Digital Recording Levels

interesting. What they are effectively doing is recording at a lower bit depth. If you start at 24 bit you can drop a bit and still end up with 16 bits of usable resolution. Part of the "problem" with 16 bit is that a lot of time you were only getting effectively say 12 bits of resolution.

They are music guys so they were probably trying to get to like -0.1 fs ;~)

I don't peak over -3, but I tend to record hot. If I don't have limiters or a way to get a sound check I don't have a big problem going lower. But as a post person I have to say that the two biggest problems I get with production (after just horrid mic placement) is too loud (usually a lave stuffed someplace) or too low. There are way too many new recordists (or cameramen, PA's pressed into service) that think if you can crank up the headphones and hear something you are good.

Short of clipping too loud can be fixed, too low? not so much. If it's really low there just isn't enough data left to work with. You can boost it but it gets really "grainy", which is natural since it is effectively a low bit recording.

The "low" they are talking about is about where production levels used to shoot for. They are talking "low" because they are music guys who lust for knobs with eleven on the scale.

The other "problem" with gaining in post used to be math errors. There used to be LONG battling threads about only gaining in 3dB increments because the math errors caused distortion. Truthfully I never heard the fabled distortion, but computers are a very long way from where they were then and they have no problem with the math now.

I'll take production peaking at -12 all day and be a happy camper.
 
My experience in this (doing dialogue editorial for games, tv, and whatnot, as well as some production sound and sound effects recording) is the exact opposite of the conclusions in that link, ESPECIALLY when low cost gear enters the equation. I'm not trying to bang on lower end gear, but typically the cheaper something costs the noisier it is. I'm dealing with this right now, as a matter of fact.

I've always been a fan of getting a nice hotter signal rather than boosting a quiet signal. The only time I've run into distortion is an actor ad-libbing a scream or yell from something quiet. If you know what you're doing and have semi-capable gear, it's really a non-issue.

Let me put it another way:
I've had to cut dialogue on a TV show where the levels were really low. I had to boost the gain on everything and then try and clean up the noise. Don't create extra work for your dialogue editor and re-recording mixer. Record a proper level and be done with it. It just makes everyone's lives and jobs easier.
 
I'm with Dave Fisk on this. It'd be like underexposing your video and then boosting the levels in post -- same thing, you're going to end up with noisy footage or noisy audio. Why not record at a good, sufficient, proper level and save yourself endless grief in post? I'm not suggesting (as the guy in the gearslutz post said) to record "as hot as possible", because that does indeed risk inadvertent clipping. But recording sufficiently hot (dialogue at generally -12dB, maybe a little hotter) is going to deliver better results than recording at -32dB and boosting 20dB in post...
 
This works well when using pro level recording gear that has very low self noise.

With consumer level gear with much higher self noise you can't boost the level without hearing the noise-floor.
 
graywolf said:
There is an interesting thread over on gearslutz experimenting with recording at lower levels and adding gain in post. It suggest that it takes advantage of the lower noise floor in digital recording.
Click to expand...

Dialog recordists have been doing this basically since digital recording of dialog took off. That's what setting peaks to -12 dBFS is all about. That plus really good limiters makes it considerably easier than in the old days of the 1/4" Nagra recorders where you had to record relatively hot and you had to really watch meters and you were constantly trying to maximize your signal to noise ratio.

What those guys are talking about is having considerably more signal to noise ratio available than they actually need, so they can record a little lower and use some of it. Music recording typically doesn't use limiters, so the extra overhead is very nice to have. You'd have to record way lower than anyone is advocating to have a signal to noise ratio as bad as analog tape -- I'm just sayin' that no one is arguing for people to get sloppy. It's just recognizing that we don't have the pressure on us to record as hot as the analog days. And that's a good thing.
 
I concur with Alex and Dave on this was well. Recording at low levels just doesn't make sense. You have to make up the gain somewhere else downstream later on, and this adds even more noise from a low bit depth and the downstream gain amps as well as the original noise from the s*%t preamps. (It just seems lower cause the initial listening level is reduced). Theoretically, for optimum performance, the level is -0.001 dBFS. But of course in the real world, 'headroom' plays into this heavily.
 
Last edited:
Hum..?

I am certainly new to digital recording, cassette tape guy from way back, so my Zoom F4 and my Sony Camcorders are about all my experience in this. But I seem to have good luck aiming for -12db peaks with the limiter set to -6db (soft). That seems to put conversational level down around -24db. With my cassette decks I shot for 0dbVU peaks which gave me a max of about 10db headroom (I considered 7db safe).

