[crinacle]

Alright so after some tests, apparently the output of the iPhone isn't quite precise even in the higher freqs. Nothing too major, nothing related to the physical impedance of the system though so everything should be easily solved with a bit of compensation.

It will take quite a bit of work though, and re-adjusting 900+ points of data is not small task. So yeah, expect that down the pipeline whenever I have the free time. In the meantime though, just take note of the differences.

Those results looks really, really good. Ridiculously smooth For whatever reason, these results look way smoother than those I'm generating from my iPad. (I will revisit this and check my FFT settings.)

Just one thought... Is the roll-off you're compensating for coming from your iPhones DAC or ADC? I presume it's the latter? Have you ever tried using a better, low z-out source for the (unfiltered) pink noise, just to confirm? I know it wouldn't be as portable (and you've clearly shown it's not necessary) to have to carry a separate DAP, but I'm just curious.

To answer your question, as above. Could be both a DAC and ADC issue since feeding in signals from my USB interface doesn't create the discrepancy in the high freqs on AudioTools but doesn't solve the bass rolloff issue.

EDIT: ON THE OTHER HAND...

The "attenuated" higher freqs on my original measurements match up to @jude's Audio Precision rig much more closely.

Jude's IE800S measurement:

My original unaltered measurement:

My "fixed" measurement:

Back to the drawing board...?

 

[csglinux]

Back to the drawing board...?

My uncompensated iPad measurements are very close to what I'm measuring through REW with my StarTech or Focusrite Forte ADCs. The Android AudioTool on my V30 gives a massive roll-off in the bass and requires tons of correction, but the only difference with the iOS/AudioTools on my iPad is a very slight roll-off approaching 20 kHz. I have an old iPhone 5s somewhere that I could try, but I don't have a newer iPhone/iOS version to test.

Anybody else tried using the "FFT Plot" app on iOS? It seems faster and more accurate than AudioTools - unfortunately my version doesn't allow you to export ASCII data from the FR measurements.

 

[malvinviriya]

I had some experience with the FFT Plot app as well. It's pretty nice and with Crin's PWN file and an IEC compliant mic it actually is a really nice tool for me to get an EQ setting

Kinda sucks that it's not able to export at all. But then again I don't think the devs had that in mind when it was designed.

 

[JohnYang1997]

Jude's new hires kemar is inaccurate. The change in acoustic impedance is faulty. Please compare to traditional 60318-4 type.

www.0db.co.kr/xe/
m.blog.naver.com/PostList.nhn?blogId=gre_nada

 

[crinacle]

Jude's new hires kemar is inaccurate. The change in acoustic impedance is faulty. Please compare to traditional 60318-4 type.

www.0db.co.kr/xe/
m.blog.naver.com/PostList.nhn?blogId=gre_nada

Inaccurate in which regards?

 

[JohnYang1997]

ie800s from 0db
measured with gras ra0045 coupler which is the traditional 60318-4 coupler.

 

[crinacle]

ie800s from 0db
measured with gras ra0045 coupler which is the traditional 60318-4 coupler.

So, you're saying that the AP rig with the RA0401/02 isn't accurate because of the high frequency dampers installed to eliminate resonance?

 

[JohnYang1997]

So, you're saying that the AP rig with the RA0401/02 isn't accurate because of the high frequency dampers installed to eliminate resonance?

Basically yes. The damped response does not represent the response at human ear drum (what human hear) Less highs.

 

[crinacle]

Basically yes. The damped response does not represent the response at human ear drum (what human hear) Less highs.

There was a debate on that topic that happened with the MDR-Z1R measurements, basically Head-Fi vs Innerfidelity in contention over the presence of a 10k peak. I'm still somewhat on the fence but I'm inclined to lean more towards GRAS' modification of the 318 coupler since the lack (or lower magnitude) of the resonant peak helps with the interpretation of measurements.

