Sean Olive provided a free download here https://aes2.org/publications/elibrary-page/?id=22696 TL;DR SoundGuys and Harman 2019 equivalent targets were more prefered than three other targets. It's noteworth that these two most prefered target are the most different from each other, of than unmodified DF.
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
A comparison of in-ear headphone target curves for the Brüel & Kjær Head & Torso Simulator Type 5128
- Thread starter markanini
- Start date
Not exactly new information that different listeners do have some variation in preference, but interesting to see it once again:
It's also interesting to note how strongly correlated age is to whether a listener falls in class 1 or class 2:
Beyond the preference stuff, I found figure 3 enormously interesting. It's clear the type 4.3 simulator does provide measurements that are more accurate to the acoustic impedance of the human ear (expected, but good to see).
Finally, I have to agree with markanini that it's interesting how the two targets that are arguably most different from each other score highest, and not from different groups either! The same group of listeners are rating these two targets roughly equally high. To me that would suggest at least 1 of 3 things:
1. Average listener preference is not very sensitive to variations in frequency response within reasonable limits as long as the bass and treble levels remain similar from one target to the next in relative levels (e.g the noted 8dB bass-treble delta that is mentioned in the paper).
2. None of these targets are ideal, and there exists a better "general" target that would be more highly preferred by the average listener.
3. IEMs bypass so many of our natural hearing functions (the majority of the HRTF) that each individual would most prefer a tailored target curve just for them, and as such the average score for a given target that is within reasonable bounds will tend to be similar for a sufficiently large set of listeners, i.e. listener 1 ranks curve 1 and 2 as 80/100 and 90/100 points respectively, while listener 2 ranks them as 90/100 and 80/100 respectively.
Which of these (if any) are true remains to be seen.
It's also interesting to note how strongly correlated age is to whether a listener falls in class 1 or class 2:
Beyond the preference stuff, I found figure 3 enormously interesting. It's clear the type 4.3 simulator does provide measurements that are more accurate to the acoustic impedance of the human ear (expected, but good to see).
Finally, I have to agree with markanini that it's interesting how the two targets that are arguably most different from each other score highest, and not from different groups either! The same group of listeners are rating these two targets roughly equally high. To me that would suggest at least 1 of 3 things:
1. Average listener preference is not very sensitive to variations in frequency response within reasonable limits as long as the bass and treble levels remain similar from one target to the next in relative levels (e.g the noted 8dB bass-treble delta that is mentioned in the paper).
2. None of these targets are ideal, and there exists a better "general" target that would be more highly preferred by the average listener.
3. IEMs bypass so many of our natural hearing functions (the majority of the HRTF) that each individual would most prefer a tailored target curve just for them, and as such the average score for a given target that is within reasonable bounds will tend to be similar for a sufficiently large set of listeners, i.e. listener 1 ranks curve 1 and 2 as 80/100 and 90/100 points respectively, while listener 2 ranks them as 90/100 and 80/100 respectively.
Which of these (if any) are true remains to be seen.
- Thread Starter
- #4
I think a reasonable speculation is that scrutiny to target deviations needs to be tempered by the range of valid IEM tuning preferences shown in the paper.
How can a target be with different parameters? Either it is a target or it isn´t.
- Thread Starter
- #6
Ask @MinimumPhase he created it.
I can speak of personal experience with v-shaped mids pairing better with elevated bass, and neutral mids pairing better with lean to moderate bass. But the bass and tilt controls seem too basic to cover this.
Speaking more objectively Sean Olive has determined that theres segmentation of varying bass level preference. Additionally, Sean has talked about the role of personalization, which broadens the scope beyond the singular target: https://danishsoundcluster.dk/wp-content/uploads/2022/11/Olive_DSD_2022.pdf
I can speak of personal experience with v-shaped mids pairing better with elevated bass, and neutral mids pairing better with lean to moderate bass. But the bass and tilt controls seem too basic to cover this.
Speaking more objectively Sean Olive has determined that theres segmentation of varying bass level preference. Additionally, Sean has talked about the role of personalization, which broadens the scope beyond the singular target: https://danishsoundcluster.dk/wp-content/uploads/2022/11/Olive_DSD_2022.pdf
Last edited:
Comparing JM-1 (tilt -0.8dB per octave, bass 8dB) with Harman2019IEV2 looks like this, so basically identical below 3kHz. So what then is all the fuss with the "new meta" ?
So I honestly don't get it. First of all, there is no unique target, as it depends on what values you choose for tilt and bass. But even if it would be fixed, say for the values given here, it only has a treble dip betweed 3 and 8 kHz and above slighly more treble. But there was so much talk by the headphone show amateurs that the midrange is more natural and all of that. How so if it is identical to the Harman2019IEv2 target? Furthermore, if it were so relevant, Sean Olive would have included it in the study.
Last edited:
He did include the Soundguys target which as @markanini has nicely shown is similar to the JM-1 target (but with a different bass boost than you have above)? And that target matched Harman for preference (slightly beating it) in that recent study. For me this shows that Harman IE2019 needs tweaking which was the discussion all along.
I would love to see a future headphone DAC with a tilt adjustment 'dial' as well as Treble and Bass adjustments.
For me this shows that there an area of the FR range where preference is pretty flat for a large group of users: an easy to measure (significant?) difference does not cause large changes in preference. More than tweaking Harman IE2019, developing this understanding further is what is needed.
The target may eventually look like a fat line or a area…
- Thread Starter
- #10
It seems you are comparing a targets intended for different models of measurement rig. Reposting a graph by Joel form his post:
Has the meta on IEMs moved away from Harman's "pinna gain"?
That means you don't like JM-1 with 10dB tilt. Nothing strange, JM-1 is just a starting point to add your preference adjustments on top of. You are right I didn't like it in the past but I think there are some inconsistencies that I just encounter, I tried with the Zero:2 and like the result...
www.audiosciencereview.com
@Jeromeof is right, Sean Olive did include a JM-1 adjacent target in controlled listening tests, Soundguys target. It scored statistically equal to Harman 2019 in listener preference.
Last edited:
True, I mixed up measurements on different rigs, my bad. But still, as the inventor of it says himself, JM-1 is not a target and it also does not resolve the elevated ear-gain region. I think it is just better using the standard Harman2019IEv2 and lower the ear gain between 2 and 6 kHz by a few dB. What has been said many times by different persons and really and seems to be true is that the balance between bass and treble is important. The soundguys target has too much treble for the corresponding bass level (or too less bass for the corresponding treble level). It is a pitty that Sean Olive did not include a target in his study remedying this.It seems you are comparing a targets intended for different models of measurement rig. Reposting a graph by Joel form his post:
Has the meta on IEMs moved away from Harman's "pinna gain"?
That means you don't like JM-1 with 10dB tilt. Nothing strange, JM-1 is just a starting point to add your preference adjustments on top of. You are right I didn't like it in the past but I think there are some inconsistencies that I just encounter, I tried with the Zero:2 and like the result...
View attachment 400283
@Jeromeof is right, Sean Olive did include a JM-1 adjacent target in controlled listening tests, Soundguys target. It scored statistically equal to Harman 2019 in listener preference.
View attachment 400285
View attachment 400287
Last edited:
