Correct Setting for Spectrum Analyzers?

The correct setting is whatever you can understand.  It is not a reliable peak meter or RMS meter.  It is there to visualize problems in the frequency response.  The peaks are averaged out based on the FFT (block) number.  So a sine wave of 1000Hz will likely contain an average of all the surrounding frequencies until the next block.  A bad answer is just don't worry about it. Use the right tool for the right purpose.  

Thank you for the reply, drewfx1.  The reason why I ask is because I am trying to confirm what a Groove3 tutorial has instructed.  I began typing up this huge post before realizing that the data coming out of Voxengo Span is totally different (amplitude wise) than other spectrum analyzers.  Obviously, I want to compare apples to apples.  I would like to "straighten out" the data before starting another thread with the issue.

In terms of apples to apples it can be difficult. If every parameter was the same then you should get the same result. Unfortunately there are lots of parameters that are involved (including the sample rate of your project) and not all parameters in a given spectrum analyzer plugin are visible or controllable or even called by their proper technical names. Some plugins also have additional features like the "slope" in Span which I think just progressively boosts higher frequencies in the display - you probably want to set it to 0dB.

Thanks Daniel and Drew!
 
Overall, I'm not unfamiliar with the way spectrum analyzers work, and historically I have always used them to check the frequencies in a mix and allow me to visualize what my ears tell me (and my ears don't seem very musical these days).  In other words, I can see the frequencies to which to cut or boost, if necessary.  And that's what I use spectrum analyzers for...checking the frequencies...not necessarily checking the amplitude of said frequencies.  One exception to that though: checking the amplitudes in a mix relative to one another in the same EQ or spectrum analyzer tool.
 
I'll just go ahead and mention my dilemma here instead of another thread.
 
You see, having watched this Groove3 tutorial entitled "Top 10 EDM Don'ts," I find myself dumber for having watched it.  The author, whom I think does a great job in his explanations and training, instructed how to adjust levels of the kick's, and the bass's, fundamental frequency using a spectrum analyzer as opposed to using the standard, track VU meter with a peak/max-level indicator.  He doesn't advocate to not using the track meter to set levels, but it just wasn't mentioned (or shown) in this particular video.  He also advises to listen and check professional songs in the spectrum analyzer to see how they stack up.  Seems like decent advice except for one, big problem:
 
The fundamental frequency of a kick or bass note does not equal the peak amplitude "held" in a track's meter.
 
For example, a bass line sample is playing and looping.  It's track's audio max-peak indicator says -3dB.  It's way too loud for the mix, so the track's volume fader is adjusted so the indicator reads -8dB, let's say.  So in most simplistic cases, that's it; you've set the bass level.  But following his advice, you then check how loud the fundamental frequency of the bass sound is so you can compare it to the kick drum, let's say, and you now find that the amplitude is nowhere even close to what the track level meter says.  So how the hell can you actually "mix" with this advice??!?

For anyone with a Groove3 subscription, I encourage you to watch the 2nd video in the "Top 10 EDM Don'ts" video series to see what I'm talking about.  Also, as a primer to his methodology for "normal" level setting, check out the 8th video in "Top 10 EDM Do's" series.
 
EDIT: To rule out the possibility of one of the aforementioned samples having a transient or a click that would force the VU meter to peak at a louder mark, I verified that the sample did not have such a sound inside of it, zooming into the waveform to verify.

It is definitely not the only one with adjustable slope.  Feed in pink noise and adjust for a flat line. The slope needs to be set for 3dB on Span for the flat line. I think it defaults to 4.5dB
 
Although from the Fab Filter website mentioned below by timidi there is this information:
 
The Tilt (Slope) setting tilts the measured spectrum around 1 kHz with a specified slope, expressed in dB per octave. The default setting of 4.5 dB/oct results in a natural looking spectrum, resembling best how loudness is perceived by the human ear.
 
Levels wise I must say I just did some tests and it is dead on accurate. I know bitflipper has been praising its level accuracy and I would say he is right for sure. It is showing exact match for peak and rms levels compared to all my other metering options.  What I do like about Span is it has many metering options too which is also nice including K system levels where 0 dB VU on the meter agrees to either K-12, K-14 or K-20 ref level.

I lost my brain earlier - it doesn't show the levels correctly because of windowing.
 
And if you are going to compare sine waves, then you need to set the slope to 0dB to get an accurate comparison of relative levels.

drewfx1
I lost my brain earlier - it doesn't show the levels correctly because of windowing. And if you are going to compare sine waves, then you need to set the slope to 0dB to get an accurate comparison of relative levels.

I was referring to the level meter on the right being accurate. Not what the spectrum display itself is showing level wise. Although I do get the fact that the spectrum level is referring to a particular frequency band. That seems to be low compared to the actual (total) level of the signal. I would not be using that as a basis of accurate level measurement.

Jeff Evans

drewfx1
I lost my brain earlier - it doesn't show the levels correctly because of windowing. And if you are going to compare sine waves, then you need to set the slope to 0dB to get an accurate comparison of relative levels.

I was referring to the level meter on the right being accurate. Not what the spectrum display itself is showing level wise. Although I do get the fact that the spectrum level is referring to a particular frequency band. That seems to be low compared to the actual (total) level of the signal. I would not be using that as a basis of accurate level measurement.

Hopefully my brain is better today. 
 
The levels between meter types should be at least somewhat similar for isolated sine waves (i.e. a single sine wave with nothing else present), but obviously if you take a single -10dB sine wave at one frequency it will have a lower total level than two or more -10dB sines at different frequencies combined. And if you know you have an isolated sine wave, there's really no reason to bother with two different types of meters, right? 
 
So yes indeed, you're right - spectrum analyzers are only good for comparison between frequency bands and you can't really compare individual band levels to the overall level, at least not in most practical applications with a DAW.
 
 
In terms of the slopes, the difference between pink noise and white noise (and associated slopes) is that pink noise distributes the noise so that it's equal by octave (like our hearing) whereas white noise is equal by frequency.
 
FFT's do their thing by dividing things into equal frequency bands rather than equal octaves, so for "exactly how much at this individual frequency vs. that frequency" we want a 0dB slope (white noise appears as a horizontal line), and for "how much in each octave" we want pink noise to be the horizontal line - which corrects for the FFT dividing things into equal frequency bands vs. octaves.

Here's an in-depth and excellent tutorial on the subject:
 
http://www.soundonsound.com/techniques/monitor-wizard
 
Here's a source for noise calibration files (there might be better ones out there but this is the first I found):
 
http://abluesky.com/support/blue-sky-calibration-test-files/