Better mixes?

This makes perfect sense to me.

http://www.looptv.net/tutorials/studio/logic-pro-k-metering-mixing-tips-monitor-calibration-with-james-wiltshire/

8)

http://www.logicprohelp.com/forum/viewtopic.php?p=531226#p531226

In theory in a floating point app with no hardware this should not make a difference, but in my experience, it does.

How so?

In theory in a floating point app with no hardware this should not make a difference, but in my experience, it does.

 

How so?

The mix sounds more "open" for want of a better word.

The mix sounds more "open" for want of a better word.

You think there's an as-yet undiscovered, or undisclosed, reason for this in the software code?

Or maybe it just encourages you to make more "open" mixes because you have headroom to mix more "openly"?

Workflow or science?

The mix sounds more "open" for want of a better word.

 

You think there's an as-yet undiscovered, or undisclosed, reason for this in the software code?

 

Or maybe it just encourages you to make more "open" mixes because you have headroom to mix more "openly"?

 

Workflow or science?

Definitely workflow, but yes, I think there may be things that go on in a DAW's summing engine that our measurement methods are not up to yet. When a guy with ears like Eric Persing of Spectrasonics and Roland fame tells me that on his high end rig, he can hear differences between two files that null, which should be bit for bit identical, I start to wonder. The guy has 30 years of amazing audio achievement and so I cannot just discount it as psychoacoustics.

My friend and fearless leader here David (who probably wants to scream at this point) is a computer scientist who can tell you chapter and verse why this cannot be, and intellectually I am with him, but my experiences have led me to conclude that other stuff is going on.

So my advice is make your detente points lower. -18 is pretty low,and if you don't want to do K-Metering (I don't) then -12 is fine.

Hey, if this guy is all wrong, still, where's the harm in doing it anyway?

The things that are correct in the video:

- Analog gear will usually sound less distorted when fed a more conservative signal (e.g. peaking at -18 dBFS out of the box)

- Some plug-ins will sound different when fed a louder or above full scale signal in a floating point system (but almost all stock plug-ins in Logic Pro will not)

- Using a lower reference point than 0 dBFS while mixing will help you avoid overloads (on the master channel and busses only)*

*An important aspect that I don't think was emphasized is that the key to using moderate levels is to regulate the internal gain of all plug-ins on a channel, not just the channel output faders.

Definitely workflow, but yes, I think there may be things that go on in a DAW's summing engine that our measurement methods are not up to yet. When a guy with ears like Eric Persing of Spectrasonics and Roland fame tells me that on his high end rig, he can hear differences between two files that null, which should be bit for bit identical, I start to wonder. The guy has 30 years of amazing audio achievement and so I cannot just discount it as psychoacoustics.

This is basic maths, and as simple as 1 - 1 = 0. There can be no difference. Any perceived difference is placebo.

You can also find videos of Greg Calbi talking about cables and Bruce Swedien talking specifically about Monster cables. Both are incredibly talented, but it's complete nonsense.

Being talented at something doesn't necessarily mean you also understand the science behind or can explain it. This is also why we have audiophiles (and religion)

Question: I used to clip plug-ins all the time having no idea that it mattered. Now I am more conscious of this and take a good look to make sure the meters are at a moderate level.

But, there are plenty of instrument plug-ins I use ALL THE TIME that have no meters at all. Like uHe Zebra2 and Diva. I probably use these instruments more than anything else, save a couple of hardware pieces that get used.

How am I to know if I am clipping the plugin if there is no meter? Diva in particular is VERY LOUD when you open it. With the channel fader at unity gain, it's extremely loud every time you open an instance for the first time.

Most plug-ins are 32 bit floating point and you cannot clip their inputs even when the signal goes above 0 dBFS.

Instrument plug-ins generate an audio signal, they have no input, so by definition they cannot clip a signal (unless it's a desired effect, for example with a distortion effect that is part of the instrument).

