This is hard to really explain, but I will try to make it clear.
If you look at raw measurements made with professional systems, like GRAS, you will see what looks like a regular frequency response graph that is "bent" and warped. When you then overlay the compensation curve, you can see that it usually follows that same "bend" in a way that makes sense. Applying the compensation curve gives you a readable frequency response, and it only appears to have been gently bent back down, visually.
If you look at EARS compensation curves, they look similar to GRAS ones, but the raw measurements absolutely do not. Instead of following that curve fairly closely, they look more flat with a single thin razor sharp peak. When the comp curve is overlaid, you see that the peak goes up to "meet" the compensation curve, but it's very clear that the compensation curve is doing way way more than it needs to do here.
The location and intensity of this thin peak changes per headphone, but it seems to always "meet" the compensation curve when overlaid. So the compensation curve can be seen as a "map" of how "loud" the resonant peak will be at any point across the spectrum.
What I have done is to use a MiniDSP 2X4 to replicate the compensation curve with EQ. Then, instead of keepkng that super wide bump, I narrowed it to match the peak exactly. The resulting graph looks very very similar to something like Innerfidelity's measurements.
So my proposal is this: an optional "real-time compensation" that is a replica of the predicted frequency response graph in the EQ view of REW, but in the SPL/All SPL views. That way you can fine tune it per headphone when doing measurements. You couls load the stock compensation curve ans use it as a "guide". One very cool idea would be to have the EQ adjust the gain of all the bands automatically to match the loaded compensation curve, and just let the user choose frequency and Q.
As it is right now, the only way to get a coherent graph is to:
-load compensation curve as house curve
-enter EQ section of REW and replicate compensation curve as exactly as possible with "view target" enabled, so the filters and target are 1:1.
-while viewing the "before" graph only, adjust the main EQ Q and gain to match what you see in the measurement
-activate predicted graph to get your compensated measurement.
With EARS you are basically measuring through a narrow little tube with a complicated ear shape that gets resonances in weird places depending on size and placement of the driver, as well as reflections or absorbtion of ear pads.
If that was hard to understand, then let me know and I'll make a video or something!
If you look at raw measurements made with professional systems, like GRAS, you will see what looks like a regular frequency response graph that is "bent" and warped. When you then overlay the compensation curve, you can see that it usually follows that same "bend" in a way that makes sense. Applying the compensation curve gives you a readable frequency response, and it only appears to have been gently bent back down, visually.
If you look at EARS compensation curves, they look similar to GRAS ones, but the raw measurements absolutely do not. Instead of following that curve fairly closely, they look more flat with a single thin razor sharp peak. When the comp curve is overlaid, you see that the peak goes up to "meet" the compensation curve, but it's very clear that the compensation curve is doing way way more than it needs to do here.
The location and intensity of this thin peak changes per headphone, but it seems to always "meet" the compensation curve when overlaid. So the compensation curve can be seen as a "map" of how "loud" the resonant peak will be at any point across the spectrum.
What I have done is to use a MiniDSP 2X4 to replicate the compensation curve with EQ. Then, instead of keepkng that super wide bump, I narrowed it to match the peak exactly. The resulting graph looks very very similar to something like Innerfidelity's measurements.
So my proposal is this: an optional "real-time compensation" that is a replica of the predicted frequency response graph in the EQ view of REW, but in the SPL/All SPL views. That way you can fine tune it per headphone when doing measurements. You couls load the stock compensation curve ans use it as a "guide". One very cool idea would be to have the EQ adjust the gain of all the bands automatically to match the loaded compensation curve, and just let the user choose frequency and Q.
As it is right now, the only way to get a coherent graph is to:
-load compensation curve as house curve
-enter EQ section of REW and replicate compensation curve as exactly as possible with "view target" enabled, so the filters and target are 1:1.
-while viewing the "before" graph only, adjust the main EQ Q and gain to match what you see in the measurement
-activate predicted graph to get your compensated measurement.
With EARS you are basically measuring through a narrow little tube with a complicated ear shape that gets resonances in weird places depending on size and placement of the driver, as well as reflections or absorbtion of ear pads.
If that was hard to understand, then let me know and I'll make a video or something!
