Recording Levels...but hopefully a new spin on it...

[zzfdvz@mac.com]

So i have read about one thousand times over that -18dbs at 24bit is where it's at as far as average level.

I recently picked up mixing with your mind, which says digital audio gets more accurate and detailed the higher you go as long as it doesn't clip.

Assuming that the book is correct...why not record hot?

If it's an issue of gain staging, a trim plugin as the first insert on each channel would be just fine, take up almost no CPU, and allow recording the hottest most accurate things possible.

I've also read that a lot of gear wants to work at -18dbs...which I can totally buy if it's older...but is it really the case with modern equipment, loudness wars, etc?

I'm not saying either side is correct...I'm just looking for very fact ladden arguments each way and why. At this point, with all of the information (and misinformation for that matter) floating around the web...simply saying "-18dbs because I said so" is kinda lame. I also acknowledge there may be no diffinitive answer...but well guided arguments would allow people to make their own choices.

Let's hope this doesn't turn into a blood bath...

[RCM - Ronan]

If your averages are around -18dbfs, we could assume that your peaks might be around -6. What is is that you are trying capture that you do not feel is accurately represented in 23bits?

[Never1]

Originally Posted by rcm

If your averages are around -18dbfs, we could assume that your peaks might be around -6. What is is that you are trying capture that you do not feel is accurately represented in 23bits?

Yepthumbsup

You have to realize that digital headroom is absolute, but our converters and the converters on consumer electronics are not what digital might see as being right.

Distortion can be introduced earlier on cheaper converters, even if we treat 0dbfs as the "limit".

That also applies on input.

when everything is getting dithered to 16 bit, in the end, it makes no sense to push our converters to the max. you can hit 20 bits in a 24 bit system and still maintain the integrity of the audio since you're still above the 16bit "watermark". It's been pretty much proven that levels that fall between -18dbfs and -10dbfs do sound better in the end.

If you don't understand this, then you haven't spent enough time recording and mixing (including the time it takes to try and fail and try again until you figure out what sounds best) in a DAW.

The "rule" also lets most plugins work in their optimal range without having to trim your gain stages, within the DAW... thus improving the quality of your audio.

RCM knows what he's talking about.

[Duardo]

I recently picked up mixing with your mind, which says digital audio gets more accurate and detailed the higher you go as long as it doesn't clip.

Assuming that the book is correct...why not record hot?

I'd have to see exactly how this was worded in the book, but it's really not correct. The whole advantage of recording at 24-bit resolution instead of 16 is the increased dynamic range. With digital audio 0 dBFS is as loud as you can go, and as you increase the number of bits you have (up to a certain point) you are able to record quieter signals. While you do get more "detail" as you increase the bit depth, all of that detail manifests itself as better resolution of lower-level signals.

When we were recording at 16-bit resolution there was more of a reason to record as hot as possible, especially when we were recording audio with a wide dynamic range...because if we didn't the quiet parts of our audio could easily get down to the level where the noise floor was audible. Now, with 24-bit recording, we have more dynamic range to play with, so getting the hottest level possible isn't really an issue.

As far as "why not" record hot...one good reason is to avoid the chance of clipping if the signal you're recording gets a little louder than you think it's going to.

I've also read that a lot of gear wants to work at -18dbs...which I can totally buy if it's older...but is it really the case with modern equipment, loudness wars, etc?

What kind of gear are you talking about, and what is "dbs"? If you're talking about 18 dB below the digital ceiling of 0 dBFS, that's a meaningless number as far as analog preamps, compressors etc are concerned. Most analog gear is set up to work best when run around +4 dBu, where it reads 0 dB on its meters, but it will have extra headroom above that. Typically you'd want to have 0 dB on your gear show up as around -18 dBFS on your recorders, or a little more or less depending on your gear, what you're recording and how you like to work. Set it higher and you could very easily clip your converters without your preamp even breaking a sweat.

The loudness wars are not something you should even worry about at the tracking stage...and when it comes to "resolution" and "details", those are two things that the loudness wars have nothing to do with...

[joeq]

Originally Posted by zzfdvz@mac.com

If it's an issue of gain staging, a trim plugin as the first insert on each channel would be just fine, take up almost no CPU, and allow recording the hottest most accurate things possible.

Isn't using a trim plug-in reducing the amount of bits in the signal?

How is this "loss of resolution" any better for your audio?

Richard Nixon used to crank the air conditioning in the White House so that he could have a fire in the fireplace on a day when it was really too warm for a fire.

At least he got some aesthetic benefit from a system that was basically canceling itself out. If your aesthetic is that you want to hear your analog front end struggling to keep up, maybe this makes sense.

So i have read about one thousand times over that -18dbs at 24bit is where it's at as far as average level.

I recently picked up mixing with your mind, which says digital audio gets more accurate and detailed the higher you go as long as it doesn't clip.

So that 1000 for, 1 against.

Where is this "new spin" we were promised?

[Thescort]

You really need to consider the consequences of one route or the other.
1. Recording hot may lead to reaching 0dBfs = Really nasty and irreparable damage to the audio.
2. Recording at a more conservative level (e.g., -18 dB) and not risking digital overs.
- What are the consequences of this? You don't occupy all 24 bits all the time, but that is ok, that is why we have 24, not 8. If your source is a little too quiet, it is easy enough to fix.

