Gain Staging

The waveform on the track display gives you a good enough idea of input level.   If its flatline with a few bumps then you recorded to quiet, but anything that actually looks like a decent waveform will drive the FX bin just fine.

Chappel,  changing the meter to pre-fader is not the same as having the meter showing the channel input, pre-FX bin.

Think of the waveform drawing as a seismographic printout.

It IS a meter that shows the time component.

If your waveform peaks are at -9dB then you know you can apply up to +9dBFS trim and you'll be ok at the fxbin.

It's perfectly obvious but I think it's ok to repeatably speak about the idea because I think some folks overlook all the info represented by the waveform drawing.

I agree that each EFX should have an input and output meter... it's 2011 already. :-)


best regards,
mike

Good example Chappel, and exactly what I was talking about.

The area you've circled on the left can cause specific problems if/when the input meter on the left isn't bouncing around happily in the middle of it's "action range" but appears to "pegged"  close to the top. This is when I think it would be prudent to trim the input down a bit.

Meters are pretty much useless in this case, as, contrary to popular belief, there is no universal "correct level" standard for plugins.

Some don't care at all and will produce 100% identical results at -200dB or +200dB.

Others don't really want to (or in a very few cases can't) go above 0dB.

Others may offer gradual saturation as you approach and pass 0dB.

A plugin's "sweet spot" is wherever the programmers decided to put it. And though some of them might try to mimic some sort of "analog levels" when designing response, there is no set standard for where that should be in a plugin. And even if there was, some designers might want to deliberately emphasize distortion characteristics while others might want to give you a bit more headroom.

IOW, you need to use your ears - it's the only way.

Danny Danzi


 I've seen some plug meters clip yet the sound didn't sound like it was clipping to me at all, so I never pay any attention to that stuff. 

One of the points is that because the plugins are in floating point digital, it means they can go hundreds dB above 0dB without clipping themselves, and that the only way they respond to level differences at all is the way in which they were deliberately programmed to do (IOW, if a plugin itself clips, it means it was designed to do so). And since you don't generally know what the programmer's intentions were, the meters can't tell you anything particularly useful.

I suppose if you played with the input level for a particular plugin and watched the meters, you might note what level that plugin liked best. But I think it's extremely unlikely the ideal level will be very consistent across different plugins from different authors.

To be clearer: sonar does not clip.  Lots of conversation elsewhere about this. 32 bit floating calculations will not clip during our use for audio.

Coming in from hardware (drivers) and going out, are clipping points. This is where 0db is the top of the room. But the sonar meters at 0db do not always reference to 0db. My hardware is 24 bit, non-floating, it can actually clip ("tops" of "wave forms" are flattened).

But, not all plugs use 32 bit processing so they can be driven over the top.

The track/bus meters are useful for identifing what plug may be unhappy.

Sweet spot?  As written, this is a programmers' decision. I think is it related to the lowest signal to be recognized as above the "noise floor." And what the effect does. Compressors are not going to be useful if the signal is too low to get to the top.

And so on.

J

Danny Danzi

drewfx1

Danny Danzi


I've seen some plug meters clip yet the sound didn't sound like it was clipping to me at all, so I never pay any attention to that stuff. 

they can go hundreds dB above 0dB without clipping themselves, and that the only way they respond to level differences at all is the way in which they were deliberately programmed to do (IOW, if a plugin itself clips, it means it was designed to do so). And since you don't generally know what the programmer's intentions were, the meters can't tell you anything particularly useful. 

Now that's interesting Drew, thanks for sharing it. I don't get too much into the scientific stuff or how things work or process, I just use my ears and adjust until it sounds right. It's nice to know though, that when the plugs clip, it's nothing to be alarmed about. Like I say though, I never watch that sort of thing, but have always wondered why a plug may show a hot meter yet no artifacts on the track either in real time or once the track is exported. Good to know, thanks. :)

The short answer is Sonar (and most modern DAW's) have absolutely tons of headroom internally. So you can only technically clip before you enter Sonar (at your analog to digital converter), or when you leave Sonar (either through your audio interface or when writing a 16bit or 24bit file). Or if a plugin is deliberately designed to clip.

So even though some meters might be in the red, as long as you keep your Master below 0dB, things can sound good - because technically you are never actually clipping.

But my advice is always, "Unless you understand exactly where you are and are not allowed to go over 0dB, you probably shouldn't". 

This has become more interesting than I thought!

Thanks to all who chimed in - it's certainly cleared a hell of a lot of fog away.

Use your ears... My mixes continue to improve, I listen to what people say? And I use my tools? And I use my ears.

I was able to clip the SONAR engine within seconds. Two tracks of audio... and the level faders were all I needed.

