[bigshot]

  This is certainly true if you use the compensated data (the red/blue lines). However IF also provides the raw, uncompensated data (grey lines).

 
Then you have to compensate yourself. It's a lot easier just to take the curve from a set of headphones that is known to be flat measured by the same person. For instance Stax.

 

[higbvuyb]

  but aren't raw data smoothed and if so, not always with the same margin? I'm asking as a general question, because if you look at goldenears graphs the answer is pretty obvious for them ^_^.

Raw data should not be smoothed (but it sometimes is). However neither of the graphs in question appear to be substantially smoothed.
 
 

   
Then you have to compensate yourself. It's a lot easier just to take the curve from a set of headphones that is known to be flat measured by the same person. For instance Stax.

Either way, you are comparing a measured response, and determining how it differs from a target response. i.e. compensating. Whether your target response is something published in a peer-reviewed paper, or 'Stax' (whether Stax is 'flat' or not is arguable).

 

[bigshot]

You missed my point. If you compare two headphone curves from the same source, hopefully they are measured the same. Using a curve from one source to be a model for another is a recipe for error.
 
In any case, all measurements made with a microphone aren't the same as how it sounds to the ear. I checked my headphones by ear using a tone sweep. I know where they stand. They are flat enough for government work between 40Hz and 10kHz. That is the range that really counts.
 
I compared the graph for my headphone (Oppo PM-1) to the Stax curve (measured by the same person) and the Stax was pretty much what I would like the PM-1s to be in the ideal world. Even better than good enough for government work. That's why I suggested those as an ideal model.

 

[Hapster]

Or you could eq it to make it look more like this. Tyll did say the FSP were the most neutral pair he's heard.




Obviously it's making large assumptions, but it's easy to follow

 

[bigshot]

I don't think the bump up in the upper mids is correct on that chart. I pointed to a chart I thought was pretty neutral earlier in the thread.

 

[Hapster]

Whoops, forgot to add the actual image.

 

[higbvuyb]

  You missed my point. If you compare two headphone curves from the same source, hopefully they are measured the same. Using a curve from one source to be a model for another is a recipe for error.
 
In any case, all measurements made with a microphone aren't the same as how it sounds to the ear. I checked my headphones by ear using a tone sweep. I know where they stand. They are flat enough for government work between 40Hz and 10kHz. That is the range that really counts.
 
I compared the graph for my headphone (Oppo PM-1) to the Stax curve (measured by the same person) and the Stax was pretty much what I would like the PM-1s to be in the ideal world. Even better than good enough for government work. That's why I suggested those as an ideal model.

That's because the point was largely irrelevant.
 
The transfer function of one calibrated, IEC-standard ear coupler is going to be similar to another, calibrated, IEC-standard ear coupler (We have an advantage here in that the OP is using in-ear monitors rather than headphones).
That's what standards are for. Again, the major difference is going to be the difference between Harman's measuring setup, and the OP's actual ear.
Yes, people use different compensation curves. However, this is totally irrelevant because we are using raw data without compensation.
 
Using your ear to calibrate his headphones is based on the erroneous assumption that your ear is the same as his ear (and, notably, human ears are not standardized).

 

[castleofargh]

 

  but aren't raw data smoothed and if so, not always with the same margin? I'm asking as a general question, because if you look at goldenears graphs the answer is pretty obvious for them ^_^.

Raw data should not be smoothed (but it sometimes is). However neither of the graphs in question appear to be substantially smoothed.

thanks. well I guessed that straight point to point graph wouldn't improve readability, there is obviously some small matter of smoothing, but I wondered if there was like a formal default value or if it's just another thing you set yourself ?
I suppose the heads come with some protocols and recommendations about all this.

 

[bigshot]

  Using your ear to calibrate his headphones is based on the erroneous assumption that your ear is the same as his ear (and, notably, human ears are not standardized).

 
You're making the erroneous assumption that when I EQ I am doing it for someone else's ears.
 
The best way to know what is audibly flat to your ears is to EQ with tones using your ears.

 

[higbvuyb]

   
You're making the erroneous assumption that when I EQ I am doing it for someone else's ears.

 
Your advice was:
"I suggested those as an ideal model." [because they sounded flat to your ears].
 
Which has the underlying assumption that he hears like you do. Or are you suggesting the OP calibrate his headphones for your ears?
 

  The best way to know what is audibly flat to your ears is to EQ with tones using your ears.

And the tones should sound, what? Equal in intensity? Can't do that, human ears don't directly perceive intensity. Equal in loudness? Can't do that because of the equal loudness contour. Calibrate to the ELC? What phon and why?

 

[bigshot]

Actually, it was my sound engineer friend's ears who EQs flat with tones for a living. We both EQed Oppo's PM-1s by ear and came up with the same results. I have him double check what I EQ. If you made the minor adjustments in the Oppos we made, it would come out almost exactly those Stax. Check it yourself however you want. You'll find out too. I'm not splitting the atom. I'm making a good sounding stereo system. People make that seem a lot harder than it is.

 

[higbvuyb]

  Actually, it was my sound engineer friend's ears who EQs flat with tones for a living. We both EQed Oppo's PM-1s by ear and came up with the same results. I have him double check what I EQ. If you made the minor adjustments in the Oppos we made, it would come out almost exactly those Stax. Check it yourself however you want. You'll find out too. I'm not splitting the atom. I'm making a good sounding stereo system. People make that seem a lot harder than it is.

Essentially the OP can use the reference target as a base, but, yes, he should be making those final adjustments himself.
 
It seems hard because it is hard. You can simplify to a degree and still be 'good enough'; however you can see the inherent problems: someone's 'good enough' may be 'base decisions off absolutely uncalibrated opinions on the internet'.
 
Quote:

  thanks. well I guessed that straight point to point graph wouldn't improve readability, there is obviously some small matter of smoothing, but I wondered if there was like a formal default value or if it's just another thing you set yourself ?
I suppose the heads come with some protocols and recommendations about all this.

It's just something you set. The raw data is just the data from the microphone (compensated for the microphone's response). Because it's real, measured data, it's going to have a bit of noise in it that makes it hard to read. Which is what the smoothing deals with. But it's like noise from a camera sensor; you have to strike a balance between noise reduction and loss of detail; there's no one ideal/standard value.

 

[bigshot]

Equalization is a process, not a destination. You start with a destination- a spit in the wind as close to flat as you can get. Then you fine tune. It isn't hard, it just isn't ever necessarily finished. I still make adjustments to my curves occasionally.