Why record and mix with such hot levels?

[blairl]

I read an observation made by Paul Frindle about the Pro Tools mixer. I think anyone that works with Pro Tools should take a look.

Level Practices

What do you think?

[tedness]

It may be wrong but the artists like it louder, the A&R guys like it louder, the promo guys like it louder, the PDs like it louder- Everybody except the mastering engineer.

[Shawn Simpson]

But the A&R guy's desire to hear it loud has nothing to do with the absolutely out of control levels people are recording. If you want your mix waveform to look like a couple of 2x4's, that can be done without recording every element on the verge of clipping.

[gwailoh]

It's the mastering engineer's job, though, to make it loud. It's very possible to record cleanly at -6dbr per Paul Frindle's recommendation and still produce a loud master. (This doesn't imply that I know whether PF's recommendation is the right thing to do or not, just that it's not necessary to slam track levels to end up with a loud product.)

[spigots]

Try this. Record a vocal, a set of drums, a single electric guitar (distorted), and an electric bass. Record them all to use up as many bits as possible without hitting 0 dBFS. Then play all of those tracks at unity gain. Do you like the way it sounds?

[lancejorton]

as i understand it, the point is you record it hot to get the maximum bits and that give you more dynamic range, a lower noise floor, and increased sonic accuracy. you don't have to max it out... i just try to keep the peaks in the top 6 db. if you turn down the level, PT plugins are pre-fader and therefore processed as accurate as possible. the PT mixer is 48-bit and therefore keeps a low noise floor even at lower levels. if you're going to an external summing bus, you just want to keep the hottest instrument withing the top 6 db. again, this maximizes your bits and therefore lowers the noise floor during D/A. recording at low levels will increase your noise level and make your mix weaker. this isn't just a loudness race, it makes the audio better. getting the EQ close and doing an initial stage of compression in the analog domain helps with this process.

ps - same thing with writing the mix. if the level is low, it will increase your noise level and make the mix weaker.

--
lance

[PTUser NYC]

This is silly.

Noise mean random uncorrelated energy. There is no way to hear noise in the recorder that is any lower than the noise in the signal being recorded.

A 24 bit word offers a theoretical signal to noise ratio of 144 db, although there are no common audio circuits which perform that well, including in our converters, so a 24 bit converter usually has a noise floor around 132 db below full scale.

If you record without any peaks higher than -18dbFS (18 db down from full scale on your converter) you will still have a distance of 114 db from the "low" peak to the noise floor.

Nothing you have ever recorded had 114 s/n ratio. The sound of moving air on a mic, the microphone electronics, and the preamp, if nothing else, will not deliver those kind of specs. Therefore, the noise in your signal is above the noise in the converter. "Getting all the bits" is a myth. If you raise the signal you are recording to get peaks higher on the meter, you are also raising the noise in the signal. The S/N of the source signal remains the same no matter where you record it on the converter, until you record it SO low, that the noise of the signal is lost in the noise of the converter.

Meanwhile, you're doing terrible things to your audio to get these higher levels. What does a mic preamp have to do to output levels above 0dbVU? When your Neve sounded great driving a Studer a few db into the red is now peaking at -16dbFS on your meter, but you want to go hotter "to get all the bits". You push the mic preamp electronics into new levels of distortion.

You also give your plug ins no headroom. With peaks at -6dbFS, how do you boost 12kHz +7db? You can't.

If you want a loud master (and you should because all of this gets turned upside down when you go to 16 bits and there is only a 96 db S/N available) then go ahead and get a good sounding mix, and then raise the volume of the output of a a compressor or something across the mix bus. Waiting until the last stage to add the gain keeps everything in Pro Tools sounding sweet. You'll be amazed at how much the "harsh digital sound of Pro Tools" is really people using it way too hot - distortion from mic pres, and clipping in plug ins.

[frenchman]

the never ever ending debate...

[PTUser NYC]

Yes, except the debate was already won and ended, and some folks just don't know it.

It is a fact that if you record a signal which has an 80 db S/N ratio, the noise in the signal is 80 db lower than the peak.

It is also a fact that as long as the noise in your recorder (or converter) is lower in level than the noise in the signal, and as long as no peaks in the signal exceed the recorder's (or converter's) headroom, then the signal has been preserved with full detail.

With a decent 24 bit converter, the S/N ratio of any signal you have to record will be seriously more constratined than the recorder's ability to record it.

Figure a 90 db S/N ratio of a signal, a rare enough occurrence. If I record on a 24 bit converter which has a real world S/N spec of around 130 db, I can record with peaks only as high as -40dbFS on the meter and still record all the detail that is possible in the signal.

