Best Recording Levels

[Bob Olhsson]

I don't know about noise getting dropped off. If you take a -100 1k sine-wave and truncate it, you will indeed get silence. If you dither it, you'll still hear the 1k in the noise!

A problem with masking is that its effectiveness is dependent on the condition of the listener's hearing. People having hearing damage can hear lots of stuff those of us with healthy hearing don't because of the missing frequencies in their hearing that would be needed to mask information.

I'm hardly an expert but its clear from what I've read and from conversations I've had with some real experts that most of what gets written about this stuff is way oversimplified and often very wrong. Among other things it was made clear to me that while dither sounds like noise, noise can't dither unless it has a very specific probability distribution, level and spectral content. My own experience has been that I can often hear truncation of the 24th bit in a recording from analog tape where the noise level is only around -60.

I do totally agree with you about the importance of reasonable levels. A lot of common converters sound like crap if you are steadily hitting them in the top 6dB. This is probably about cheap power supplies and lack of headroom in the analog stages.

[PTUser NYC]

Originally Posted by Bob Olhsson

Well a whole bunch of real scientists say no, that's not what dither does.

Dither randomizes the bottom bit making quantization sound like noise rather than like a chattering noise gate. Yes the result sounds like noise but the dither has in fact prevented the distortion and is not merely masking it.

Thank you Bob, of course you are right.

Dither is added before the word truncation, not afterwards. Masking would be an effect of noise afterwards.

Still I think (hope?) I was right about everything else I wrote. The main point being that observing 0dbVU is better than recording really hot to preserve resolution that is already eaten up by noise in the signal being recorded. And that the noise in the signal DOES mask the lower noise in the converter, and so further resolution is impossible past that which is present in the signal.

Thank you for the correction, i was throwing something in and I admit I didn't really think it through. It's always good to be corrected and learn. :)

[M-ManLA]

Originally Posted by DrFord

My 24 bits.

HAHAHA Nice!!! I should start using that.

[Carl Kolchak]

Originally Posted by O.G. Killa

Well his room is incredibly loud to sit in. It was even worse when he was over at the Mastering Lab and you'd actually sit in front of him. Now in his new room, you sit behind him. So it's a little bit more bearable, but it still feels pretty loud. It's also not like he continuously listens that loud. but that's where he makes all the major decisions about the mastering.

You can hear some recent stuff he's mastered for me in the last year or so. See the links below...

http://www.megatrax.com/core/cd_deta...pe=The%20Scene

http://www.megatrax.com/core/cd_deta...pe=The%20Scene

http://www.megatrax.com/core/cd_deta...pe=The%20Scene

http://www.megatrax.com/core/cd_deta...pe=The%20Scene

http://www.megatrax.com/core/cd_deta...pe=The%20Scene

http://www.megatrax.com/core/cd_deta...pe=The%20Scene

http://www.megatrax.com/core/cd_deta...arquee%20Music

http://www.megatrax.com/core/cd_deta...arquee%20Music

http://www.megatrax.com/core/cd_deta...arquee%20Music

Hi Derek,

that's all really top notch stuff - well done!

I didn't meter anything, I just listened, and everything gives the impression of being very dynamic and open (even if technically it has a narrow dynamic range).

It all sounds really nice, with great depth, particularly in the floor toms. And there's still plenty of thwack without any distortion or shredding going on, even on the guitars etc.

Out of all the tracks I listened to, only "Dr Spook" & "I'm in Heaven" seemed to have noticable squish or pumping going on - and I'm certain that was an esthetic choice you went for in the mix, before it went to mastering.

I really like the work both you and Gavin have done here!



In terms of how this realtes to this thread, and for the benefit of others:

I think that's one of the advantages of having your monitors calibrated that high:

You're forced to create more dynamic masters, because it's physically unbearable to crush it all up to 0dBFS.

The downside is that, in less experienced hands there would be a danger of mastering at too low a level for a 16bit format, where you could re-introduce the noisefloor, and exacerbate it by requiring the end user crank up their noisy old amp.

Conversely, if your monitors are calibrated too low, it's really easy to squash the living daylights out of a master, in an attempt to make it appear loud.

Those are both pitfalls people need to be aware of, and I'd recommend (certainly for tracking / mixing) calibrating to 83dBSPL for music.

If you're doing post, it's generally considered 85dBSPL for movies on a proper mixing stage, and 79 - 82dBSPL for T.V. in a smaller mixing suite - calibrated from -20dBFS RMS pink noise, C weighted, slow.

