What levels should I be shooting for?
 

I just finished up a doc for the film festival circuit and thanks to some wise advice on here I brought all my levels down to a max of -6dbs with general level between -12 and -18dbs (previously the max was 0), which brings me to another question. When I'm recording people for docs in the future should I be aiming for those same levels on my sd702T? Sometimes I use a schoeps cmc641 and sometimes a countryman lav on a g2senn. Is it all the same? Also, for recording ambient noise is there any general rule of thumb for recording the ocean or someone walking in the woods? I imagine alot of it is by feel and subjective but I'd love a hint or two from the pros around here.

When recording at 24-bits, line-up at -20dbfs. At 16-bits, -18 or -12 is typical. Remember your goal is to maximize signal-to-noise while protecting your recording against any overmodulation.

The space above your reference point (for example, -20dbfs=20 db headroom) is to allow for lots of dynamic range with little danger of hitting 0. -18dbfs works too - so long as you're quick on the mixer and paying attention, but -20 is now the standard.

One of the great benefits of using a high-quality audio recorder (like the SD702s) is recording at 24-bits to capture the best possible field sound. Be sure to understand your post-process, though, as some of the older (and not so older) NLEs have trouble with 24-bit files - or resample upon ingest.

Re: ambient -- observe the same standard, but since there are fewer transients, you can often crank the gain into the -teens with no danger of overmodulation.

Eric, thanks very much for replying. I use 24bit when recording and have pp2 for an NLE. So in HD projects I think it converts to 16bit, but I could be wrong. But one quick question. When I line-up at -20dbfs, I'm not exactly sure what that means. Does that mean average voice level or something else?

It's a reference point to help you match your levels on all equipment - from production through post-production - and help make sure you don't overmodulate and lose recordings.

Typically, you'll generate a 1 KHz tone at the reference level (on the 702T there's a menu option for where the tone goes -- outputs only or outputs and the CF/drive) and record it for 30 seconds or 1 minute at the beginning of the day. This will give anyone else that handles the recording a reference point from which to align their audio gear.

This was much more critical when most production used analog equipment, but it's still important with digital gear. If someone else is going to use your recording - make sure you note the reference level, sample and bit rate, etc.

If you're working single-system and sending a mic or line level to a camera, you'll use this reference tone to align the audio record level on the camera with your mixer -- so you can work off the mixer without having to constantly see the camera's meters.

In theory, you can use all of the headroom above the reference level for your recordings, but it's critical you never hit zero on a digital recorder - so aim your peaks 8 to 10 db below zero, with the average level around your line-up level (either -20, -18, or -12, depending on the situation). No matter how fast you are on a fader - it's virtually impossible to prevent loud noises, plosives, etc. from hitting zero if you don't allow enough headroom.

Also, if your projects allow for real audio-post away from the NLE, preserve the original 24-bit files so someone with ProTools or a legit DAW can sub-in the higher-rez audio. This process can be somewhat automated if you have a time-code match between video and audio.

Levels, compression, and finishing for delivery on DVD or tape for broadcast is an entirely different can of worms -- but your product will sound best in the end if you follow standards throughout the process and try and always record the best audio possible.

-20 is your reference level for tone. This level is to basically to line up equipment with the same level, just a reference. Dialog could be say at -12, where -20 comes in handy is the level of HEADROOM you have for the occasional yell or shout, with 24bit it wouldnt clip as fast as 16bit as you have a higher dynamic range. It shouldnt clip anyways if the limiter is on, but dont just depend on the limiter, but try to maximize your headroom without going to 0 all the time, but their might be the occasional time that you do.

-20 was originally made up by hollywood to more or less transfer 35mm to digital with little hassle. By referencing to -20, they can fit a potential 20 DB of headroom. DVD'S can handle that range, VHS can not. The cinema definetly can. In broadcasting situations, the reference is at -12 usually.

If a source is referenced at -20 and will be broadcast, the level of the program is usually turned down a little in post and a limiter is applied for a max of -10 DB

Not to knit pick, but the -20dbfs standard has nothing to do with DVDs, VHS, or Hollywood. Well, maybe Hollywood.