The technique of "using moderate levels" to avoid clipping the input of plug-ins has its limits: the only way to check for sure wether or not the audio at the input of a plug-in goes above 0 dBFS is to insert a meter plug-in before that plug-in. "Using moderate levels" doesn't mean much in itself, and is not a guarantee that you're not going to clip the input of a plug-in. This workflow only works if you never raise the gain (too much) inside your plug-ins. If you're using a compressor with automatic gain on, or an EQ and you use the gain slider, or a gain plug-in, or any other plug-in that can affect the gain of your audio in a significant manner, then "Using moderate levels" isn't enough of a precaution to avoid clipping the input of a plug-in.

An example would be an instrument that is really loud, followed by a compressor that compresses a lot, without make up gain. You would see a perfectly reasonable "moderate level" on the channel strip, and yet the signal at the input of the compressor could be WAY above 0 dBFS. So if that compressor was a hardware compressor or a 24 bit fixed plug-in, then you'd be clipping your signal, even though you're using moderate levels.

But like I said, most plug-ins cannot be clipped anyway.

Note that I am not recommending pushing your levels liberally in the red. I am just stating how things work in Logic. I still do recommend you work with moderate levels, if for no other reasons than the ability to use meters and faders in their "sweet spot" areas, where they are the most flexible and efficient. What I am trying to say is that if you're concerned about clipping the input of a plug-in, then using moderate levels is only going to put luck on your side, but you still can't tell wether or not you're clipping the input of a plug-in. Inserting a meter plug-in before that plug-in and checking the level of the signal at the input of that plug-in is the only way you'll know for sure wether you're clipping the input of the next plug-in or not.

I had an interesting conversation about this with renowned Logic-based engineer, Chuck Zwicky. He told me that the possible reasons for my arriving at my conclusions are two-fold:

1. Some third party 32 bit plug-ins just don't sound good when receiving hot signal from another plug-in, even if they are not clipping. They just don't.

2. By controlling the flow from plug-in to plug-in on channel strips to keep the levels moderate with pre-fader metering, I am paying a lot of intention to the details of my mix, and that kind of focus almost always leads to better mixes.

When Chuck speaks, I listen, so I will accept that.

But like a conspiracy theorist, I am still suspicious that we do not really entirely know what goes on in the summing engines of DAWs yet

Most plug-ins are 32 bit floating point and you cannot clip their inputs even when the signal goes above 0 dBFS.

 

Instrument plug-ins generate an audio signal, they have no input, so by definition they cannot clip a signal (unless it's a desired effect, for example with a distortion effect that is part of the instrument). ..............

Thanks!

Most plug-ins are 32 bit floating point and you cannot clip their inputs even when the signal goes above 0 dBFS.

Waves plug-ins output 24 bits and you can easily and audibly overload some of them.

UAD have many plug-ins that - at least on paper - have recommended operating levels. Most are at -18 dBFS @ +22dBu, while some are different (Studer, Ampex, Manley Massive Passive). While I can't comment on the validity of UAD's own information, I'm quite sure the recommended operating levels for at least some of their emulation plug-ins are true.

One of the most widely used compressor plug-ins even today is the Waves Renaissance Compressor. This plug-in has a very soft knee and a hard-wired saturation limiter on the output. Overloading this plug-in will always lead to an uncontrollable threshold even at the minimum threshold setting (0) and can easily lead to overloading of the output saturation as well. There is no input attenuation option available, so it can only be avoided by using a conservative gain structure.

Instrument plug-ins generate an audio signal, they have no input, so by definition they cannot clip a signal (unless it's a desired effect, for example with a distortion effect that is part of the instrument).

Somewhat irrelevant since they can go above 0 dBFS on the output which can lead to overloading the next plug-in in line. So good gain structure still applies to instrument plug-ins.

The technique of "using moderate levels" to avoid clipping the input of plug-ins has its limits: the only way to check for sure wether or not the audio at the input of a plug-in goes above 0 dBFS is to insert a meter plug-in before that plug-in. "Using moderate levels" doesn't mean much in itself, and is not a guarantee that you're not going to clip the input of a plug-in. This workflow only works if you never raise the gain (too much) inside your plug-ins.

Which is why you should do the latter, i.e. not raise the gain too much inside plug-ins.

If you're using a compressor with automatic gain on, or an EQ and you use the gain slider, or a gain plug-in, or any other plug-in that can affect the gain of your audio in a significant manner, then "Using moderate levels" isn't enough of a precaution to avoid clipping the input of a plug-in.

