Final mix going over, but meters ok

[cretaceous]

So I play my mix in Reaper and the master VU is all ok - but when I render it briefly goes over by 0.2
Then I use the gfx analyser and get a better indication of level but I can't pinpoint that over loud bit (I know meters are algorithms and some might be faster and more precise).

I could see the peak in the render dialog window, but had no way to see exactly where it was.
I resorted to doing a screen grab of the render dialog window and laying it over a screen grab of the mix window in Photoshop and scaling to match, so I could see the exact point where the 'over' occurred and then finally adjust levels.

Previously I just stuck a brick wall limiter on for a tiny over like this, but there must be a better way to do this, surely?

Advice please!
Thanks

[Fex]

Originally Posted by cretaceous

there must be a better way to do this, surely?

Turn everything down! There's no reason to aim for 0dB.

[cretaceous]

Yes but everything is at -1 or less - I just have one tiny hump in the snake.

[planetnine]

Originally Posted by Fex

Turn everything down! There's no reason to aim for 0dB.

Originally Posted by cretaceous

Yes but everything is at -1 or less - I just have one tiny hump in the snake.

...as in 15dB further down

Oh, and Fex isn't entirely correct, you obviously need to turn your monitors back up



>

[cretaceous]

Thanks for the suggestions guys and I appreciate the place you're coming from, but they don't address my question.
I was hoping there might be a persistent level tracker - much like the persistent spectrum analyser - so you get to see a record of the levels as you play the mix.
I would like to see the peaks as well to help me balance - my left ear has worse hearing than my right ear.
Maybe there's a VST.

[Fex]

Originally Posted by cretaceous

Thanks for the suggestions guys and I appreciate the place you're coming from, but they don't address my question.

They do. You asked if there was a better way.

Are you mixing, or mastering?

Do you want your music to look nice, or to sound good?

If you're mixing, for your ears, you are, seriously, way too loud.

Originally Posted by cretaceous

I was hoping there might be a persistent level tracker so you get to see a record of the levels as you play the mix.

Render the tracks to wav. The wave form is a persistent level tracker so you get to see a record of the levels as you play the mix.

What might really help you, if the meters aren't catching clipping on playback, is more sensitive meters as a VST on the master track.

[cretaceous]

Theoretically it's mastering, but it's only for release to my fellow bandmates (or the occasional person who might find it on the internet).

Why can't I happily go to just under "0" ?

As regards the wav render - thanks, way better than my Photoshop workaround, but still seems a bit of a kludge.

[planetnine]

It's good practice to leave plenty of headroom when recording/setting up a mix (along with other prep duties), so that you don't have to worry about peaking while actually mixing, which ideally needs to be a quick-ish process (even spontaneous), so you ears don't get too used to the sound.

That said, REAPER uses 64-bit floating-point format maths, so you can't actually clip (as in distort) the channels and most plugins. Don't overdrive the master though, as fixed-point renders and your audio interface will clip (and distort).

You can always normalise your mix afterwards (mix decisions should not be made on whether you're close to 0dB or not), and of course you can compress and limit to 0dB or close if a particular intrinsic loudness is your game.

You shouldn't worry about this at mix time though.


>

[ashcat_lt]

Guys? Good advice I suppose to not shoot for final volume in the mix stage, but it's not the issue here. Sometimes, for whatever reason, you actually do want to get it up to final level. If you can't count on the meters to tell you what the final file's levels are going to be then you're screwed.

OP - Are you by chance resampling in the render process? A lot of times if you are hitting a limiter to keep your peaks in check, and then put a lowpass filter after the limiter, you'll end up seeing peaks higher than limiter would have allowed through. A large part of downsampling involves such lowpass filters, and it could easily have the same effect.

[Lawrence]

Originally Posted by cretaceous

Yes but everything is at -1 or less - I just have one tiny hump in the snake.

Lot of unnecessary worry over 0.03 db when you can just pull the master down 0.03db.

At any rate, the 0.03db difference is most likely due to inter-sample peaks and Reaper may not have any way to meter those. Most meters can only read levels at the sample points, they cannot calculate what may be a waveform curve a little higher in-between that during reconstruction in the analog world.

Go get the free SSL ISM meter and stop stressing over this kind of minutia.

And maybe make a feature request for Reaper to add "true peak" metering.

P.S. If you read up on digital recording this is exactly why you shouldn't be at -1, inter-sample peaks.

[clepsydrae]

Originally Posted by ashcat_lt

Guys? Good advice I suppose to not shoot for final volume in the mix stage, but it's not the issue here. Sometimes, for whatever reason, you actually do want to get it up to final level. If you can't count on the meters to tell you what the final file's levels are going to be then you're screwed.

Yeah, if OP is talking about final mixdown, you don't want to turn overall levels down 15db and then normalize a 16bit fixed point file afterwards...

In general rendering should not clip if the master meter didn't clip while playing, and understanding why this happened is important. I'm not familiar with ashcat_it's theory, so maybe that's an explanation. Perhaps there are plugins that can cause this as well? Perhaps a plugin with a random component to it might clip some times but not other times?

