MixPre-D, setting limiters

[Bruce Watson]

Managed to snag a used MixPre-D. Got it feeding my camera directly (no recorder, yet). Used the tone generator to set camera to -12 dBfs when MixPre-D is showing 0 dBu. This in turn drops the noisy hiss from the mic preamps in the camera way down, just like it should. So far, so good.

Now I'm trying to figure out the limiters. If I set them to +6 dBu, that shows up in the camera as -6 dBfs. So that any sudden loud sounds from talent (sudden laugh, clapping hands, setting down a hard object on a hard surface, that sort of thing) should "shoulder off" and not clip. Yes?

But, am I setting the limiters artificially low? I could, for example, set the limiters to +12 dBu, which would show up as +0 dBfs at the camera -- the point where the camera audio starts hard clipping. But that seems too high. Is it?

BTW, the application is dialog, but not narrative. Things like interviews, cooking shows, product demos, etc. Sometimes using lavs on talent (wired and Senn G3), sometimes using boomed (wired) AT 4053b hypercardioid (especially interviews).

So, please help a MixPre-D newbie out here -- what's the "proper" level for limiters when running from MixPre-D straight to camera?

And tell me a little more (the manual is, um, less than forthcoming) about what the limiter actually does. I'm thinking it's basically applying a curve to the incoming signal. Sort of like the way an equalizer does with frequency, but in this case, with the signal power (across all frequencies). Yes? If so, how is this roll off characterized? IOW, how fast is it? Is there a way to set it up so that no matter how much power I feed in, it doesn't exceed, say, +12 dBu output? So that clipping in the camera is impossible? And if that's possible, is it desirable?

 

[Rick R]

FWIW, I set the output limiters about 4dB below clipping . Typically around +16dB on a 302.. the mixer/recorder aligned w/ a full scale reference level.

 

[Bruce Watson]

Anybody want to explain limiter theory? Or give me a URL to an already existing explanation? I'll I'm coming up with on my own is more questions.

 

[Josh Bass]

Why set 0 to -12? Generally folks set it to -20.

 

[Steve House]

Exactly - s SINEWAVE tone at -18 or -20 will result in voice waveforms hovering around -12

 

[Noiz2]

You have to think the numbers differently with the SoundDevices products because they use a different system, I'm assuming this hasn't changed from the MixPre. Their meters are in dBu and your recorder is dBfs (probably) and the -20 people are talking about is dBvu.

So talking numbers is going to end up in confusion, especially since nobody was putting the scale reference suffix to their numbers. We are also crossing over here with some scales being analog and some being digital, the first can go above 0 the latter can't.


I'm jealous, the MixPre just has mini pots to adjust the limiter with no idea of where it is set.

I pretty much leave the maxed out and trim back my recorder so that blowing out the input gives me about -3 (dBfs) on the recorder. This is my "clip proof" setting. Then I do all my level adjusting with the MixPre. I also mark the trim on the recorder so I can get back there is for some super quiet sound I turn up the recorder. Remember I'm mostly doing FX so some are REALLY loud and others are really quiet.

I have gotten to where I can set the recorder off the tone so I don't have to over load the mic to check the limiter point.

In your case I would probably set the camera as low as it will go and set the limiter to stay -3 - -6 below clipping.

 

[Noiz2]

Anybody want to explain limiter theory? Or give me a URL to an already existing explanation? I'll I'm coming up with on my own is more questions.

The SD limiters are anti clipping devices. Plugin limiters etc are used for a variety of purposes and most info on limiters will be about them.

SD's limiters look at both the input signal and the output signal and if either crosses over the threshold they lower the input gain. This works really well as a clipping limiter and doesn't produce many of the side effects that limiters are used for in say music production.
You can think of SD's limiters as a really nice safety net against clipping. If you set them right they will almost never get used but will save your bacon when the actress decides to add a scream she didn't rehearse.

 

[Mike Paul]

Setting the Limiter threshold to +6db might be a little low, but it will depend on what your nominal levels are while mixing. If you are rolling along with your dialog levels averaging around the 0db mark on the MixPreD, then you only have 6db of headroom before the limiters activate. Since you only want the limiter to activate when unexpected peaks occur (shouts, claps, Method Actors, etc.), then setting that threshold a little higher will give you more headroom before they kick in. If you raise the threshold another 6db (to give you 12db of headroom before limiting) and your camera is clipping or close to it, then back off the camera level a little bit to compensate. Since the MixPreD is the first stage of your signal path, make sure you set your gain stages to optimal levels here before it gets to the camera.