There is no saturation in digital so don't try to push it!!!

[PDC]

Originally Posted by Never1

Yepthumbsup

You have to realize that digital headroom is absolute, but our converters and the converters on consumer electronics are not what digital might see as being right.

Distortion can be introduced earlier on cheaper converters, even if we treat 0dbfs as the "limit".

That also applies on input.

when everything is getting dithered to 16 bit, in the end, it makes no sense to push our converters to the max. you can hit 20 bits in a 24 bit system and still maintain the integrity of the audio since you're still above the 16bit "watermark". It's been pretty much proven that levels that fall between -18dbfs and -10dbfs do sound better in the end.

If you don't understand this, then you haven't spent enough time recording and mixing (including the time it takes to try and fail and try again until you figure out what sounds best) in a DAW.

The "rule" also lets most plugins work in their optimal range without having to trim your gain stages, within the DAW... thus improving the quality of your audio.

RCM knows what he's talking about.

I would not blame this on consumer gear completely. There are many audiophiles and causal listeners with better A/D and D/A than many studios have.

Another reason not record near the "last bit" is because the mix engines in most DAWs can't tolerate it. The average plug-in is not going to be able to handle it.

The best reason is that preserving the waveforms properly without distortion is our job. It is the job of the mastering engineer to square them all off making the waveform veiw of a region in PT look exactly like the block view.

[Dean Roddey]

I recently picked up mixing with your mind, which says digital audio gets more accurate and detailed the higher you go as long as it doesn't clip.

I'm not one of those folks who argues to record everything at what I would consider silly low levels (like *peaking* at -18dbFS on the converters) but this is not true.

The reason a lot of people think this (and I did, too, until I finally had it beaten into my head) is that they believe that by recording at higher peaks, then you are 'spreading out' the samples, i.e. you have more sample levels between the samples of a note at one volume and a note at a higher volume. So that must mean there's more accuracy involved.

But that's not the case. You have X number of dB of possible range and you have a grid of 16 millionish slots laid over that range. All those slots are the same size. So you are using the same *resolution* everywhere on the whole possible range (a separate thing from dynamic range.)

If you went on to process all the data in the DAW at that 24 bit integer level, then there could be some argument for the higher levels in terms of resolution. But most DAWs these days are probably using floating point or fixed point 32 bit or 64 bit summing engines. So there is a whole universe of values almost between slot X and slot X+1 on the 24 bit integral sampling scale. So it doesn't matter if there were 10,000 or 50,000 sample slots between two notes in terms of resolution within the DAW because in a floating point summing engine, there are probably millions of possible values between 24 bit sample values 10,000 and 10,001.

Since the DAW will represent the values during summing/processing as values between 0.0 and 1.0 (though it can go over 1.0 sometimes), then if sample value 10,000 was (just randomly picking a value) 0.00312 and sample value 10,001 was 0.00313, then a huge range of floating point values between those two sampled values. You have 0.003121, 0.00312089, 0.0031293 and so on. The number of actual digits that can be represented is limited by the 32 or 64 bit'ness of the summing engine. A 64 bit summing engine would have a ginormous number of possible intermediate values between those two sample values.

So when all the processing is going on between the raw samples and the final output to the master bus, the amount of resolution available for intermediate calculations is huge in a floating point DAW. The only thing that would get you more resolution in terms of the sampling processing the converters would be to add more bits to the overall range, i.e. go to 32 bit sampling.

Now you've made each sample slot considerably smaller, so each sample value can more accurately respresent the level that was captured. But most folks would probably tell you that that's a waste of time at this point because 24 bits is always way more than enough.


And all that is separate from how you push your analog gear, which is more to do with good gain staging for minimizing noise and getting the sound you want out of that gear, some of which sounds nice when pushed hard. As long as you adjust the last one in the chain to get the level you want to the converters, you can do whatever is optimal up stream from there.

[steelyfan]

Clarity is only Clarity, it's doen't even make my top 10 list of essential musical requirements. Not comparing music styles at all here, but The Police recordings don't sound better than Jimi Hendrix's, and they're recorded way cleaner with alot more clarity. I've never understood this quest for supreme clarity.

[MarkRB]

It's about line level and is based in science not opinion.

Pro gear operates at +4dBu, which means a constant signal coming out of your pre at 0VU measures 1.228volts.

Your AD also has a predefined (sometimes switchable) 0VU or line level.

AES recommends -18dBFS, so a constant signal (i.e. 1kHz tone) coming into your AD at 1.228volts will read -18dBFS in your recorder. Presuming that is how your gear is set up. Other common 0VU levels are -16 and -12 or even -20dBFS.

A VU meter is an intentionally slow meter in order to show you an average reading of level. If you hit it with a constant tone at line level it will read 0VU and your recorder will record that tone at -18.

However if you hit it with a transient source, like a close mic'd snare your VU meter will hit zero, but it's not telling you the whole story (as it is measuring an approximate average level).
You will find that your snare that peaks at 0VU will actually be peaking at around -6dBFS in your DAW due to the fast transient energy of the signal.

Ergo 0VU = -18dBFS. Plenty of headroom for both your analogue and digital gear. No spin.

[Brad McGowan]

Mark is right. It's all about maintaining proper headroom in both your analog and digital gear.

Brad