 

[csglinux]

Basically yes. The damped response does not represent the response at human ear drum (what human hear) Less highs.

My understanding is that the new in-ear simulator is supposed to extend the current standard to more accurately represent the actual ear impedance at higher frequencies. At least, this is GRAS' claim. I don't know how we'd disprove that without inserting microphones into a large-enough sample of human cadavers to ensure statistical relevance?

 

[JohnYang1997]

My understanding is that the new in-ear simulator is supposed to extend the current standard to more accurately represent the actual ear impedance at higher frequencies. At least, this is GRAS' claim. I don't know how we'd disprove that without inserting microphones into a large-enough sample of human cadavers to ensure statistical relevance?

That's what b&k did. They stayed true to human.
The 60318-4 did not specify the amplitude of the peak at 13.5k. So gras just met the standard but actually the result change the frequency reaprese too much with shallow insertion earphones.

 

[csglinux]

That's what b&k did. They stayed true to human.
The 60318-4 did not specify the amplitude of the peak at 13.5k. So gras just met the standard but actually the result change the frequency reaprese too much with shallow insertion earphones.

It's unfortunate to have created another cause for measurement discrepancies, but maybe there was a reason for what GRAS did, beyond just $$$s?
Do you have any links, references or citations to show that B&K couplers measure closer to that of the actual human ear than the newer GRAS couplers?

 

[JohnYang1997]

It's unfortunate to have created another cause for measurement discrepancies, but maybe there was a reason for what GRAS did, beyond just $$$s?
Do you have any links, references or citations to show that B&K couplers measure closer to that of the actual human ear than the newer GRAS couplers?

The main reason gras did that is for measurement consistency. It masks a lot of the inconsistency from different insert depth etc.
As for the comparisons, I don't have direct comparison but there are indirect comparisons.
I believed that the n5005 measurements from harman slides is from bk's new hires head. And naver earfi does have n5005 measured as well.
There are not much measurements done with bk's new head on the internet yet.
The comparison between gras new and old couplers can be seen from any measurements done by jude and 0db, earfi, speakerphone.

 

[jude]

Basically yes. The damped response does not represent the response at human ear drum (what human hear) Less highs.

Here's GRAS's whitepaper on their HR simulators:

High Resolution Ear Simulator By Morten Wille (October 2017)

The high Q, 1/2-wave resonance at 13.5 kHz is part of the specification that defines the main volume, but is actually outside of the standard's defined response tolerance (which stops at 10 kHz, as illustrated in Figure 3 in the whitepaper).

Please read the entire whitepaper (linked above), but here's a short excerpt from it:

The standard IEC 60318-4 (former IEC 60711) Ear Simulator was designed in the early 1980s and mimics the input and transfer impedance of a human ear. While the input impedance was based on measurements on human subjects, the transfer impedance was based on the assumption that the ear canal is a simple cylindrical volume with a hard termination. Obviously the human ear canal is not a cylindrical cavity and the tympanic membrane is at an angle to the tapered ear canal. This questions the validity of the transfer impedance, particularly at high frequencies.

When the Ear Simulator was designed in the early 1980s the need for high frequency measurements above 10 kHz was limited. Within the hearing aid industry 8 kHz was considered adequate. Modern hearing aids and consumer electronics such as headphones require measurements at frequencies up to 20 kHz and beyond. When measuring with the Ear Simulator, the Device Under Test (DUT) is typically coupled by means of an ear canal ex- tension and a rubber pinna...

In the whitepaper, you can see in Figure 7 and Figure 8 the response differences between a typical 60318-4 simulator and the newer simulator.

That's what b&k did. They stayed true to human.
The 60318-4 did not specify the amplitude of the peak at 13.5k. So gras just met the standard but actually the result change the frequency reaprese too much with shallow insertion earphones.

Trivia: The "GR" in "GRAS," by the way, stands for Gunnar Rasmussen, who co-developed the original 60711 standard (now the 60318-4 standard). Gunnar Rasmussen left Brüel & Kjær and founded GRAS in 1994.