The limiters on the MixPreD have the following specs:
Threshold: Adjustable from +6 to +18
Ratio: 20:1
Attack: 1mS
Release: 500mS

What this means is that once the threshold is reached, the mixer will output 1db of audio for every 20db of input above the threshold. The 1mS (millisecond) Attack time means it activates virtually instantaneously, and the 500mS release time means that the limiting function tapers off much slower than the attack, which makes the overall effect less noticeable (no sudden peaks and valleys)

 

[Bruce Watson]

You have to think the numbers differently with the SoundDevices products because they use a different system, I'm assuming this hasn't changed from the MixPre. Their meters are in dBu and your recorder is dBfs (probably) and the -20 people are talking about is dBvu.

So talking numbers is going to end up in confusion, especially since nobody was putting the scale reference suffix to their numbers.

I gave scale reference suffixes in the original post. But this question is about setting the limiters. And about how the SD limiters work. I can't even get SD to tell me. All I've gotten from them so far is that the setting on the mixer is where the limiters start to compress. And that the compression is 20:1, whatever that means to SD. And so far SD isn't sayin'.

What I would like to know is, where is the ceiling? IOW, if I set the limters to, say, +8 dBu, where is the ceiling from there? +20 dBu? Higher? Lower? Where is it, really?

EDIT: OK, got it from Mr. Paul's explanation below. Never mind...

 

[Bruce Watson]

What this means is that once the threshold is reached, the mixer will output 1db of audio for every 20db of input above the threshold.

That's what I wanted to know. Why the heck can't SD say that in their user guide? Of course, they probably do and dyslexia boy here can't actually see it. Sigh...

So, there's no "hard ceiling", just a "firm ceiling" eh? And it sounds like it's a step function change, and not a rolloff. IOW, you go from 1:1 to 20:1 right at onset. I would have thought they would shoulder the power curve some. Hmmm... they must have found that's either not necessary or not desirable, or both.

But the point is, in fact, to not use it at all, right? If you're hitting the limiter, your levels are too high and you should back off until you aren't hitting the limiter except under exceptional circumstances. For me, like when someone sets a glass container down too hard on a granite countertop, or drops a metal stirring spoon into a glass bowl. Big transients that need to be capped, but never the talent's voice.

Hmmm.... I'm thinking then that I should set the limiter threshold to around 3 dBu before clipping at the camera. That'll make it very difficult to clip at the camera, and by the time I reach clipping it's probably going to sound like crap already. Probably will have already overloaded the mic, or the wireless (when using). Hmmm... probably don't want either of those to ever overload.

(sound of scratcing of head)... hmmm.... Getting good sound at the extremes isn't going to be simple is it? Better add some more testing time into the schedule...

 

[Bruce Watson]

I pretty much leave the maxed out and trim back my recorder so that blowing out the input gives me about -3 (dBfs) on the recorder.
.
.
.
In your case I would probably set the camera as low as it will go and set the limiter to stay -3 - -6 below clipping.

Thanks for pointing me in that direction. I get the logic of this.

 

[Josh Bass]

Im not sure i understand everything youre asking/saying in those last couple posts. Ratios regarding this stuff expressed as x:1 always mean what was stated above. Are you asking how long does it take for the limiter to fully kick in once triggered? (I.e. going from off to the full 20:1 amount if limiting)? If so the answer would be 1 ms. Yes, its a very quick attack, but it is not literally instananeous. If you could see it slowed down you WOULD see a curve (i think this is what youre referring to) where the instant the limiter is triggered, no limiting occurs, and then over the duration of the following millisecond, the gain reduction is gradually applied until it reaches the full 20:1 of reduction. Again very quick, yes, but not literally instantaneous.

Yes, no ceiling for the limiter; the ratio determines where the final volume on the limited audio ends up. Even with it limiting, it could still overmodulate if the input signal was too high (if people are talking normally and then someone detonates a bomb next to the mic.). 20:1 is pretty high ratio, but they do go higher in other applications. generally with dialogue if you keep the regular peaks between -20 to -12 or -10 on the cams meters/dbfs scale (NOT the mixer---that would be 0 to +8 on the mixer, the first red light should be your highest for regular peaks), that shouldnt be able to happen with limiter engaged.

 

[Bruce Watson]

Im not sure i understand everything youre asking/saying in those last couple posts. Ratios regarding this stuff expressed as x:1 always mean what was stated above. Are you asking how long does it take for the limiter to fully kick in once triggered? (I.e. going from off to the full 20:1 amount if limiting)?