Actually, I am going for about the same with the digital if you convert back and forth from dbVU to dbFS. Of course the reel to reel guys had more space about like 16 bit vs 24 bit, I guess. I was conservative, just like I now have to be with video, because my takes were usually one chance. The main difference is where that noise floor is with the cassettes it was -30 to -40 max, with digit it is at least -60 and quite often quite a lot more.

I do realize that they were talking about studio music recording where traditionally they shot for the bleeding edge with big tape machines, so the idea is kind of revolutionary over there.

But, I really am interested in your opinions, which is why I posted the link here. Only, I am now not sure we are talking from the same viewpoint or not. Which is why I added this post with numbers in it.
 
I used to record the graywolf describes. Now I record hot. Almost everything I do goes to the web. In post, I always found myself adding about 6dB of gain and compressing and limiting if I was recording at regular, safe levels. When dealing with web content, I figure you are fighting music and computer sound effects so it better be as loud as the music and background noise. On one gig, the producer ended up complaining that my levels were too low and he couldn't hear it very well when he listened to the recording. In order to save the client, I spent the night remixing and mastering the sound. Since then, I record relatively hot and let the volume hit the hard limiter every once in awhile. I figure that I might as well do a little of my own compression before it hits post and nobody has ever complained that it is easy to hear the audio. If the editor wants lower sound levels, they can reduce it in post. Or compress it.

In any case, I am dealing with equipment with good limiters and preamps. I could see a possible argument could be made for recording at lower levels if you have really bad equipment. Such as if your gain staging is all messed up or you have some type of digital gain being used or there is some noticeable preamp hiss being created by a junk recorder. But I would characterize those as operator mistakes or not using the correct equipment for the job.
 
When recording digital at 24-bits, you should set the 0VU level at -18dBFS (or -20dBFS) and peak, say, 10dB above this.

This gives plenty of headroom and allows for inter-sample peaks.

If you record at 16-bits, then the 0VU level would be -12dBFS and peak only 6dB above this.

This is the way I have been working since I first started recording digitally in 1982.
 
cpreston said:
I used to record the graywolf describes. Now I record hot.
Click to expand...

You should NEVER record hot with digital.

You cannot record over 0dBFS and anything higher just gets clipped.

Even levels slightly lower than 0dBFS can be clipped and distorted due to inter-sample peaks.

Record as I describe above - then - if you want it "hot" do this in the DAW so you get a hot recording you desire without any clipping or distortion.
 
When I say HOT I am peaking below -3dBfs. 3dB is the area where intersample peaks could cause clipping in the D/A converter. This is actually analog clipping not digital clipping. I don't play the headroom game because I have limiters that are set to prevent any clipping.

I'm not very dogmatic about things so whatever is working for you go for it. Recording too low IS a problem. If your peaks are hitting -12dB at least you are probably in the zone, really depends on dynamics. If your RMS is below -20dBfs or so you are starting to get into problematic territory.

I'm doing a show now where one interview has peaks around -15dBfs, but the RMS is like -34dBfs. It is a BIG PITA. when you boost the level you boost the nose. Noise reduction needs data to work with and because they recorded so low there is very little to work with so you head into artifact city really quick. So right now we are in this balancing game with the director of how much noise she can take. I may pull out my old Behringer Noise suppressor since it was great on tape his and doesn't have artifact issues...

Anyway there is no magic with digital, it's just data. The more data you collect the more you get to use. Going over (digitally) looses data so that is bad, going in too hot screws with the A/D converter and gives you errors in the data and going out over -3dBfs can give you errors in the D/A converter. Aside from those limits the more the better, data wise.

The big variable is really on the analog side. Brand X A/D converter may sound better at some point lower than max data and better always = better. And the D/A converter may not like getting hit so hard and might sound better lower? I'm not concerned with the D/A converter since I don't have any control over it, and the end user is going to get a mix anyway. My concern is getting the best data I can into a file for an editor. If I wasn't using limiters I would be recording lower so I could play the headroom game and keep from clipping.
 
Like most things in life, it seems a well balanced approach is the best.

Also, I would think that what you give the client is not the raw recording.

That thread I posted about have move on a bit. Seems to be 3 opinions: 1-Record hot (Music studio hot) and hope you didn't over do it. 2-Record at a sensible level and boost it in post. 3-Everyone on this thread is nuts.

Those who try 2, are still claiming that they get the cleanest sound they ever have.

As usual, most of the arguement comes from every ones differing definition, rather than actual practices.
 
