digital - preferred recording level?

[psychicoctopus]

What level do you like to lay down on a digital recorder? Hot as possible w/o overs, very quiet, or somewhere in the middle? Am I correct in thinking that the recording will have less bit resolution if the signal is lower than digital full scale?

In other words, if you record on a 16-bit system, and your audio signal is only hitting 50% of full scale, will it only have 8-bit resolution because half of the bits are just recording zeros?

[NewYorkDave]

Can't really answer that without knowing what sort of music you're recording. If you're doing something like, say, jazz or classical, with an extended dynamic range, you'll of course wanna leave yourself some headroom. Rock music generally requires less headroom, but I don't care for the current mania for keeping the level up as close to 0dBfs at all times. THEN the mix goes to the slaughter... um, I mean, "mastering" house, where the last remnant of dynamic activity is summarily expunged.

Perhaps we can take some guidance from the longstanding broadcast practice of setting average level 20dB below maximum level. In digital tracking, that would mean keeping your average level around -20dBfs. In my facility--and this is common practice elsewhere as well--we use -20dBfs as our reference level. Conveniently, in most professional digital audio gear, the analog output voltage happens to be +4dBU (1.228 volts RMS) when the digital level meter reads -20dB.

I avoid using compression when tracking. But if you're tracking a compressed signal, you can of course afford to set your average level a bit higher.

[jajjguy]

Since there's no difference in tone when you record "hot" to digital, there's no harm in leaving plenty of headroom. It's pretty rare that you'd need all the dynamic range that 24 bits (or even 16 bits) offers, so recording average peaks at -20dB doesn't reduce your resolution at all. Even if you want to make the loudest record ever, you can still record at -20dB and pump it up in mixing. I've found no loss of resolution or fidelity of any kind when adding 20dB or more in the mix.

If you're talking about intentionally overoading a preamp or something, that's different, but the level actually recorded to hard disk doesn't matter much -- unless it clips!

[soundguy]

I usually cal my zero for -18dB. Ive never once over these years had an editor down the line complain or resolution or anything like that. Of course if I hand them tracks with overs, I dont hear the end of it, so... Err on the side of caution.

dave

[cyrusjulian]

I used to kill myself trying to record everything as close to the peak before clipping. And then when I would record like a demo or something I didn't really care about, I would notice that I couldn't really tell the difference between the two. So I agree with all the others that you should leave some headroom. However, I can say that I can totally tell the difference between 16 bits and 24 bits. I'd say try and get 24 bits if you can and that way you don't have to stress so much about getting a really good level. Oh well, just my opinion.

Cyrus

[jamoo]

Just per levels, if you record near peak you will almost always have to reduce your level. If you don't, you may not. There's a potential time/hassle savings I suppose. If you need to boost you always can, and the sound stays pretty much the same, just louder. From my experience, once you get the sound in the recorder it's all math. The art is getting it there.

..

What level do you like to lay down on a digital recorder? Hot as possible w/o overs, very quiet, or somewhere in the middle? Am I correct in thinking that the recording will have less bit resolution if the signal is lower than digital full scale?

In other words, if you record on a 16-bit system, and your audio signal is only hitting 50% of full scale, will it only have 8-bit resolution because half of the bits are just recording zeros?

[bigtoe]

i saw a good meter patched into a 3700. it's amazing how off the stack meters are on em... the same would probably go for MDM's and lower cost computer interfaces...

i peak at -6 on DAT and try to keep the same going for other mediums...

Off to work...working on sunday sucks.

Mike

[Mr PC]

I am recording drums using sm57's on snare and toms, a beta 52 on kick, and Studio Projects C1's as overheads. I run the kick, snare, and OH through a Sytek mic pre, and toms through the Digi 002r.

When going through the Sytek, I'm barely turning the knobs past 2.5 and am near clipping. It's not that I'm anal about getting close to 0db, but that it seems like I'm barely using the mic pres on the way in.

I'm playing the drums, and am not that heavy a hitter.

Could someone educate me here?

Mr. PC

[MoreSpaceEcho]

check this...

http://messageboard.tapeop.com/viewtopic.php?t=6706

[Cellotron]

In general I would go as hot as possible without risk of overs. For every 6db below digital zero you lose a bit of dynamic resolution - which leads to an increase in the apparent noise floor. A lot of time when people think they are recording at 24 bit they are actually recording as low as 14!

Digital recordings "distort" in 2 ways: when samples go over digital zero which leads to horrendous clcks and clipping noise - or when it tries to represent incredibly soft sounds with very little bit depth. This is the reason why people use dither - so that the fizzling out of reverb tails and fades can be hidden by a low level noise. Giving the most possible bit depth can help to minimize this problem.

For multitrack recording - the answer really lies in the multitrack summing algorithm you are using - not all DAW's sound the because they use different math to combine multiple tracks into one soundstream. It might be possible some of them might sound better when the multiple tracks are at a gain level that allows them to be summed without attenuating or adding gain to the overall output - but I might be barking up the wrong tree with this idea. Anyway - the DAW I use - SAWStudio - uses a slightly unusual 32bit double precision integer math with several calculations done at 64bit and sounds more transparent than most 32bit floating point DAW apps to me.

Best regards,
Steve Berson

[jajjguy]

One more thing: half of 16 bits isn't 8 bits. Each extra bit means you increase the dynamic range by some factor (probably 2), which means that 16 bits is like 1000 times more dynamic range than 8 bits. Okay, I'm making up the numbers, but the principle is that recording at half of full scale (instead of full scale) does not mean you get half as many bits of resolution.

[black mariah]

I record quietly, then bring the level up digitally. I can set it so that it makes it as loud as possible without clipping, and that's what I do. It doesn't affect the sound as far as I can tell, and it makes sure everything is nice and lound without squishing all the dynamics.

[psychicoctopus]

... so there is a relationship between signal level and bit depth, but it isn't linear when you look at it on the screen. Looking at cool edit's vertical scale (16 bits), it looks like -18 dB is about 20% of the height of the waveform display. And if each -6 dB corresponds to one bit, then even if you're only averaging 20% of the on-screen height, the recording has only suffered by 3 bits of resolution. Ok, now I'm satisfied