True, but this is also why many or most plug-ins with auto gain have a compensation feature, often a linked one, e.g. Waves L-series or FabFilter Pro-L (holding Option while gaining will lower the output simultaneously).

But like I said, most plug-ins cannot be clipped anyway.

"Most stock Logic Pro plug-ins" since a lot of people will interpret your sentence as "most plug-ins in general", which is not true.

Note that I am not recommending pushing your levels liberally in the red. I am just stating how things work in Logic. I still do recommend you work with moderate levels, if for no other reasons than the ability to use meters and faders in their "sweet spot" areas, where they are the most flexible and efficient. What I am trying to say is that if you're concerned about clipping the input of a plug-in, then using moderate levels is only going to put luck on your side, but you still can't tell wether or not you're clipping the input of a plug-in. Inserting a meter plug-in before that plug-in and checking the level of the signal at the input of that plug-in is the only way you'll know for sure wether you're clipping the input of the next plug-in or not.

It's not as diffcult as that.

Watch your peak and watch your fader offset relative to unity each time you insert a new plug-in and you know. The exact same amount of luck as when using a meter or pre-fader metering.

I had an interesting conversation about this with renowned Logic-based engineer, Chuck Zwicky. He told me that the possible reasons for my arriving at my conclusions are two-fold:

 

1. Some third party 32 bit plug-ins just don't sound good when receiving hot signal from another plug-in, even if they are not clipping. They just don't.

Poppycock. Plug-in processing is maths, pure and simple. No magic or mystery, and it can be proven with a simple null test. Null tests are irrefutable evidence. Failure to understand that a null test is valid proof is a failure to understand pre-school maths. It really is as simple as 1 minus 1 equals 0. Add two tracks, but polarity invert one of them. If there is no residue then there is no difference in sound. Q.E.D.

Now, there are cases when you're dealing with real-time error correction vs. offline error-correction when a null test can fail to show the problem, i.e. null testing a ripped (offline error corrected) CD against the hard disk based source and coming up with no differences in the data, but the physical medium (with real-time error correction) actually sounds different due to a bad burn that triggers the real-time error correction in the CD player. But this is a very special condition and has nothing to do with a ITB mixing file vs. file or track vs. track situation. But I digress.

2. By controlling the flow from plug-in to plug-in on channel strips to keep the levels moderate with pre-fader metering, I am paying a lot of intention to the details of my mix, and that kind of focus almost always leads to better mixes.

This on the other hand is good advice and can lead to a more controlled sound when using some 3rd party plug-ins. I won't use the word "better" since the perceived effect of distortion is subjective.

But like a conspiracy theorist, I am still suspicious that we do not really entirely know what goes on in the summing engines of DAWs yet

As Clarke's third law states:

3. Any sufficiently advanced technology is indistinguishable from magic

Instrument plug-ins generate an audio signal, they have no input, so by definition they cannot clip a signal (unless it's a desired effect, for example with a distortion effect that is part of the instrument).

Somewhat irrelevant since they can go above 0 dBFS on the output which can lead to overloading the next plug-in in line. So good gain structure still applies to instrument plug-ins.

That's exactly what I was trying to say: the instrument plug-in cannot clip the signal it generates. It can however overload the next plug-in in the chain. Watching your channel strip meters on the Mixer may not help prevent clipping the input of a plug-in placed after such a hot instrument.

Telling a newbie: "just watch your channel strip meters and keep reasonable levels" may not prevent them from clipping the input of plug-ins, especially with a hot instrument feeding for example into a compressor. You may clip the input of the compressor while your channel strip meters show you a perfectly reasonable level (since the signal has by now been compressed).

As for all the 3rd party plug-ins you mention that have recommended operating levels... we're no longer really talking about above-0dBFS-clipping, since those plug-ins may sound different at different levels even when the different levels are all below 0dBFS. It's obvious for example that a guitar amp will not sound the same depending on the level of the signal that feeds it, even if that signal stays below 0 dBFS. It's obvious that any dynamic plug-in with a threshold parameter will sound different depending on the level of the signal at its input, even if that level always stays below 0 dBFS. Etc...