Not quite. Power = energy / time. I wasn't asking about the time part, I was asking about the energy part. So, if I set the limiter to start at, say, +8 dBu, how much input do I have to supply to get an output of +9 dBu? If the limiter starts right up at 20:1, then it would be an input of +28, to get an output of +9, yes? That's a rather sudden onset, I was thinking that it might be more gradual, like an input of +13 to get an output of +9, then an input of +23 to get an output of +10, then an input of +43 to get an output of +11. So, 5:1, then 10:1, then 20:1, for example.

But what I'm hearing, I think, is that it goes from 1:1 to 20:1 in a single step. No shoulder. More binary. The limiter is either on or off. Nothing inbetween. No gradual build up.

Am I close? Or not?

 

[Josh Bass]

Sounds like youre referring to whats called a "knee" on compressors, where instead of simply being on or off, the compression gradually increases as you near the threshold.

Limiter is different, no knee. Its either on or off, thats the design (think of it like the audio hitting a brick wall--that wall is your limiter). Everything works normally til you hit your threshold (wherever you set it), then it kicks in at 20:1. Thats why you set it so high, so when it kicks in, you really need it (also why ratio is so high, so it really hammers those "overs" back down) What youre talking about sounds like a compressor with a knee.

 

[Noiz2]

Interesting. The MixPre's (the D's predecessor) limiters had different specs.
"Thresholds independently adjustable from +6 dBu to +18 dBu, 10:1 limiting ratio, 5 mS attack time, 100 mS release time"

On the practical level the MixPre feels like a brick wall limiter, so the D should if anything be a harder brick ;~)

So while 20:1 (or in my case 10:1) doesn't sound that drastic remember that 6dB is twice as loud and in your case you would have to be 60dB over the threshold to have your out put go up by three dB. In practice you can set the limiter by making a loud enough sound to kick in the limiter and set it so you won't clip your recorder, and treat it like a brick wall.


What I was saying about numbers were in reference to folks talking about what number to set things to. They were referring to different scales and so that info while correct was not of a lot of use. Your camera is probably using a digital FS metering and the MixPre uses dBu, the numbers don't mean the same thing. In your first post you did give suffixes but talked as if they related in some fixed way, and they don't. Or rather they will as long as you don't change any settings. For instance 0dBfs on the recorder is you clip point, that point on the MixPre is around +28 dBu so the offset (in theory) is +28. You had the limiters set at +6dBu and said that related to -6dBfs at the camera. That relationship (12 dB offset) however was because of your gain settings on the camera. Unless you glue down the cameras gain control you can't depend on the numbers equating in the same way next time.

The traditional system was to calibrate your "zero" to a known level. This gave you enough head room and kept your signal enough above the noise floor. Tape saturates gradually and so your head room was a soft ceiling. It also gave a reference so people would know what your "zero" was so they could set theirs the same for playback.

Digital is a different beast. There is nothing beyond "zero" so the ceiling is rock hard but there is no "saturation" either so you can go right up to the ceiling with out problems. Actually -3 is the safe limit because of overshoot in the D/A conversion. And you have essentially no tape noise so your noise floor (with a decent recorder) is VERY low. Your file will always play back the way you recorded it because everyones "zero" is the same in digital.

If your average dialog is sitting around -12dBfs and peaking around -6dBfs you have a good amount of headroom. If you set your limiters to kick in at around -3dBfs (recorder side) you will be essentially clip proof.

On the MixPre side your metering is analog and it's reading dBu. It clips at around +28 dBu. You can run the MixPre quite hot.

As I said I have my limiters at the maximum which for the MixPre is +18. That leaves plenty of headroom on the MixPre side and you then just need to send it something loud enough to make the limiters kick in and adjust the recorder in so that with the limiters working you hit -3dBfs.

Once you do that, then send the recorder tone and note where that sits. From then on you can set off the tone. I generally also double check with a loud sound if I'm doing something like dialog where an over can't be redone easily.
On my recorder I set the tone to be -20dBfs and all is goodness.

 

[Noiz2]

...

But what I'm hearing, I think, is that it goes from 1:1 to 20:1 in a single step. No shoulder. More binary. The limiter is either on or off. Nothing inbetween. No gradual build up.

Am I close? Or not?

As Josh said this is what is called a "knee". I have seen knee settings. You don't want that with a limiter used like this. If you use one more like a compressor on say bass tracks then a knee can be a good thing. But when you are using a limiter to stop from clipping you want a hard ceiling you can depend on. You are getting a tiny knee in the 1ms it takes to go from off to fully on, that is where your power curve is (all 48 sample of it).