It's a broadcast standard set by S.M.P.T.E. - the Society of Motion Picture and Television Engineers. It's officially known as RP155 which specifies -20db DBFS (decibels full-scale) for test tone to be recognized as 0VU (analog level) in the U.S.

The standard is designed to help maximize the potential headroom of an audio signal, minimize analog transmission loss, and maximize signal-to-noise. Interestingly, the U.S. SMPTE standard offers a bit more punch than its -18db DBFS European equivalent.

This all has to do with audio signals driving a certain amount of electrical power -- stuff that's not really worth delving into.

This standard is important for the production chain - from acquisition through post. Master levels for commercial DVD distribution, as I mentioned in the earlier post, are an entirely different animal. Same goes for final mixes delivered to networks for broadcast programming. The model number "LM-100" will give any TV mixer the chills...

The real benefit of 24-bit recording is the staggering increase in audio resolution over 16-bit recordings, especially in faint, transient, and low-level sounds (not in the loudest, clearest sounds captured adequately by 16-bit equipment).

** 24-bit recordings will clip just as easily as 16-bit recordings! ** However, the increased resolution of 24-bit allows you to set lower record levels (while actually increasing recording resolution) thereby decreasing the likelihood of clipping/overmodulation - via increasing the headroom. (Does this make sense anymore!?)

Another major advantage with 24-bit is that with that extra resolution, it's easy to boost (normalize) recorded levels in post production with virtually no noise, so even record levels considered 'low' by 16-bit standards will still produce outstanding results.

Whew. That's it for today. I hope this helps.

Nicely put. ;)

Sorry - posted in wrong thread (Mods please remove!)
 

Reposted in correct thread now!

Just to adda note to the discussions of -20 versus -12 line-up levels. The type of signal can make a difference in what the meter reads. If you adjust your NLE levels so its peak-reading meters are sitting at -20dBFS with a sine tone and the switch to a pink noise signal recorded at -20dBFS RMS without resetting levels, the NLE meters will read about -12dBFS. In other words, a sine wave tone and a broadband signal consisting of complex mix of frequencies that mimics real-world signals set to exactly the same RMS average power level will meter about 12dB apart. That's why tone at -20 to set system gain staging and then ride levels so speech stays around -12. And tone at -20 and speech hovering around -12 digital will both read about 0VU on trasitional analog VU meter.

About the -20 reference level made up by hollywood to more or less match the headroom of 35mm movies transferred to DVD's with little hassle. This info is from Tomlinson Holmans "Sound for digital video".

24 bit doesnt just give you more headroom but also more bits. This is critical when applying effects in post as you have more bits to use. 24bit also helps when mixing tracks for less noise. There's a reason that 24bit/48khz is generally the standard for double system. 24BIT audio is kinda like 10bit video vs 8bit video, you have a higher color range with 10bit video, you have a higher dynamic range with 24bit audio.

BTW 24bit is realistically about 18-20bits, but will help a lot in processing.

well the one thing I have seen is that when mixing for _ANALOG_ gear like betaSP, using a mix level of -12db=0dvu works best for how that analog gear was designed. if when working in FCP, I'll mix at a normal -18 , peaks at -12db, as speced by ALL the local TV stations for digital delivery.
when going to betaSP I then boost 8db for a ref level of -12 and layout to betaSP. I also flatten the mix by using a real analog compressor at 1.5 or 2:1 to flatten the peaks. yes a flatter more pumped mix, but thats analog. actually 1800's stink because they indicate clip at just over +3dbu while 75's will take +6dbu before indicating clip. problem is, too many who don't know this so if you have even a flicker of the clip light on a 1800... well they get all silly when in fact you are still within being ok. sometimes... analog is a gentler kinder place even if the tradeoff is some S/N ratio and freq response.

Over here in PPM land I think we got -18 because we always used to set tone at 0 dm and peaked at +8 db. Therefore ten db was stuck on top of that for headroom. (to allow for the time it takes for the PPM needles to rise etc..

0 db and +8db where magic numbers so it would have gone against the grain to choose a decimal division of ten such as -20.

We used to line up tone at 0db ppm4 to -4 vu. I wonder if people do that here with beta SP stil. I'l ask VT today and see if they give me a funny look.