The main reason gras did that is for measurement consistency. It masks a lot of the inconsistency from different insert depth etc...

In the whitepaper (linked above), you can see in Figure 4 and Figure 9 the comparison of the two (standard 60318-4 and the newer simulators), in terms of transfer impedance with varying canal lengths.

...I believed that the n5005 measurements from harman slides is from bk's new hires head...

I don't believe Harman has published any measurements from the Brüel & Kjær 5128 yet.

...There are not much measurements done with bk's new head on the internet yet.
The comparison between gras new and old couplers can be seen from any measurements done by jude and 0db, earfi, speakerphone.

If you listen to my discussion with Sean Olive yesterday, you'll hear at around 21:55 that Sean and his team are now also using the newer GRAS ear simulators and anthropometric pinnae in their GRAS 45CA. And they also just recently acquired the Brüel & Kjær 5128 that you're referring to.

"So we have what you have now. We have the latest (GRAS) coupler and the latest pinnae...and we also have the new B&K HATS 5128..."

...There are not much measurements done with bk's new head on the internet yet.

We will likely also be doing measurements with the Brüel & Kjær 5128 in the future, so should be able to make some direct comparisons at some point.

The original 60711 (60318-4) standard is nearly 40 years old. You'll continue to see changes and advances over time, as the needs for more accurate measurements and more accurate modeling of human perception beyond 10 kHz are becoming increasingly important.

 

[JohnYang1997]

Here's GRAS's whitepaper on their HR simulators:

High Resolution Ear Simulator By Morten Wille (October 2017)

The high Q, 1/2-wave resonance at 13.5 kHz is part of the specification that defines the main volume, but is actually outside of the standard's defined response tolerance (which stops at 10 kHz, as illustrated in Figure 3 in the whitepaper).

Please read the entire whitepaper (linked above), but here's a short excerpt from it:



In the whitepaper, you can see in Figure 7 and Figure 8 the response differences between a typical 60318-4 simulator and the newer simulator.



Trivia: The "GR" in "GRAS," by the way, stands for Gunnar Rasmussen, who co-developed the original 60711 standard (now the 60318-4 standard). Gunnar Rasmussen left Brüel & Kjær and founded GRAS in 1994.



In the whitepaper (linked above), you can see in Figure 4 and Figure 9 the comparison of the two (standard 60318-4 and the newer simulators), in terms of transfer impedance with varying canal lengths.



I don't believe Harman has published any measurements from the Brüel & Kjær 5128 yet.



If you listen to my discussion with Sean Olive yesterday, you'll hear at around 21:55 that Sean and his team are now also using the newer GRAS ear simulators and anthropometric pinnae in their GRAS 45CA. And they also just recently acquired the Brüel & Kjær 5128 that you're referring to.

"So we have what you have now. We have the latest (GRAS) coupler and the latest pinnae...and we also have the new B&K HATS 5128..."



We will likely also be doing measurements with the Brüel & Kjær 5128 in the future, so should be able to make some direct comparisons at some point.

The original 60711 (60318-4) standard is nearly 40 years old. You'll continue to see changes and advances over time, as the needs for more accurate measurements and modeling human perception beyond 10 kHz are becoming increasingly important.

The key points here are
1, 60318-4 only specify deep insertion(at reference plane) and the resonance peak is at 13.5k. With shallow insertion the peak moves to lower frequency.
2, As the peak moves to lower frequency the 10k tolerance is not valid any more. That's why old measurements suggests that they are not guaranteed to be accurate after 7khz.
3, The main issue is that even with that big of change in acoustic impedance they still managed to be within the standard.
4, continued the last point but in a different direction, the standard is indeed old and needs improvement.
And I would love to know the initial motivation of gras making hires couplers. In my mind they were screwed over by sony and sony went to bk later on.