OP, what plugins are you using?

Have you tried an online render instead of an offline to see if it prevents the clip?

What is the FX chain you're using on the master bus?

What are the project settings (sample rate, etc) and to what format are you rendering? (screenshot of the render dialog might be easiest).

Digression: Something that caught me out once: if the master fader is at +0.2 (accidentally or because you were using it as a shortcut to make the monitors louder and it didn't get back exactly to 0dBFS) and then you enable the volume automation envelope for the master track (which i do for fade out, etc), Reaper will make an flat envelope curve at 0.2dBFS and also sets the fader to 0 so it can now be used as an overall trim. It's a great feature, but you look at the envelope and it looks like it's at zero, you look at the fader and it's at zero, and you look at your plugin chain and it's limiting and not going over 0, yet the master still clips due to the 0.2 boost from the curve. In this scenario, though, the master fader will show a clip, and it sounds like yours is not.

[clepsydrae]

Originally Posted by Lawrence

At any rate, the 0.03db difference is most likely due to inter-sample peaks and Reaper may not have any way to meter those.

I'm a little confused about this point; if the master meter didn't show the ISP, would the meter in the render dialog? Even if the OP is rendering to a different sample rate or whatever, would the render meter somehow detect ISP's while the master meter wouldn't?

In other words, aren't ISP's only relevant when doing DA conversion, or are they also sometimes relevant when changing sample rate or bit depth?

OP, you could try using an ISP limiter last in the master buss and seeing if it works... I use the free Limiter No6: http://vladgsound.wordpress.com/plugins/limiter6/
...you can deactivate all sections but the last one, set to "ISP precise" and the ceiling to whatever.

Anyway, the academic question of why the master meter didn't clip but the render meter did is still interesting to me...

[whiteaxxxe]

ISPs are not shown in "normal" meters. as far as I know - VladG explained it once, but I didnt get it in whole, because he is a nerd, in a very positive way, a genius I think, so when he talks techincalities ... oh, well ... - its due to that ISPs as the name says must be measured in a way that has to do with partly guessing. these ISPs are no samples, so every method that measures the values of a sample must fail. interesting with these ISPs is, that they are under normal circumstances impossible in digital, because they represent a value bigger than 1, whereas in digital everything is only 0 and 1. so, more than a string of some 1s is impossible. theoretically.

practically meters measure sample values. ISPs are no samples. they come to exist when a wave rises relatively fast to its peak at 0db, is followed by another peak at 0db and after that the waveform falls down again relatively fast.

so if the waveform is "generated, the curve spills over the 0db between the 0db samples. that is the explanation for that.

it has nothing to do with the Reaper meters are not good or not capable. you must use the limiter no6 to kill ISPs, I dont know of any other method. and to measure these, there is a ISP meter that comes with limiter no6, that gives you a clue, whats going on.

on the other hand: you are mixing way to hot if you run into ISPs. even in mastering this shouldnt occur. although nowadays ... I think Skrillex invented ISPs. or Metallica. or Nigel Tufnel?? anyway, ISPs are the new "11" on the master volume ... :-))

[clepsydrae]

Originally Posted by whiteaxxxe

ISPs are not shown in "normal" meters.

Yeah, what I'm wondering about is whether they can become relevant when resampling in some esoteric way. It seems unlikely, and I can't come up with a plausible scenario, which is why I question Lawrence's theory.

The crux of the OP's question, at least to my nerd brain, is why the master meter didn't clip but the render did, assuming we have that correct, and assuming there isn't some user error going on.

[Lawrence]

I assume because the master meter can't measure ISP's and that really, really minor clipping was maybe ISP clipping but there's no logical reason for him to be up at -1 anyway.

Digital meters can do that, calculate or measure reconstruction or measure what they call true peaks, some products just choose not to. I wouldn't exactly call the math that does that "guessing".

The most important part of it all is that nobody even cares and nobody can really hear 0.03 changes anyway, so just pull the master down a little. It's just unnecessary minutia that has no real practical value to anyone just making decent music. Don't obsess over stuff like that, it has no real value... relatively speaking.

If you must obsess over it, put this on your master and go back to work... http://www.solid-state-logic.com/music/X-ISM/

The funniest thing about all of it is that most people can't even hear 0.03 of digital clipping anyway, especially not in a track with 4 guitars shredding on the latest amp-sim.

My other theory is that he printed an mp3.

[karbomusic]

I'm a little confused about this point; if the master meter didn't show the ISP, would the meter in the render dialog?

IIRC whenever I render something near the limit, I see overs in the render dialog that never existed in the mix environment and just assumed Reaper was guesticalcing ISPs. I'd have to test but that's what my failing memory is telling me. I'm almost certain the render metering differs from the master meter 'depending' on circumstances. Would it not be plausible to assume the rendering does this pseudo ISP calculation since it is no longer on the fly?

I'm like Lawrence though, it was a clear enough sign to go back to the mixer and drop it down some, move on, live life, prosper.

