[NoBruno]

I am using the MeterPlugs Loudness Penalty Analyzer. I run my final mix through the Maxim Limiter Plugin with the Ceiling set at 0.1 dB and try different threshold settings on Maxim until I find one that minimizes any audible artifacts from Maxim while yielding an optimal loudness penalty (LP). Now the question is what is an "optimal" loudness penalty?

Here is what I am thinking:
• You need to shoot for some LP. If your LP is 0+, that means you are below the target loudness for the streaming service and they my apply positive gain to your track. If the amount of gain added exceeds the difference between your max peak and full scale digital, then it will introduce digital clipping.
• On the other extreme, you don't want to drive Maxim so hard that there are easily audible artifacts. This will typically result in higher overall gain after LP, but at an obvious price.
• I am interested in the gray area between these two extremes. In this regime, the tradeoffs would seem to be straightforward. When comparing two possible Maxim settings, weigh effective gain against any sonic differences.
• I think of the effective gain as the gain introduced by Maxim less any resultant LP, or
• Effective Gain = Maxim Gain - LP,
○ (Maxim automatically inserts makeup gain equal to the absolute value of its threshold, so a threshold of -10 dB introduces gain of 10 dB)
○ (I am using the absolute value of LP, which is displayed as a negative number in the plugin)
• So if I am compare two Maxim settings it might look like this:
○ Setting 1: Threshold: -6 dB LP: 2.4 Effective gain: 6 - 2.4 = 3.6
○ Setting 2: Threshold: -8 dB LP: 3.9 Effective gain: 8 - 3.9 = 4.1
• In this example, I can get an extra .5 dB by driving the Maxim a little harder. I just need to decide if that is worth any sonic difference and make my choice of tradeoff.

EXCEPT, this approach assumes there is no further limiting in the signal chain from the point where the streamer's loudness penalty (or makeup gain) to the final listen. I don't know if this is true. For example, what if the streamer wants its service to sound louder overall, and so it introduces another gain stage and then limiting before sending to users? In that case, the artifacts introduced by the streamer's limiting might be worse than if I had just driven Maxim a little harder and done the limiting myself. So I guess my question is: does anyone know whether all full-scale peaks pass through the streamer's processing untouched or whether there is any additional limiting there? I have seen engineers in a number of forums recommend mastering with peak levels at -2dB. I can't think of any reason for this recommendation unless there is some limiting on the streamer's end that would trim the tops off of signals above this point.