Hope that helps.

[Fullsound Recording]

Get your audio in the yello and your good

[Picture Start]

Originally Posted by Bob Olhsson

Well a whole bunch of real scientists say no, that's not what dither does.

Dither randomizes the bottom bit making quantization sound like noise rather than like a chattering noise gate. Yes the result sounds like noise but the dither has in fact prevented the distortion and is not merely masking it.

Yes, because distortion is correlated to the signal and noise is, by definition, not. A fundamental difference that sounds like splitting hairs, but an important distinction. Randomization by noise changes the LSBs fundamentally from one form of matter, if you will, to another.

[maxcarola]

Originally Posted by O.G. Killa

What noise? Your mic preamp that you are cranking up is going to have more noise in it than your protools system. Most mic preamps have a noise floor of about -90dB, protools is around -118dB to -120dB. Your mic and mic preamp will have AT least 30dB MORE noise in it than your DAW. So, if you are trying to record at louder levels the only thing you are doing is putting MORE NOISE into your tracks because you are turning your mic preamps up to get the signal as hot as possible without clipping.

If you are recording into a DAW, you are NOT USING TAPE. There is no need to record as hot as possible since there is no tape noise/hiss to compete with. You are only going to make your recordings noisier by turning up the preamps.

Calibrate your studio and everything will fall into place. The 192IO is set from the factory to -18dBfs (on the PT meter) = +4dBu = 1.228 Volts = 0 VU (on an Analog VU meter).

This means, the old analog technique of "keep the needle right around 0 on the VU meter" translates to "Keep the signal right around -18dB on the Protools meter (which is just under halfway up the meter)".

Some of the best tracking engineers I've seen, record everything with the faders set to "0" and change the mic preamp gain to place things proper ly within the mix WHILE TRACKING. Most people today don't really do this because pulling the fader down in PTHD doesn't really effect the sound of the track (until you get down around -90dB on the fader). Whereas on an analog console as soon as you start pulling the fader down you are changing the sound (since the fader is a voltage controlled amplifier/variable resistor).

So, to answer the original poster's question... if you are going to record and mix completely in the box you are better off keeping the levels lower for better Signal to noise ratio and to keep intersample peaks from clipping plugins and such.

THIS IS IT! THE BEST ADVICE ON RECORDING SO FAR!
It really makes a difference in quality

[Emcha_audio]

Originally Posted by O.G. Killa

Basically you are making your end product noiser by recording louder.

This is false sorry to bust your bubble. And I will give you an example. coming from the analogue days I have also seen the same result in digital.

If you record a lead of violin playing soft while a chime is also playing at the same time, the chime will be heard well. And as your where actually recording the violin you will wind up wanting to boost the gain of the violin so it pierces what ever else are on the other tracks, so you will boost the gain, and also probably raise the fader. At that same moment, you are actually also raising the chime. Now imagine that you are recording the violin and the unfortunate chime is actual RF signal that comes through from a bad shield connection or from the psu, well the same happens. It's also very apparent when you use compression.

On the other hand, if the violin is playing very loud as you record it, the chime will be less heard. And as it is very loud then you won't need to boost the gain very much or even raise the fader, you may in fact need to lower the fader so that it's not too over the head of the other tracks. And by doing so you are actually also bringing down the chime, as you lower the fader in db thus bringing down in db the chimes. Now again if the chimes were for this example RF interference, the fact that you recorded the violin hot will overshadow the RF interference that you have, and unless you compress the signal, then the Rf interference will never be brought up in a way that will be hear-able over the violin or the other tracks for that matter.

It's a technique that has been used for decades and decades in the analogue world and is also very applicable in the digital world. An other example of actual application of this is by the act that when you mic a drum, you always mic as close as possible of the head of the snare, HH, toms, kick to minimize leakage of the other parts of the drum. Because of the simple fact that the more you track hot, without peaking over the 0dbu of the digital signal, the more you have the primary sound that is being recorded. And unless your "noise" is so high that it is having exactly the same DB as the instrument, the instrument will always drown the "noise" pushing it in the lower parts of the bits depth Raising the fader, will raise the instrument even higher, yes the "noise" will be so, but in the same proportion as the instrument, and thus still be drowned by it. Compression excepting, your "noise" will never be noticeable as there will be a point where you wont be able to bring the level of your track higher without peaking.