As for all the 3rd party plug-ins you mention that have recommended operating levels... we're no longer really talking about above-0dBFS-clipping, since those plug-ins may sound different at different levels even when the different levels are all below 0dBFS. It's obvious for example that a guitar amp will not sound the same depending on the level of the signal that feeds it, even if that signal stays below 0 dBFS. It's obvious that any dynamic plug-in with a threshold parameter will sound different depending on the level of the signal at its input, even if that level always stays below 0 dBFS. Etc...

Including common guitar amp emulators like Amplitube et al

Including common guitar amp emulators like Amplitube et al

I'm sorry - that's what I meant: guitar amp modeling plug-ins.

Hehe, this thread is getting rather pedantic

It seems that the more plug-ins try to emulate analogue gear, the more notice DAW users have to pay to gain staging.

I don't think it is pedantic, Lagerfeldt knows what he's talking about and it turns out that even seasoned pro's can learn from him and his meticulousness (engineers must be meticulous, that comes with the title), even though he uses the word 'placebo' in the wrong way - he means perception bias and aural illusions (now that's pedantic - by me ).

All in all, it is a very useful thread about an important aspect of mixing, very useful to refer people to - it's bookmarked! Thanks everybody, so far!

Telling a newbie: "just watch your channel strip meters and keep reasonable levels" may not prevent them from clipping the input of plug-ins, especially with a hot instrument feeding for example into a compressor. You may clip the input of the compressor while your channel strip meters show you a perfectly reasonable level (since the signal has by now been compressed).

Which is exactly why I'm not doing that in my posts here, but rather trying to put some nuance into the discussion.

The fact is that many plug-ins will clip or behave differently with different levels, and it's especially true for Waves, UAD and TDM plug-ins in general.

So the gist of the video posted in the beginning of this thread is valid, though the explanations behind and the specific advice was not correct or optimal, but that should be rectified now by the additional information found in this thread.

I don't think it is pedantic, Lagerfeldt knows what he's talking about and it turns out that even seasoned pro's can learn from him and his meticulousness (engineers must be meticulous, that comes with the title), even though he uses the word 'placebo' in the wrong way - he means perception bias and aural illusions (now that's pedantic - by me ).

Heh heh, pedants unite! And I shall remember to use "perception bias" or perhaps the more common "confirmation bias" the next time

I like to record my acoustic guitar using a sm57 inside the guitar body panned center, Soundhole pickup panned left, and a saddle pick up panned right. As long as I don't clip at the interface, my signal sounds good.

So using three separate inputs, I output all channels to a common stereo bus. Adding a Gain plug in to the Aux 1 (common bus) channel strip, I increase the gain to about +6 and set that AUX output to Bus 2. Bus 2 is now the input to my Audio channel on which the guitar will be recorded.

You can see the levels in the left pic are all 'clipped' based on a 24 bit system, but not so in the 32bit floating point system. There is no distortion. Playing back the track, my signal is now peaking at 0dBFS (basically normalized with a crest factor around 12). (Center pic).

Repeating the same thing to experiment, I substitute the Gain plugin with a UAD Plug in making sure that my signal going to the audio strip does not exceed +6dB (that's right +6dB). After this signal is recorded, I can use the Sample editor to normalize the signal from 0dBFS to -6dBFS.

The input to the UAD /Gain plug in is around the -18dB peak level and then cranked up so that Logic will naturally brick wall the signal. Either way, this process will commit the guitar to the way it was recorded and panned. However, in most cases, the end result is similar if I record as three separate channels at a lower level. In this case I use a Gain plug in on the Stereo output and raise the gain to comfortably hear the guitar source(s).

Waves plug-ins output 24 bits and you can easily and audibly overload some of them.

Many of Waves plug-ins work in Double Precision nowadays and dither their output to 32-bit floating.

Here's the list as it is today:

Double Precision | Waves

That said, I'm all for gain staging myself.

Ah, pedantic perhaps, but certainly edifying.

Could perhaps someone explicate, or at least put forth conjecture as to the grounds for the assumption that said double precision and it’s proclivity for floating point math should somehow obnubilate a similar single point device.

Or

Since there seems to be a debate as to whether the difference between double precision and single precision is as much about storage capacity as precision, are double precision plugins that much better?