Recording levels

Goreliscious

Senior Member
I was on another forum asking about recording guitar cabs when someone said that recording as high as possible before clipping, (around -1dB), is too high for digital recording and it's what you do with analogue recorders.

They posted a link to an article, (the link posted below if you want to read it), with techy jargon I don't understand but in general the jist is that with digital you should record between -20dB and -12dB to leave yourself headroom.

Does this sound right in general/something I should follow for drum recording too?

http://www.massivemastering.com/blog/index_files/Proper_Audio_Recording_Levels.php
 
I agree, never red line. When recording on tape they had to saturate the tape with as much signal to rise above the tape hiss and noise, recording on to digital does not require this so rather than strain the signal (think of truck tires on VW) you keep it below 0db.
I record my drums with a digital portable and the sound quality of the recordings sound very good. After I put the recording through a program, in my case AudioMagix I now have the option of tweaking it without worry of creating distortion I have an example here and please, I am not adding this drum part to be analyzed but to show you an example of what I am saying. This is recorded on a $200CDN Roland on a house kit in a reherasal studio.

http://soundcloud.com/user4976306/dead-heads-drum-solo-yamaha
 
yep that advice is pretty good
for guitars and most other instruments keeping the input between -20 to -12db
is a good rule of thumb.

the reason is two-fold: headroom and focus
The ad/da converters in digital recording are calibrated to a zero scale at -18db
what this means is -18db is the level the input expects and is designed to preform it's best at.
 
Really good answers thus far, just to add a couple of thoughts -

Keep in mind that most DAW metering is peak, where analog decks are RMS or averaging meters. With -18 being the typical digital equivalent to 0db on an analog deck - know that on your loudest sounds your peak meters will and should read higher than -18 - as much as 6-12db higher with drums Which still should put your drum channels never rising above -6db to -8db - and that's full tilt maximum. And unlike analog tape systems where there were tangible audio reasons for sometimes running that hot - there are really none when it comes to driving a digital system that hot.

Best advice - poke around your DAW and see if you can set the metering or run a plug-in that displays with RMS or average metering - oftentimes you can find something that will actually show both at the same time. I don't think you need to run it all the time, but lokking at your signals through an RMS meter lets you see all this in action much better. And once you have the gist of it, you should be fine with just standard peak metering.

One last thought - this whole discussion of levels has talked in terms of analog vs. digital, but the reality, all of this is only applicable when discussing analog vs, 24bit digital. And if you can, you should always be recording 24bit.

The difference between 24 bit and 26 bit is HUGE. So if you have an older or budget system that forces you to track 16bit - then you'll really need to dismiss most of this. 16 bit just doesn't have a dynamic range that allows you to throw away the top 10-18db - the low level resolution just simply isn't there to do that. So with 16bit, I would advice doing the 90's tap dance of recording as hot as possible without going over.

With the far better advice being, always record 24bit whenever possible.

David
 
Cheers for the replies. The techy jargon and acronims went over my head but I got the jist. I have a basic "16-bit uncompressed" 8 track so I guess I'll ignore the -18dB guide. Having said that though I mic'd my guitarists amp at -12dB then at -1dB and the -12dB came out better. Though I did the same with my snare drum and I couldn't tell the difference.
 
Dcrigger is absolutely right. If you're using a 16-bit system, keep the headroom. If you can change your system to 24-bit, then you can afford less headroom. I have some software that runs at 32-bit, but that's not particularly commonplace.

The number of bits determines the total dynamic range available to you. If you imagine the amplitude (volume) as 'steps' then a 16-bit system has fewer 'steps' to the top than a 24-bit system. They get to the same level ultimately, but the steps in 16-bit are larger and hence less accurate. If you want to get even more technical, you also have the 'sampling rate' which determines the number of samples of sound taken per second.

If you imagine sound as a wave, a computer can't actually process the entire wave. So, instead it takes a 'sample' of the wave at defined periods. By something called the Nyquist-Shannon theory, the number of times a sound is sampled a second is approximately double that of the maximum frequency produced on playback. If you assume that the maximum frequency Humans can hear is 20KHz (which is what is stated, although it's usually lower) then you need to sample the signal at 40KHz - i.e. 40,000 times a second. In practice, there is some space allowed, so the lowest modern sampling rate is usually 44.1KHz. For things like phone services, the encoding is actually much poorer, but that's to save bandwidth and isn't necessary for the purpose.

A CD therefore, is encoded at 16-bit and 44.1KHz sampling rate. Which is adequate for good replication of sound. There are now much better technologies (including DVD and Blu-Ray) that have much higher theoretical bit-rates and sampling rates.

This is all basic digital audio theory. If you have questions, feel free to ask. This is part of my degree.
 
Cheers for the replies. The techy jargon and acronims went over my head but I got the jist. I have a basic "16-bit uncompressed" 8 track so I guess I'll ignore the -18dB guide. Having said that though I mic'd my guitarists amp at -12dB then at -1dB and the -12dB came out better. Though I did the same with my snare drum and I couldn't tell the difference.

Yes in your case, you'll need to keep the levels up as best you can. That said - it isn't as though every below -10db on a 16bit system is horrible, because that absolutely isn't the case.

The degree any of this will be a problem for you is when you have parts that have wider dynamic variations in them - say a snare drum part that is whomping backbeats most of the way, but has a little whisper quiet part in the middle with delicate little rolls. Something like might have a 25db dynamic swing as played - and recorded it peaking at say -4db would make the softest part -29db, which though down there should still sound OK. On the other hand, set he loud peaks at -12 and then the soft part is sitting at -37 - which is not the greatest of fidelity.

But... we must keep in mind, lots of pretty great sounding records - including orchestral ones - were recorded at 16 bit, and obviously CD's are 16 bit... so it isn't that horrible down there, until you get way way soft.

So don't overreact to this need to keep levels hot - because rule #1 with digital - 16 bit or 24 bit.... is never, never, never go over zero (and if you've hit zero - you've gone over!). All long way of saying having your guitar hit -12 was probably perfectly fine - and having the snare hit at -1 is just too risky when tracking a live player. Backing that off a bit will be much safer and still sound perfectly fine.

Have fun,

David
 
The only way I found to truly know the exact level that distortion occurs is to measure it. When I get a new piece of equipment it goes on the bench with a signal generator inputted and both an AC millivoltmeter and oscilloscope on the output of the unit. I can then accurately measure what signal level, compared to the units internal VU, peak level indicator or graphs, at which point distortion first occurs. I've found many different devices where their meters weren't calibrated as accurately as they should have been. In the analog world recording to tape, you could easily pin the meter and saturate the tape to 3% distortion and it will add phatness to the signal. In the digital world of 0s and 1s it readily turns into indistinguishable garbage, lol. I still like to track and mix to tape. After this process it can go all digital if need be since all the basic parameters are set.

I do record fairly hot to tape, usually at 250 or 320 nanoweber per meter and depending on the tape formulation, I still have 10 or 12 DB of headroom. On a digital device I usually calibrate it's meter system between -10 to -12 below 0 using a 1K sign wave when tracking.

Dennis
 
Cheers for the replies, much appreciated. So it sounds like around -12dB to -6dB is gunna suit my drums? Thank god baldness is in the family otherwise when it finally happens to me I'd be looking back thinking this bloody saga of understanding phase cancellation and 16 vs 24 bits did it to me!
 
Back
Top