[Rynsin]

I'm getting a bit confused with the wonderful graphs over at HeadRoom. I'm unable to figure out what the difference between "frequency response", "frequency response compensated", and "frequency response raw." I currently have access to HiFIMAN HE-400s and Beyerdynamic DT990s, and I'm interested in comparing them to the NAD VISO HP50s.
Here's the "frequency response" graph of these three headphones:

From this graph, I read the HE-400s will have less bass than both of the other two headphones. However, when I look at the "frequency response raw" graph, I see this:

... which seems to indicate that that DT990s will have lower bass response than the HE-400s! Not only is this the opposite conclusion of the previous graph, but it seems to be the opposite of what I observe when listening.
 
I'm also confused why the line for the HE-400s appear to be almost exclusively below zero. Is this because HeadRoom measures the frequency response of all their headphones using the same power (relative to the resistance of the headphones)? If that's the case, can I assume that changing the volume on my amp will shift the frequency response curve vertically?
It looks like the "frequency response compensated" graph shows all the curves shifted so that they are zero at 1000Hz.

"Compensating" for frequency response by shifting each curve to be 0 at a single point seems kind of silly to me. Wouldn't it be significantly better to shift each curve so that it is equally above and below 0 (such that the integral of the amplitude from 10Hz to 20,000Hz is equal to zero)? I don't really care how the frequency response curves compare at a set volume, because I'm likely to adjust my amp when I switch headphones. I care much more about the sound signature of the headphones once I've set them to a comfortable volume.
 
Obviously, the people at HeadRoom are pretty smart, and I realize that I'm probably just not understanding something that they're doing. Anyone care to fill me in?
 
tl;dr -- what is the difference between "frequency response", "frequency response compensated", and "frequency response raw" on the HeadRoom graphs?

 

[LCfiner]

RAW: actual sound level measured at microphone.
 
COMPENSATED: response is adjusted via mathematical “HRTF” (Head related transfer function). There’s articles on Innerfidelity about different types of HRTF compensations and why it’s used in measurements.
 
“Frequency Response” The result from the HRTF “compensated” result is smoothed out with a smoothing function in an attempt to make the graph easier to read. 
 
I just focus on the Compensated result. the smoothed version obscures too many details for me, personally. Headroom and Innerfidelity always use HRTF so the raw response is typically not what is compared around the sites or here when posters refer to measurement results. So just focus on Compensated, imo

 

[Rynsin]

Ah, that makes sense! Thank you so much for clarifying that for me.
 
I'm still a bit confused as to what determines the 0db line (obviously, it's what the microphone picks up as 0db, but that's a function of how much power is used to power the headphones). I'll do a few experiments and likely figure it out myself.
 
For my own vindication, where could I have found the information you posted? I'm surprised that HeadRoom/headphone.com doesn't have a page explaining their measurements, but I certainly couldn't find it. Or is this the kind of thing I should just know?

 

[audioman83]

I see that it's been 8 months since your question Rynsin, but I'll answer anyway. the 0db line in the middle has nothing to do with power requirements, or even how loud the cans were during measurement. The chart does not tell us what db spl the measurement is taken at, or what power was used to to drive the cans, or what weighting or response time was used. notice that the chart says dbr which means decibels to reference.
 
So if at a point in the x axis the colored "line" hits +3 that means that at that frequency (pitch) the sound is 3db higher than flat (0). Flat, aka 0 across the chart, is where every frequency is the same loudness* as every other frequency.

 

[Agent of Chaos]

  RAW: actual sound level measured at microphone.
 
COMPENSATED: response is adjusted via mathematical “HRTF” (Head related transfer function). There’s articles on Innerfidelity about different types of HRTF compensations and why it’s used in measurements.
 
“Frequency Response” The result from the HRTF “compensated” result is smoothed out with a smoothing function in an attempt to make the graph easier to read. 
 
I just focus on the Compensated result. the smoothed version obscures too many details for me, personally. Headroom and Innerfidelity always use HRTF so the raw response is typically not what is compared around the sites or here when posters refer to measurement results. So just focus on Compensated, imo

 

  I see that it's been 8 months since your question Rynsin, but I'll answer anyway. the 0db line in the middle has nothing to do with power requirements, or even how loud the cans were during measurement. The chart does not tell us what db spl the measurement is taken at, or what power was used to to drive the cans, or what weighting or response time was used. notice that the chart says dbr which means decibels to reference.
 
So if at a point in the x axis the colored "line" hits +3 that means that at that frequency (pitch) the sound is 3db higher than flat (0). Flat, aka 0 across the chart, is where every frequency is the same loudness* as every other frequency.

 
OH GOD these are old posts and I'm praying one of you can answer me
I noticed that some pros here actually avoid compensated FR graphs and route for measured raw data, Even Tyll himself prefers to look for raw measured FR because as he claims to be "inaccurate" anyways my understanding to the situation is that compensated is affected by HRTF and not all people are created equal, Their head/ear size, shape , etc differs but for that they use an average human head simulator, The raw measured is just the artificial ear (mic) response without considering HRTF where the target curve is harman's, Am I right anyone? please confirm this 

 

 

[Music Alchemist]

  The raw measured is just the artificial ear (mic) response without considering HRTF where the target curve is harman's, Am I right anyone? please confirm this 

 

 
Yep! The raw measurements are just going to show the actual measurements without altering them. But remember that those mesaurements are always going to be influenced by the equipment that is measuring it. (As for compensation curves, the Harman curve is only one target. There are many others, such as diffuse field, free field, and proprietary ones that companies like Golden Ears and Sonarworks use.)

 

[akg fanboy]

   
 
OH GOD these are old posts and I'm praying one of you can answer me
I noticed that some pros here actually avoid compensated FR graphs and route for measured raw data, Even Tyll himself prefers to look for raw measured FR because as he claims to be "inaccurate" anyways my understanding to the situation is that compensated is affected by HRTF and not all people are created equal, Their head/ear size, shape , etc differs but for that they use an average human head simulator, The raw measured is just the artificial ear (mic) response without considering HRTF where the target curve is harman's, Am I right anyone? please confirm this 

 

Habibi, the harman curve is not always the target. Harman curve is just the ideal response from harman to replicate a speaker, it basically boosts the bass from what is usually considered neutral for a headphone. A target is the reference a company or manufacturer is comparing the headphone response to.

 

[Agent of Chaos]

  Habibi, the harman curve is not always the target. Harman curve is just the ideal response from harman to replicate a speaker, it basically boosts the bass from what is usually considered neutral for a headphone. A target is the reference a company or manufacturer is comparing the headphone response to.

Thank you tiptop guy

 

[Agent of Chaos]

   
Yep! The raw measurements are just going to show the actual measurements without altering them. But remember that those mesaurements are always going to be influenced by the equipment that is measuring it. (As for compensation curves, the Harman curve is only one target. There are many others, such as diffuse field, free field, and proprietary ones that companies like Golden Ears and Sonarworks use.)

Thank you too, Music Alchemist.

 

[TheSonicTruth]

So which of the three does Sonarworks' calibration plug-in base its corrections on? Compensated, Raw, or plain old Frequency Response? And if I wanted to use a parametric of my own to do corrections with, which of the three should I reference?