs or so.

If you have an instrument that doesn't have any dynamics, like a synth on a sine wave patch, the peak and the rms are about the same. So the recording level should peak around -18dbfs or so.

It all depends on the crest of the sound you are recording.

Setting recording levels is like playing horseshoes, getting close counts.
Click to expand...
 
My guidelines:

If every track has the fader at 0, my master buss should not clip.
If every track has the fader at 0, the mix should at least sound like it's in the ballpark.


I usually record my first tracks with the body around -18 (peaks fall somewhere near -12 or -8). Then every track after that I set the recording gain to record the new track at the appropriate relative volume to what already exists. It's like mixing with the mic gain knob.

The awesome upside is that you can really hear the song come together and know instantly if your overdubs are the right tone, and you can perform your final mix without any meters or visual feedback of any sort since you will never clip.
 
RPRecording said:
The correct way to record is to adjust the source to output a level that's well below it's clipping point, but well above it's noise floor. Then adjust your input gain such that you see average meter levels of around -8, maybe -12 db. This will give you enough additional input headroom that, if you get a level spike from the source, you're not likely to hit 0db on the input level meter.
Click to expand...

The correct way to record is to adjust the source to output a level that's well below its clipping point but well above its noise floor. Adjust your input gain such that you see average meter levels of line level. If your digital converters are calibrated to -18 dBFS line level, this means -18 dBFS. Going to -8 or -12 means that you're overdriving your input chain by 6 to 10 dB and, especially with modest preamps and converters, causing the audio to sonically fall apart before it even hits digital, which doesn't care and has extremely low noise floor specs, but will record the damage you've done in all its pristine glory.

Recording at line level (or even a few dB's under, on the conservative side) helps to keep things sounding bigger, cleaner and gives you more control in post processing. In 24 bit systems, recording at -28 instead of -18 wouldn't give you a big enough drop in level to be concerned with noise floor.
 
BBlack said:
Thank you, that last post was fantastic. I really appreciate it.

One thing - are you all referring to a peak meter or an RMS meter when you're suggesting levels? Not sure if this is dumb question or not, as I'm not sure which is more common. I'm using Logic, and I'm pretty sure it uses peak meters.
Click to expand...

Welll....the short wise a$$ answer is....both....and neither! Thats really helpful isn't it....seriously though, we are talking about both sort of.

RMS is AVERAGE level but on a DAW, there really is no RMS (a DAW get's bit's, not voltage or current so there's not really an equivalent...consider two bytes of data... 11110000 or 00001111. The average number of bits set in either of these is 4, but the two bytes represent VERY different values so, applied at this level, the very term, "AVERAGE", becomes meaningless).

What folks are saying is get the meters to sit, ON AVERAGE, around -8 to -12. At the same time, be sure that it NEVER PEAKS at or above 0.

In a DAW, meter levels start to get really weird. On an analog console you would hear references to +4 dbu or 0dbV etc.... The confusion stems from the fact that a "db" is a meaningless thing. It's not a specific number, it's a ratio. In the case of dbu or dbv it's a ratio of a signal level TO a specific electrical value. Electrically, a value of 0dbu=.775 volts.

In the DAW world you'll hear folks refer to 0dbfs. This stands for 0 db's, FULL SCALE. This refers to the point where all bit's are on (which, depending on how the analog stages are calibrated, could be just about anything, voltage wise). While it doesn't refer to a SPECIFIC voltage, digitally speaking, it doesn't matter. This is the point at which all bits are on and whetever voltage it takes to get this, the fact that all bits are on means that NO HIGHER VOLTAGE CAN BE REPRESENTED. You've reached the point of turning sine waves to square waves.

Just keep it simple. get your meters to sit around -8 or -12 but never peak at or above 0.

If you're into the techie stuff you could pick up a book called "Studio Recording Procedures" by Mike Shea. He get's pretty deep into the electronic engineering aspects of recording. A good book but some of this stuff will give you a headache.
 
Last edited:
RPRecording said:
RMS is, more or less, AVERAGE level. What folks are saying is get the meters to sit, ON AVERAGE, around -8 to -12.
Click to expand...
There *is* a correlation between dBu/VU on the analog side and dBFS on the digital side. And that is the lefel in digital dbFS that the A/D converter will output when receiving an input of x amount of dBu. Typically the calibration between the two of them that most engineers look at is the equivalence at 0VU (typically +4dBu). This varies from machine to machine, and can range anywhere from -14dBFS on some older European gear, to -15dBFS for DAT/ADAT specifications to -18dBFS for many early 2000s interfaces sold in North America, to -22dBFS or higher for some all-in-one DAW boxes sold in Japan or late model devices sold elsewhere.

