Headphones are IIR filters? [GRAPHS!] - Page 2

 nice work, will be watching :)

Quote:

Originally Posted by arnaud 

I could not check your reference but found this one where the measured stax headphone had non minimum phase characteristics (which causes troubles for the creation of an inverse filter): http://www.extra.research.philips.com/hera/people/aarts/RMA_papers/aar90.pdf

Even if certain headphone drivers oder measurement techniques produce non-minimum phase results you can still calculate a minimum phase filter.

Phase doesn't really matter with single driver headphones anyway, so all you need to get right is the magnitude response.

what does a square wave test represent? The decay of bass frequencies? From what I know about loudspeakers, you don't need to worry about phase with single driver headphones. Makes me wonder what type of filter roll off they use in the tf10's tho.
 

It indicates a lot of different things.  Generally a straight top that doesn't curve inwards on the the 30/50hz indicates tight bass.  The more concave it is the looser it is.

The 300/500hz wave can indicate smoothness and coherency.  It should rise quickly, overshoot a bit, and settle just as fast.  Larger secondary overshoots when settling usually indicate some sort of incoherency, poor imaging, or indistinct transients.  The longer these continue the worse it is.  Regular small amplitude ringing along the top indicates a sharper ringing at a specific frequency.

Quote:

Originally Posted by maverickronin 

It indicates a lot of different things.  Generally a straight top that doesn't curve inwards on the the 30/50hz indicates tight bass.  The more concave it is the looser it is.

I don't think this can be generalized like that. Better look at the (non-smoothed) FR and not time response data.

Unless you know how to read impulse responses, which is cumbersome without some tool, don't try to read step or square wave responses.

Quote:

Originally Posted by xnor 

I don't think this can be generalized like that. Better look at the (non-smoothed) FR and not time response data.

 

Unless you know how to read impulse responses, which is cumbersome without some tool, don't try to read step or square wave responses.

Do you just mean the bass in particular?

I know it's not perfect and I'm not an expert in the math behind it IME the correlation is pretty good.

The results look the way they do becaues of non flat FR and associated non flat phase response and loss of sound pressure -> high pass.

Yeah it's mostly the tight/loose bass I'm talking about. Have you seen the Sony XB measurements?

Yeah.  Really bass heavy pairs like that end up being convex instead.

I was doing some testing on matching frequency responses across different models of headphones (as per the findings of the paper here and here) and noticed that the impulse response indeed seems to follow the frequency response. This is nothing new in the context of this thread of course (the link between impulse and frequency response has already been established), but I thought the 'phenomenon' was interesting to see anyway.

I used the program HOLMImpulse to generate these graphs, but can't find an exact explanation of how it calculates the impulse response data. I do know that it seems unable to generate an impulse response unless the input data are an actual recorded audio signal (as opposed to a text file that only contains frequency response data), but that's it.

In any case, below in the black line is the (non-HRTF corrected - and these were not carefully or accurately measured) frequency response of the Yuin PK3 buds, and in grey the response of stock Creative buds. The pink response is the Creative buds with an EQ that attempts to emulate the PK3's response on the Creatives.

Same thing as above, but the reference (black line) is a pair of AKG K 314 buds.