How? Inquiring minds want to know.

IF anyones interested I just did a quick test.

Load a sinewave in TTS-1, played a chord, adding one note at a time, building to 7 notes.

On the audio track receiving the TTS audio, set input trim to max (+18db), set fader to max, the first note alone triggers the overload lights...

Send the track to a bus, set bus input trim to max, set bus fader to max, bus meter is permanently pinned on max.

Audio at this point in an analogue system would be so distorted you'd be getting almost pure square wave output...

Send the bus to "Master" bus, use Master fader to keep level below Full Scale (about -3dbfs) so the exported 24bit .wav is not clipped, export Master bus to .wav


Examining the .wav you see each added note increases the level as expected but there is no distortion. The track and bus meters were permanently pinned on max the whole time but the sine waves are not affected.

It would be interesting (although useless) to know just how hard the internal bus can be pushed before distortion (or a numeric overflow) sets in.

Karyn

It would be interesting (although useless) to know just how hard the internal bus can be pushed before distortion (or a numeric overflow) sets in. 

By my calculations, 32bit single-precision floating point should clip at about +771dB.

And 64bit double-precision floating point should clip at about +6165dB.  

To put this in perspective:
If we sum 2 tracks that peak at 0dB, we get a maximum possible peak (the 2 tracks' 0dB peaks happening at the same instant) of ~+6dB. 
4 tracks would be ~+12dB. 
8 tracks would be ~+18dB.  
10 tracks would be +20dB
100 tracks would be +40dB
1000 tacks would be +60dB

To get to +771dB, you would need to sum more than 340,282,346,638,529,000,000,000,000,000,000,000,000 tracks, with each one of them peaking at 0dB at the same instant.

Let's not even talk about 64bit.

mike_mccue



"To get to +771dB, you would need to sum more than 340,282,346,638,529,000,000,000,000,000,000,000,000 tracks, with each one of them peaking at 0dB at the same instant."

But you only need a single instance of Sonitus EQ being clipped to make it all sound awful.

:-)

Sonitus EQ is 32bit fp, and it won't clip either...

If your clipping at all Mike, you ain't clipping where you think you're clipping.

It still amuses me why people are talking about being up so close to 0db FS and clipping. You can do a complete production and be well clear of any clipping anywhere. On tracks, plugins, busses, final mixes and in mastering (close rather than well clear!).  It is possible to end up with a loud punchy master and not clip anywhere on the way there. (Keep your monitor levels up, it helps keep signals down within your DAW)

As far as gain staging is concerned having some sort of reference is always going to be your friend and of course K system levels will help in this regard too. Firstly if you have your VU's monitoring even your main stereo buss you can easily get the correct levels recorded onto tracks. If you are working even at K -14 and your average levels are around -14 db FS then you still have a clear 14 db of headroom up your sleeve. I still say it is nice to have a calibrated VU (either real or virutal eg BlueCat) showing you what is going on while you are setting levels for tracking stages.

By opening up individual tracks in an editing program and doing some level sweetening operations you will also ensure you are on the money level wise. Its nice to tame a few peaks and raise the overall rms track levels if you can. Helps in the mixing stages. You are not going off trying to trim up weak tracks for example.

From there it is into and out of plugins and yes it is easy to forget the input and output meters on many plugins. They generally won't show rms levels but will be peak more than likely but that is still good as it will tell if a plugin is being clipped on the way in or out at least.

The BlueCat meter plugin is good because it does show both rms and peak levels, can be set for K system and draws very little CPU as well. They are very handy to just insert before or after any plugins or any point in your signal chain if you are concerned about levels anywhere. I think accurate level monitoring is actually quite important and better than your ears for this sort of thing. Your ears will only tell you if things are very bad and smashed out of whack level wise. By the time you start hearing it, a lot of damage can already be done. And while you might not hear it with everything going it might show it self up somewhere else in the music.

I have found that the Mastering Multiband compressor in Sonar is one plugin that you do have to pay attention to in terms of its input levels at least. Sometimes I find the signal is just falling 3 or 4 db short all the time and it is very easy to just trim that up a tad. It will change how all the thresholds obviously are going to respond. So do it first!

If you do use a compressor on a track and you are consistently getting say - 3db of gain reduction, then you need to add that amount of gain back into the makeup gain to get the signal back to where it was. That is what makeup gain is for.  The makeup gain control is also good for stealing an extra db or two here and there. An equaliser is another place gain can be lost especially if you are cutting areas rather than boosting. If you dip out a big (or even smaller) section of the spectrum you will need to trim the output level back up again to restore signal levels. Sometimes by a lot, but it is then you really start to hear how that equed signal really sounds once it is back to normal level wise. And it is usually nice!