Here is something about headphone measurements.
Headphone Measurement Proceedures - Introduction and Equipment | InnerFidelity
Curryman is from Head Acoustics so maybe he can explain.
Headphone Measurement Proceedures - Introduction and Equipment | InnerFidelity
Curryman is from Head Acoustics so maybe he can explain.
The big question is the transfer characteristics (HRTF) of the ear canal of the individual. Would the following indirect method work?
1. Measure a loudspeaker with a measuring microphone, equalize its response perfectly flat
2. Put a tiny microphone in the ear canal, measure the response of the above equalized loudspeaker from the same distance (you will get your individual HRTF)
3. Leave the microphone in the ear canal, put on the headhones, equalize them so that you get the same response as in 2.
1. Measure a loudspeaker with a measuring microphone, equalize its response perfectly flat
2. Put a tiny microphone in the ear canal, measure the response of the above equalized loudspeaker from the same distance (you will get your individual HRTF)
3. Leave the microphone in the ear canal, put on the headhones, equalize them so that you get the same response as in 2.
This reminds me of mobile phone cameras.
They are known to have really crappy lenses.
But then the camera chip has so much resolution these days.
So they just make an image of a square grid and do software correction.
Maybe that will happen to Audio as well eventually.
Just do a measurement and then digital correction, no need for making things linear.
Luckily there are still people in Germany and in Japan who knows how to make perfect lenses.
And I guess we still have a few people here who knows how to make linear audio.
They serve a niche, but they still exist.
Patrick
They are known to have really crappy lenses.
But then the camera chip has so much resolution these days.
So they just make an image of a square grid and do software correction.
Maybe that will happen to Audio as well eventually.
Just do a measurement and then digital correction, no need for making things linear.
Luckily there are still people in Germany and in Japan who knows how to make perfect lenses.
And I guess we still have a few people here who knows how to make linear audio.
They serve a niche, but they still exist.
Patrick
Lens digital correction isn't something that's reserved only for mobile cameras, almost all major manufacturers do it. The problem with close-to-perfect optics is that they happen to be heavy and expensive. If you can offset that with digital correction, then it's a fair trade.
Now, if only in headphones we were left with the "expensive and heavy" problem to get perfection. The harsh reality is that for headphones there still isn't a standardised way how to measure their tonal response and therefore no standardised goal for headphone acoustic design. Which is why the vast majority of headphones sound extremely poorly and performance variation in terms of tonal balance is huge.
Here's the tonal balance of some of the more popular models.
Then there's sample to sample variation -
And inter-channel differences due to cup cavity volume imbalance -
All of it can be addressed by digital calibration.
Now, if only in headphones we were left with the "expensive and heavy" problem to get perfection. The harsh reality is that for headphones there still isn't a standardised way how to measure their tonal response and therefore no standardised goal for headphone acoustic design. Which is why the vast majority of headphones sound extremely poorly and performance variation in terms of tonal balance is huge.
Here's the tonal balance of some of the more popular models.
An externally hosted image should be here but it was not working when we last tested it.
Then there's sample to sample variation -
An externally hosted image should be here but it was not working when we last tested it.
And inter-channel differences due to cup cavity volume imbalance -
An externally hosted image should be here but it was not working when we last tested it.
All of it can be addressed by digital calibration.
Hm, big variation indeed between different models. Were they measured by using an artifical head (B&K torso)?
I was thinking about the best measuring method. Would the following be a good universal measurement? We can start from the assumption that the frequency response of an omnidirectional (pressure sensitive) measuring microphone is flat.
- Place the artifical head in an anechoic chamber
- Use a reasonably flat wideband loudspeaker at some distance (2 to 3 meters) in front
- Drive the loudspeaker with pink noise
- Equalize the driving signal of the loudspeaker so that the microphone output becomes flat. This will equalize both the loudspeaker frequency response AND the HRTF of the artifical head. We don't know yet, what is due to the loudspeaker and what is due to the HRTF, but it has no importance for now
- Now replace the loudspeaker with the headphone, using the previous equalized signal. Any irregularities will be due to the headphone only, so now you are measuring the REAL frequency response of the headhone on the artifical head. This should be equalized for the particular headphone.
Of course, the frequency response of a headphone on a real human could be measured the same way (as the HRTF differs from one individual to the other).
I was thinking about the best measuring method. Would the following be a good universal measurement? We can start from the assumption that the frequency response of an omnidirectional (pressure sensitive) measuring microphone is flat.
- Place the artifical head in an anechoic chamber
- Use a reasonably flat wideband loudspeaker at some distance (2 to 3 meters) in front
- Drive the loudspeaker with pink noise
- Equalize the driving signal of the loudspeaker so that the microphone output becomes flat. This will equalize both the loudspeaker frequency response AND the HRTF of the artifical head. We don't know yet, what is due to the loudspeaker and what is due to the HRTF, but it has no importance for now
- Now replace the loudspeaker with the headphone, using the previous equalized signal. Any irregularities will be due to the headphone only, so now you are measuring the REAL frequency response of the headhone on the artifical head. This should be equalized for the particular headphone.
Of course, the frequency response of a headphone on a real human could be measured the same way (as the HRTF differs from one individual to the other).
The target curve is a tough one. There have been many and indeed the best way to determine one would be to listen to a pair of headphones and then calibrated speakers until both sound the same. It took us around a year to do it mostly because it was done by hand and ear and we were lucky that the headphones we used were reasonably close to the target.
Using a dummy would be tough because only recently there have serious offerings for headphone tonal response measurements. We tried the usual suspects around 5 years ago and were forced to develop our own measurement system. Even then it's not perfect and some headphones behave differently from others, which is why we have 4 engineers who compensate for measurement artefacts by hand and ear.
Using a dummy would be tough because only recently there have serious offerings for headphone tonal response measurements. We tried the usual suspects around 5 years ago and were forced to develop our own measurement system. Even then it's not perfect and some headphones behave differently from others, which is why we have 4 engineers who compensate for measurement artefacts by hand and ear.
Are those free graphs (mic near the driver no other interface than a shielding barrier)? Or taken on a head and torso dummy with a rubber ear replica, the mic at the base of the ear canal, then reshaped by that whole system's reverse transfer function?
So those curves are a differential result to the total transfer function of some proprietary coupling system to your company. Thus we can't interpret them to some published standard. In other words they tell nothing definite to us more than how much they deviate from your non disclosed standard target.
Using the method that I described it could be done objectively (hand and ear intervention excluded), and due to the recent advancements in DSP technology it can be done quite easily.
I suppose you haven't got access to an anechoic chamber, therefore you rely on the more complicated subjective method.
MONACOR - ECM-2500
This mic modified is not that bad up to ca.10kHz to use in ear. I am working on a method too.
This mic modified is not that bad up to ca.10kHz to use in ear. I am working on a method too.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.