As practically all of us semi-pros here use digital video only, and digital PC based editing systems, would it not be practical to refer only to dBFS digital level scales??? I know the analog history, but it is still confusing for beginners.

And what levels to aim for? When shooting aim as high as you safely can (use limiters) to minimize hiss and system noise, even when shooting something quiet, pushing levels down never downgrade the signal. When editing bring the levels to delivery standards specified by TV stations, can not go much wrong. That is: dialog around -20 to -12 dBFS, max peaks at -6 dBFS.

ok, lets make things simple

In today's world, we're more concerned with shooting digital video whether its SD or HD.

First scenario-using a mixer and camera

Send tone out of the mixer and balance the levels on the camera to -20, record say 30 seconds of tone before program start. Once the levels are set on the camera, we control dialogue levels on the mixer and usually will look for peaks on the camera around -12, with the occasional yell going above that

Second scenario- using a mixer and recorder

same thing, send tone out of the mixer and adjust channels on the recorder to -20, record 30 seconds of tone, levels are adjusted on the mixer once the levels are set on recorder

Third scenario- just a seperate recorder

Most recorders do have internal tone, which you record -20 internally before program start, you may use one mic only and dont need the mixer

The point, which ever method you choose, you can use the -20 reference to line it up with another piece of gear. Say you are now transferring your sound files onto a deck, or computer or a DAW, you will use -20 level to adjust other
equipment to match that level, we're talking digital here.

Say now, you are lining up your -20 digital level to an analog equipment, you would play your -20 tone recordings and set your analog gear to 0,
basically -20 digital equals 0DB analog.

I hope this helps

I'm assuming you mean to set the mixer tone level to -20 first, right? (SD302 default is 0)

Or am I mixing myself up somehow?

SD302 1kHz test tone is at -20 dBFS (0 dB VU) level as defaul, just like it should be.

It is also possible to get a full level test tone out ot the SD302 via the menus, which is at 0 dBFS (+20 dB VU). This can be used to test the system match, that is, set the level excatly so that levels are maximized but recorder clipping is impossible.

Unless your in the U.K. in which case -18 = 0db analogue (ppm4) and we always aim to peak at +8 db analogue which is -10dbfs. 8 db above tone.

The U.K. method makes more sense I think actually as it refers directly back to the old practice and is a logical progression.

Oh while I'm at it here in the U.K. broadcast world we have pretty much avoided V.U.s completely and use PPM's which actually makes it a lot simpler.

Does this mean that broadcast cams (varicam, digibeta, etc) record 24 bit audio, since their meters indicate that peaks should be at -20? (this is opposed to cams like the XL2, DVX, etc. where the peaks want to be at -12) .

They're actually the same. (Huh? 0 = -20???? --- YES!) You're getting digital and analog mixed up. The meters on the Sound Devices 302 are analog meters, selectable for either VU or Peak reading, that are calibrated so that 0dBVU on the meter equals an output signal voltage of 0dBu, 0.775 volts. When that signal is input to the DV camera or a digital recorder, it should line up to -20dBFS on the recording device's digital reading meters. So 0dB on the SD302 IS -20dB on the recorder or camera. Same signal level but different meter calibrations.

Steve, Petri - thans for the comebacks.

Let me try to see if I've got it.

The 302 manual says that tone level can be set to -20dBu, -10, etc and that 0dBu is the default

Their recommended calibration procedure is to generate 0dBFS and set this to just below clipping on the recorder.

I think this is where I was getting confused between the 0dBFS and 0dBu - I was mentally not paying attention to the distinction between dBu and dBFS when I saw the comment in the manual about 0dBu being the default and I was reading it as 0dBFS.

Anyhow, I calibrated my recorder per the instructions to use the 0dBFS tone and set it at just under clipping on the recorder, and when I sent the default 0dBu tone from the mixer, the recorder indeed read -20dBFS as it should.

I think the reason that they recommend the "just under clipping at 0dBFS" method is that some recorders don't have exactly the same scale as the 302, and even if 0dBU on the mixer reads -20dBFS on the recorder (which would be an easy way to set things up) you might still get clipping on the recorder before the 302's limiters kicked in.