On the other hand, the lower you track, the more you can bring up the level of your track, and as the "noise" (that will then be higher in the bit depth) will be in proportion higher than if tracking hot, the more you raise the fader, the more you will bring in front the "noise" before you track actually peaks. Introducing more garbage in your mix, than you would have if tracking hot.

It's the reason why good engineers always try to have the highest pure signal and hot (in db) while doing the track, to make sure that any "noise" is so low, that even if they end up raising the gain of the signal afterwards, the "noise" will still be in-audible. And as most time when you record hot, you actually bring down the gain by lowering the faders, and not raising the gain, then you actually diminish even more the "noise" thus having less garbage in your mix.

Let me give you a more practicle visual aid of this.

Usually (unless you have a [bleep][bleep][bleep][bleep]ty system) "noise" will be in the range of -50 and lower dbu.

If you record a Violin again (yes I love violins) at -4 dbu, then your "noise" is still at -50 dbu. Which is louder? easy right? the violin ofcrouse. Now if you raise the violin to -2dbu what happened? You only raised the "noise" to -48 dbu.

ON THE OTHER HAND..... tadadaaaam

If you record your violin at -20 dbu ... then your "noise" is still at -50 dbu. The violin is still louder but the differences between sound to noise will be significantly lower, and you will probably want to raise your violin higher, and if you raise it to -2dbu, then you also raise the "noise" to -32 dbu. That is a very, very large mark up for the "noise". Which will be made even worse as you use compression to lessen the dynamic range of your violin so you can boost it more because you tracked it to low.

Now I hope this little imagery clears things up a bit more why it's better to track hot.

Not to mention that by tracking hot, when you actually use reverbs and all, that have trailing ends in the lower bit depth, when you dither, or compress your things, you will also have less "noise" in the reverb due to having tracked hot, because the "noise" will be low in the bit depth itself.

[Picture Start]

You are talking about leakage and not recording system noise, in the case of digital: quantization distortion in the LSBs and the actual noise floor of your analog signal paths, all measurable and specified. Your whole parable about chimes and violins is valid... But for another discussion. This is about recording levels with respect to the myth that you need to engage all 24-bits in the system in order to get good sound as well as how hot the signals hitting the pro tools mix bus should be. the answer to both is: not as hot as some people think.

[Greg Malcangi]

Emcha_audio: What OG Killa stated was absolutely correct in regards to digital and I'm afraid it is your bubble which needs bursting.

Your posting is a very good explanation of SNR (Signal to Noise Ratio) and I'm sure everyone here would readily agree the need to maintain as good a SNR as possible. The discussion here is not about the need to maintain a good SNR but about recording levels using a 24bit digital system.

A common misconception with digital recording is that to achieve the best SNR you have to record as hot as possible (without hitting 0dBFS). This is a misconception because of the very nature of 24bit digital. 24bit has an incredibly huge dynamic range, the noise floor, for all practical recording purposes, is non-existent, it's hundreds of times lower than the noise in your mic-pre and recording environment. For this reason, you don't have to be concerned with recording as hot as possible because you are not fighting the noise floor of the recording medium with digital as we once were with tape. Therefore, most (though not all) 24bit systems are set so that -18dBFS = +4dBu (line level for professional analogue gear). So, by recording say to -6dBFS or higher, the "recording as hot as you can" brigade are actually recording probably 5 or more times louder than the normal operating level for which your mic-pre was designed! You see the problem here?

Let's use your example: You record a violin, let's say averaging roughly -18dBFS and the noise floor of the recording environment and analogue circuitry, let's be generous and say is at -70dBFS (your digital noise floor is down at roughly -138dBFS). Now, let's do another take, and whack up the mic-pre so the violin averages roughly -6dBFS. OK, to make take one sound as loud as take two we are going to need to add 12dB of gain bringing our noise floor up 12dB in the process. However, by turning the mic-pre up so it output 12dB more violin it also boosted the environmental noise floor by 12dB. There are two differences in these two takes: 1. Take one has 12dB more digital noise than take two, it is now up around -126dBFS or so (and therefore still insignificant and irrelevant). 2. You had to drive your mic-pre 6 or so times harder than with take one and therefore six times harder than it was probably designed for. In all likelihood your mic-pre is not going to sound as good.

Driving your analogue circuitry (pre-amp, etc.) several times harder than it was designed for is going to have a massively larger impact on sound quality than whatever happens down at the digital noise floor of a 24bit digital system. So while good knowledgeable engineers would try to get the hottest signal practical in the analogue world, in a 24bit digital world, the good knowledgeable engineers aim for round about -18dBFS.

G