[clepsydrae]

Originally Posted by karbomusic

IIRC whenever I render something near the limit, I see overs in the render dialog that never existed in the mix environment and just assumed Reaper was guesticalcing ISPs. I'd have to test but that's what my failing memory is telling me. I'm almost certain the render metering differs from the master meter 'depending' on circumstances. Would it not be plausible to assume the rendering does this pseudo ISP calculation since it is no longer on the fly?

Interesting. Now that you say that I seem to recall seeing a "clip light" on the render meter with a peak of 0.0 and thinking to myself "maybe the render meter works a little differently"... my memory also fails me.

Assuming the render meter is working off the rendered output, say, 16bit fixed, it stands to reason that there would be occasional differences in what clips when (e.g. a 64bit float value that is technically a tiny bit over 0dBFS but which quantizes down to 0dBFS in 16bit fixed, or possibly the reverse happening from the addition of dither to the 16bit signal). But it seems like it would be very rare?

If the render meters are working in a significantly different fashion (e.g. ISP detection or otherwise) it'd be interesting to know how.

I'm like Lawrence though, it was a clear enough sign to go back to the mixer and drop it down some, move on, live life, prosper.

Yeah, well, different strokes. :-) I like to understand anomalies when I see them to make sure that I understand my tools. Half the time I track stuff like that down I find out new things that are useful in practical ways. The other half of the time it's just interesting for its own sake.

[karbomusic]

Originally Posted by clepsydrae

Yeah, well, different strokes. :-) I like to understand anomalies when I see them to make sure that I understand my tools. Half the time I track stuff like that down I find out new things that are useful in practical ways. The other half of the time it's just interesting for its own sake.

That's the creative recording me, the geek me would sit there for hours with a calculator and spec sheets etc. I just try to separate the two these days and never wear both hats at the same time. As to actually why it peaks, it's total assumption on my part simply because it "smelled" like it was tracking ISPs. I have no data to support that.

[Lawrence]

Originally Posted by clepsydrae

I like to understand anomalies when I see them to make sure that I understand my tools. Half the time I track stuff like that down I find out new things that are useful in practical ways. The other half of the time it's just interesting for its own sake.

To be completely fair to you, my reaction was partly a nagging ongoing impression of minutia chasing in general so I do apologize if I unfairly targeted your question. In general, when something doesn't seem to add up audio engineers do often really want to know "why", so the question is in indeed a valid technical question.

Sorry if i came off like a jerk. I so often see people chasing pointless minutia (contextually speaking, in a sound quality way, thinking it's important) that I do tend to forget that sometimes it's just simple curiosity.

"Why does what looks like 1+1 seem to add up to 2.03 (or whatever)" is a valid question.

[clepsydrae]

Thanks -- and no sweat. :-)

[Mr. Data]

I experienced this only when I rendered to a lossy format like mp3 (eg. when converting). In such cases the level can go "over", because the result is not an exact duplication of what's happening in the DAW when just playing back.



-Data

[Lawrence]

Here's S1's ISP switching on it's level meter plugin while reading a test tone at -0.01. It always goes a little over so it's good to not push the levels that far since there's no good reason to anyway.

That reddish tint on the tips of the meters kinda means... "Uh, you probably shouldn't be up here anyway." or "Danger Will Robinson!".



No clue if that one is as accurate as the SSL meter (I doubt it, the SSL meter is a cpu hog) but it serves the purpose if you really must push the scale that hard.

Otoh, you'd probably need to be in a really, really great space with really really great speakers and ears to even hear it clipping by 0.05 anyway so I doubt it will impact 99% of the population.

P.S. While it certainly doesn't hurt anything, I've never really understood the need for render meters. In most cases you'd think that you'd already know everything you need to know from the master meter, or the meter on whatever bus you're rendering.

[cretaceous]

Wow - I triggered a whole discussion - thanks to every one who contributed. Really great informative comments from you all!

Right well, um, yes, I was rendering to mp3 - this was just a quick render to stick on my iPod to play in the car for "car audio" test, so I skipped the wav render. Will try not to do that again, honest.

On this one I had no fx on the master - individual tracks had eq, minimal compression and a couple had a bit of reverb, one had reafir and another had a noise gate.. but I try not to use too many fx.

I need to read an authoritative but simple guide to compression mastering, but there's so much out there it's hard to know which are good for middling levels of experience.
I've read enough to know I should avoid compression on the master, as I don't know enough to be sure I'm improving things.

Thanks again

[clepsydrae]

Originally Posted by cretaceous

I was rendering to mp3 - this was just a quick render to stick on my iPod to play in the car for "car audio" test, so I skipped the wav render. Will try not to do that again, honest.

No shame in that -- render to whatever suits your needs. But it is a good guess as to what happened; as others have pointed out, the encoding of mp3's entails changes to the audio that can sometimes result in higher levels, and when you're pushing up against 0 it might trigger an over. I've heard that leaving 0.3dBFS headroom is a good rule of thumb if you're heading to mp3s, but don't quote me.

Of course this theory presumes that the render meter is showing the post-mp3-encoding stream... Seems logical that it would, but I'm still curious to confirm that, and to know of any other technical differences between it and the master meter, whether ISPs or whatever else...