Typically, folks in the HR community round it off to 0VU/+4dBu analog converting to -18dBFS (+/- 4dB). This means that to *average* or RMS around -8dbFS (with digital unity gain) would mean averaging or RMSing on the analog side at around +10VU (+14dBu), which is rather hot. It also means leaving only 8dB of crest factor headroom on the digital side to fit the peaks in without clipping, which is cutting things pretty tight unless one is applying some pretty moderate compression during tracking.

I think there's a problem with semantics here. If one says the average peak level - that is the average level of the transient peaks themselves - is somewhere around -18 to -12 dBFS, that would sound pretty good to me. But an actual RMS or average energy level of -8 to -12 dBFS is coming in rather hotter than I would recommend.

G.
 
Hhhmmm... I think folks are over complicating this. The OP said that he was just recording his first tracks so inundating him (assuming it's a him of course) with talk of dbu, dbm, dbV or dbv (which is really the same as dbu) may not be particularly helpful.

The fact is that if he's looking at the meters on his DAW software mixer, then there's a better than 99% likelihood that he's looking at 0 dbfs PEAK levels, not RMS. And that, at this level, there is NO correlation with any real world voltage or power value. The correlation between real world voltage or power input and 0 dbfs output only exists at the audio interface where the actual analog voltage signals are converted to digital. Once the audio interface does this conversion, there is 0 dbfs (all bit's on). The DAW software gets the digital stream and has no way of knowing whether the converter was calibrated to output 0dbfs given a -10 dbV or a +4 dbu input.

Moreover, for many of the audio input devices available, you'll find that the manufacturer may not provide anything that actually gives a calibration spec for exactly what input level will produce a digital output of 0 dbfs.

Take a look at the spec sheet for an M-Audio 2626...

http://www.m-audio.com/products/en_us/ProFire2626.html

You can find line output levels but nothing that lists what input voltage or power level will provide a 0dbfs digital output. Similar thing with the Focusright Saffire Pro 40...

http://www.focusrite.com/products/saffire/saffire_pro_40/Specifications/#Specifications

RME is pretty good about this. As you can see, for the Fireface 800 they provide specs for the various software gain configuration(s) available in the device...

http://www.rme-audio.de/en_products_fireface_800.php#5

Specifically...

Input/Output level for 0 dBFS @ Hi Gain: +19 dBu

Input/Output level for 0 dBFS @ +4 dBu: +13 dBu

Input/Output level for 0 dBFS @ -10 dBV: +2 dBV

Dont over complicate this

In short, if there's a RED led on your audio interface thats labeled "Clip" you should probably avoid sending a signal INTO the interface that causes that LED to light up with the input gain knob turned all the way down.

Secondly, dont crank the gain knob up so high that the RED LED comes on. Do this, and you've probably adjusted the audio interface to get good levels, with out clipping, at the audio interface level.

Both of these statements assume that your input source is playing as loud as it's likely to go.

Now look at your DAW meters. If you've set up your inputs this way, then in all likelihood, with the DAW mixer channel fader set to 0, you will probably be seeing peak levels below 0. If they're just barely below 0 you might turn the gain, on your audio interface, down a bit more until your DAW's PEAK reading meters show you getting PEAKS of around -8 to -12.

BTW...I would second one of the previous post's...Bob Katz's book is excellent. Aditionally, if you go to his web site (www.digido.com) , you can download a stereo wav file thats a calibrated -20 dbfs pink noise burst...

If you import that into a channel on your DAW, set that channel and your Stereo out bus both to 0, your stereo out meters should read -11 db.

The file is -20 dbfs but, again, your DAW probably is a peak meter (like the ones I see in Cubase).

Lastly, 0VU is not the same as +4dbu. By the spec (ANSI C16.5-1942)...

Volume Unit (VU) defined: The reading of the volume indicator shall be 0 VU when it is connected to a 600-ohm resistance in which is flowing one milliwatt of sine-wave power at 1000 cycles per second

Source..
http://en.wikipedia.org/wiki/VU_meter

It's the same as 0 dbu...

dBu or dBv

dB(0.775 VRMS) – voltage relative to 0.775 volts. Originally dBv, it was changed to dBu to avoid confusion with dBV. The "v" comes from "volt", while "u" comes from "unloaded". dBu can be used regardless of impedance, but is derived from a 600 Ω load dissipating 0 dBm (1 mW).

Source..
http://en.wikipedia.org/wiki/Decibel
 
SouthSIDE Glen said:
...the average level of the transient peaks themselves - is somewhere around -18 to -12 dBFS, that would sound pretty good to me. But an actual RMS or average energy level of -8 to -12 dBFS is coming in rather hotter than I would recommend.

G.
Click to expand...
Yup.



(Character limit)
 
You must log in or register to reply here.
Back
Top