Simply put, it matters not at all how much lower the noise in your converter is, as long as its still below the noise in the signal being recorded.

Recording hotter than that does not increase detail in the recording.

What recording hotter does do, is eat up headroom for your plug ins, and ask your analog electronics to work in ranges much hotter than they were originally deisgned for.

Imagine a Neve 1073 mic pre. It is designed to sound best (purposeful overdriven color aside here) when operating around a level that would result in an output around 0dbVU. This was where it operated when interfaced with a Studer tape machine 20 years ago.

In those days, 0dbVU wasn't the end of the line. You could record in the red, above 0, and get tape compression effects etc. Boards were set up to give quite a bit of headroom over 0dbVU (sometimes more than 24 db!) to accomodate transients and hot signals.

Then along came digital. With every instantaneous voltage slice being recorded as a numerical value, the designers of the system have to identify a point where the highest number possible (full scale, the end of headroom) would be located electrically. They knew that gear and recordings made on analog equipment routinely included levels above 0dbVU, so they decided on a plan to have a level that corresponds to 0dbVU on a VU meter show up on a digital system at -18dbFS, or 18 db below full scale.

This meant that digital equipment could accomodate up to 18db of headroom over 0dbVU.

When you are operating your Neve "correctly" it should be showing up with peaks around -15dbFS or so, 3 db or so into the headroom of the device, and average levels near -18dbFS on your converter.

When you record hotter to get phantom resolution, you not only get no extra resolution, AND eat up plug in headroom, but you ALSO get more distortion from your analog gear, which is now running hotter than intended to accomodate higher levels in this quixotic quest for untenable resolution.

When you look at a 16 bit file, the numbers tighten up. Suddenly the noise floor is only 96 db below Full Scale, and mixes that peak at -12dbFS are only going to have 84db of resolution if recorded at those levels, plus hit records are loud, right?

This is why you work in 24 bits, with reasonable levels, get a great sounding mix (you'll be shocked at how much air and detail was lost in the analog chain when recording hotter) and then when you're about to print it, you note the highest peak of the mix. Lets say it peaks at -10dbFS.

I'll use the Rennaisance Compressor across the Master Fader just for the output gain. Even if the compressor set to 1:1 ratio, and a full scale threshold (so it never acts) the output gain is active. I boost the output gain so that there is 0.1db left of headroom, andplay the mix through to check that I have it correctly.

Then print the mix - full resolution was observed during tracking and mixing, and then captured faithfully as best as possible into a 16 bit ready level. You gained resolution and headroom in the analog gear, and preserved 10db or so more of S/N in the final 16 bit master at the end.

So what debate? Where is a correct view that opposes this?

Saying that you like the sound of pushed gear doesn't count. I like it too, I do it all the time, when I'm doing it on purpose. But until you choose to push something for color, you should check to see where your converter is set to interface with 0dbVU (mine is at -14dbFS) and try to keep levels around there.

[Mt.Everest]

as i understand it, the point is you record it hot to get the maximum bits and that give you more dynamic range, a lower noise floor, and increased sonic accuracy. you don't have to max it out... i just try to keep the peaks in the top 6 db. if you turn down the level, PT plugins are pre-fader and therefore processed as accurate as possible. the PT mixer is 48-bit and therefore keeps a low noise floor even at lower levels. if you're going to an external summing bus, you just want to keep the hottest instrument withing the top 6 db. again, this maximizes your bits and therefore lowers the noise floor during D/A. recording at low levels will increase your noise level and make your mix weaker. this isn't just a loudness race, it makes the audio better. getting the EQ close and doing an initial stage of compression in the analog domain helps with this process.

ps - same thing with writing the mix. if the level is low, it will increase your noise level and make the mix weaker.

--
lance

Lance,

make your mix weaker? Increased sonic accuracy? maximize your bits?? no.

please read PT User NYC's post, re-read, take a nap, then read again. Then do a search for "record levels in PT" here on the DUC. You really have your facts wayyy backwards. Im really surprised at how many ppl STILL have these myths imprinted in their minds. I think Nika's Digital Audio book should be bundled with every DAW purchase and have some kind of eye scanner that wont let you use your DAW until you have read it cover to cover. I too am sick of this debate, and sick of getting PT sessions from other engineers with Kicks, Snares, and Guitars hovering at -3dBfs. I think that PT should do away with its meters so people stop worrying about this bits nonsense and focus on their front end.
and Lance, what do you think is happening to your audio and analog front end when you have peaks at -6dBfs? You do realize that -6dBfs is +12VU right?

somehow tho, I think this debate will never end...
MT