Originally Posted by LoveKnight
So what about the graph that Dan gave us himself last couple months ago? Seem like if innerfidelity or I mean Tyll measures the Alpha Dog then we will know more precisely but I think each company has at least one or two dedicated methods to measure.
It's pretty much this. Different hardware and rig setups, different software, different (or no) compensation curves, etc. When looking at any set of measurements, it's important to know all of these details in order to best be able to "read" the measurements and get good information from them. There's not a ton of consensus about all this from what I can tell, hence the sometimes wildly different results. And there are simply some measurement setups and methods that are just not good overall or only occasionally provide useful information.
The best thing one can do is get their hands on a lot of gear and look at a lot of measurements so they can find which methods most often correspond to what they hear (or find a way to mix attributes of measurements). The key is not tricking yourself into hearing what you see in measurements, much like it can be the same way with how subjective impressions influence what we hear or expect to hear. Easier said than done.
I'm sure there will be better, more accurate, and more repeatable ways to get headphone measurements in the future.
The Russian graph being referred to looks like they use too much smoothing, for one. Smoothing can "cover up" dips, peaks, and other imperfections that are actually quite audible (too little smoothing can make it appear rougher than what we'd really hear, though). Based on my experience with the MD and AD, I'd be hesitant to use this site's measurements before some other sources of measurements.