Have I got it?

Jim, you got it. Another way to set the recording levels for maximum dynamic range is to send a full maximum level tone from the 302 and set it to just below full-scale (0dBFS) on the recorder. That will be about +20dBU actual signal level, so when you send a tone that reads 0dBVU on the 302 meters, it will read about -20dBFS on the recorder meters and the 302's limiters will kick in just before the recorder clips.

No his explanation was confusing. 0dB tone from the mixer to -20 at the camera inputs.

Mixer output and camera input both must be set to line or mic, but not one to mic and the other to line.

Regards,

Ty Ford

It is confusing and I usually am a pretty good writer/explainer. How's this....

1: Turning on tone on the SD302 (or 442) that reads 0VU on the mixer's meters means the mixer is putting out a 0.775v signal level on its outputs.

2: According to the standard, a camera, recorder, or workstation that's recording digitally should have its recording level controls adjusted so that a tone with a level of 0.775v at its inputs will result in its meter reading -20dBFS.

3: Real world signals ain't simple tones and meters react differently to tones and complex signals. Setting tone that averages 0.775v to read -20dBFS on the recorder and then switching to voice that also averages 0.775v without changing any settings will actually result in the recorder's meter reading closer to -12dBFS with the same setup.

Your 3: is still confusing to me Steve. I know what you're trying to say, but if I put on my novice brain, I don't understand.

Lets see......Tone doesn't have the peaks a human voice or piece of music does. When you align tone at -20 on the camera, you should also check your metering on the mixer to make sure it is set to a Peak reading mode.

I ran into this issue with a friend recently. He had most of the calibration right but was concerned that the readings were so low. He wasn't setting the mixer to read peak. After I showed him how to do that, he saw the correlation between the mixer metering and the camera metering and the concept "clicked" in his head.

The other most frequently missed problem is that you have to have both mixer output and camera input set to mic or to line. THEY HAVE TO BE SET THE SAME ON EACH DEVICE. Mic to mic or line to line.

Regards,

Ty Ford

Peak reading mode
 

How is this done? Is it possible to explain this here in writing? If it's different for each mixer, then let's use an SD 442 for example.

-Chris

With the SD442 there's a meter mode pushbutton control on the front panel. Each push switches to the next of 4 possible modes and then after the 4th one, back to the first to repeat the cycle all over again. It starts out in VU meter mode, then a button push switches it to Peak Reading, then to a combination VU/Peak mode where the VU is a continuous series of LEDs illuminated ending at the VU level followed by a sequence of several dark and then a single LED showing the Peak, then finally to a Peak Hold which shows the same reading as the Peak meter but holding the indication for a few seconds before decaying to give you time to actually see the sucker. You can easily see on voice the difference in the VU and Peak on the same signal by setting the mode to VU and speaking into a microphone while adjusting the fader until the meters read 0VU and then pushing the button to switch to Peak mode while continuing to speak taking care to keep your voice the same. The meter immediately jumps up to read + 6 to +8 dB higher and it actually would be higher than that but the real peaks are occuring too fast for your eyes to see the LEDs illuminate. Or send it some music while in VU/Peak mode and see how far above the bar indicating VU the peak single LED will land. Another interesting demonstration is to use a tone plug to send tone to an input, use the fader to set it to 0VU level, and then switch the metering between the modes. Unlike voice that reads signifigantly higher in Peak mode than it does in VU mode at a given perceived identical volume level, the steady sine wave reads exactly the same in both modes.


(Have a new 442 and I've been running those experiments myself just to get gut-level familiar with how the meters respond in different situations.)

The trick with recording is that even a single cycle transient, far too short for any meter to respond to, that goes above 0dBFS (a level represented digitally by 1111111111111111 if recording 16bit) will clip with potentially audible distortion. So you want to set the average signal levels low enough that under no circumstances will the waveform EVER rise that far yet high enough that the highest peaks just almost reach it so as to realize the best signal to noise ratio while preserving the full dynamic range of the original performance.

The SD 302 has the same button pushing method to switch among the metering modes.

thanks steve, much more simple than I thought.