To clarify;

When I say record "hot", I'm talking about recording as hot as whatever system I'm working on will allow at that given time. It's also important that talent actually be as loud as the scenario will allow (this gets into the whole mumbling debate and 'realism'). When I do my own videos (product demos and educational stuff), I speak loudly and I record hot. My peaks are right below clipping, but I'm know what I'm doing and I know what I'm going to say and how I'm going to say it and I know where the limits are.

Like I said, if you know what you are doing and know what your gear is capable of, record as hot as you can, whatever level that ends up being. Try to avoid having your post people boost your levels later on. It creates unnecessary work for them, which could create animosity, especially in episodic TV and reality TV. I've had to cut dialogue on reality TV, where I got picture late (because people in editorial couldn't make up their minds), and I had very little time to do my job. Getting dialogue to a usable level in post is not something I wanted to deal with because that meant I wasn't cutting. Leaving that to the re-recording mixer...just getting a proper level...is not something they want to deal with either when it's something that should have been done to begin with in the TV world. Maybe if you have loads of time it's not a biggie, but for me, it's creating a step in the process that doesn't need to be there.
 
Yeah, it's a real pain having to compensate for low audio levels, especially when they're not consistent, so peak normalizing won't work. I have even resorted to auto-gain plug-ins for rush jobs, unlike hardware based AGC, the good ones offer parameter adjustments.. still not a desirable solution though.
 
Last edited:
It's also worth noting here that lower-cost recorders from the likes of Tascam, Fostex, and Zoom tend to meter just a touch higher than the actual recorded signal. If you're shooting for -12dB, you might want to hit -10dB or even -8dB on the recorder just to be sure.
 
davefisk said:
To clarify;

When I say record "hot", I'm talking about recording as hot as whatever system I'm working on will allow at that given time. It's also important that talent actually be as loud as the scenario will allow (this gets into the whole mumbling debate and 'realism'). When I do my own videos (product demos and educational stuff), I speak loudly and I record hot. My peaks are right below clipping, but I'm know what I'm doing and I know what I'm going to say and how I'm going to say it and I know where the limits are.

Like I said, if you know what you are doing and know what your gear is capable of, record as hot as you can, whatever level that ends up being. Try to avoid having your post people boost your levels later on. It creates unnecessary work for them, which could create animosity, especially in episodic TV and reality TV. I've had to cut dialogue on reality TV, where I got picture late (because people in editorial couldn't make up their minds), and I had very little time to do my job. Getting dialogue to a usable level in post is not something I wanted to deal with because that meant I wasn't cutting. Leaving that to the re-recording mixer...just getting a proper level...is not something they want to deal with either when it's something that should have been done to begin with in the TV world. Maybe if you have loads of time it's not a biggie, but for me, it's creating a step in the process that doesn't need to be there.
Click to expand...

I am not a sound guy and there seems to be varying opinions on this post. A lot of good and interesting info, some even confusing. But I agree with davefisk from my personal experience. Fear of clipping has caused me to almost always record a bit too low and then increase the gain in post. I think I am going to start recording just a touch hotter. Thanks guys
 
Rick R said:
I have even resorted to auto gain plug-ins for rush jobs, but they offer parameter adjustments, but still not a desirable solution.
Click to expand...

Same here. The promise of having them at least do a rough even it out pass is tempting but I almost always end up redoing it by hand and regretting the time wasted.

I can hear mixers puking, but on a big film stuff arrives at the mix as a roughed in premix (admittedly done by a human mixer). On no/low budget posts it would be a real help to basically be able to automate that premix. On really simple stuff it sometimes gets in the ballpark but a realtime pass riding a fader generally works better, and usually faster.
 
I am surprised no one mentioned dual level recording, either in camera or audio recorder.

I generally set the main level around -12 with peeks up to -6, then set up the second channel to record 6 dB lower.
In post, I pick the higher level without clipping.

Sometimes the talent talks at one level for the audio test, then talks louder later during the scene. This covers it well, at least for me.
 
Last edited:
Digital Frank said:
I am surprised no one mentioned dual level recording, either in camera or audio recorder.

I generally set the main level around -12 with peeks up to -6, then set up the second channel to record 6 dB lower.
In post, I pick the higher level without clipping.
Click to expand...

Two reasons I think. First this only works if your A/D converter didn't clip. If you clipped the preamp you just have a lower and higher version of the clipped sound. So providing your inputs are set at two different levels it should work, but some recorders have it as a setting which makes me think it's a digital domain setting which isn't going to help much.

The second reason is that in the article that was reference they were recording low because they thought it sounded better. Though I gues if you are setting the level in the analog domain then you could do an easy test to see if you like your recorders sound recorded at the lower level.?
 
You must log in or register to reply here.
Back
Top