Tanchjim Origin IEM Review

[staticV3]

While that would make for a nice demo, I am not trying to get perfect sound from these demos, nor is the author claiming that you will. It is clearly about the relative differences, as I said before. Which, you can hear, and are valid, because they are all done using the same equipment. And actually it comes across better than you are expecting.

May I suggest trying it the right way, then going back to the YouTube recording to see if it carries over?

 

[deadkrillin]

apparently the IE study didn't let them adjust treble preferences to begin with.

Sean olive mentioned a while ago that they are developing a new target for iem's. I was hoping this would clear out the 2019 IE curve once and for all and we'll know if the IEF curve is neutral for most listeners. however as amir mentioned its better to EQ to your taste starting with the measurements. Transducers and how they work with each other in tuning also plays a decent effect (~20% preference) on the preference imo but id like to see them study this. people who reviewed IE900 say its the best quality bass they've ever heard. Id like to hear that one.

Re: treble adjustment, there had to have been some adjustment since there is a difference between the 2016 and 2017 curves. Unfortunately we don’t know what exactly went into this, because the latter methodology wasn’t published. Frankly, it strikes me as a bit odd how little the absence of the methodology behind the curve is discussed as a legitimate issue with the output of the research, even in the “anti-Harman” circles.

Re: the new Harman IE target, if we’re putting in guesses, my guess is it will look a lot more like the equivalent Harman AE/OE curve on the 5128 (ie. take the 5128 into the IEC room and record the flattened Revel F208 with the -30/0/+30 angle average for a HATS measurement —> apply Harman AE/OE filters). This means on 711, it will likely have meaningfully more low-midrange… but maybe the same amount of bass, as the IE curve already has more than the OE curve in a roughly equivalent amount to the average difference between IEMs measured on the two systems.

 

[Chromatischism]

May I suggest trying it the right way, then going back to the YouTube recording to see if it carries over?

If I had the capability, I could. I don't know how.

And likewise, I would suggest trying this in the way that I did as well. You are implying that without modifying the sound the result is going to be pretty bad. If that were the case I would have been immediately turned off (and the measurements would have been off, as well, which is not the case--the microphone does not know the difference between a sweep and music). While YouTube audio has its flaws, the translation is better than you think. Good enough, anyway. It's listenable music. And since I'm starting from a neutral baseline I'm not getting additive effects of curves on top of curves (like if I used a stock IEM that had hotter treble). So it's 0+1 rather than 1+1, if that makes sense.

 

[amirm]

The video talks about how iem's bypass the pinna entirely and that makes it more inconsistent to the listener including the insertion depth at 8k vs headphones measurements where the treble is captured more consistently in the measurements vs what the listeners hear.

That is a theory in dire need of verification as far as impact on listener. The story of the fidelity of a headphone, speaker, and IEM is well told before 8 kHz. We don't even know if what the fixture measures at those frequencies relates to reality, marketing stories around 5128 fixture notwithstanding.

 

[MacClintock]

If I had the capability, I could. I don't know how.

And likewise, I would suggest trying this in the way that I did as well. You are implying that without modifying the sound the result is going to be pretty bad. If that were the case I would have been immediately turned off (and the measurements would have been off, as well, which is not the case). While YouTube audio has its flaws, the translation is better than you think. Good enough, anyway. It's listenable music. And since I'm starting from a neutral baseline I'm not getting additive effects of curves on top of curves (like if I used a stock IEM that had hotter treble). So it's 0+1 rather than 1+1, if that makes sense.

I don't get your problem. Just take one IEM of your possession with a reliable published frequency response and EQ it to the other IEM with a reliable published frequency response. That takes about 1 minute (if the data are available, but squig, HypeTheSonics, AutoEQ, oratory, ASR and Headphones.com should serve fine).
Do you also drive your car sitting in the backseat using sticks to press the pedals and some strings to move the steering wheel?

 

[staticV3]

If I had the capability, I could. I don't know how.

I'd be happy to help.

Could you please check if you own any IEM that's listed in Crinacle's 5128 database?

IEM Graph Comparison Tools | Hangout.Audio

The world's largest public database of IEM measurements.

crinacle.com

If the tool doesn't respond properly, try refreshing the window.
You need to pass Cloudflare for it to work properly, and sometimes the Cloudflare check doesn't load correctly upon first open. A refresh fixes that.

You can also check csglinux's 5128 database:

What are the best in-ear monitors?

IEM database

www.hypethesonics.com

If you have something from Moondrop or Truthear, that'd be ideal as those have shown excellent quality control so far.

 

[deadkrillin]

The story of the fidelity of a headphone, speaker, and IEM is well told before 8 kHz

I worry that this statement is implying something you don't mean, can you clarify? It almost reads like you're saying "because Harman's research—limited by the measurement instruments available at the time—was limited to a bandwidth up to 8-10kHz, that this is the only band of audio that matters for how a listener would qualitatively judge their experience with an audio device."

I think regardless of the veracity of the claims re: the 4620's—or the GRAS Hi-Res coupler's, for that matter—accuracy above 8-10kHz, I fear this is a dangerous oversimplification, no? Plenty of listeners could have an experience ruined by a nasty peak somewhere above 8kHz, even if the rest of the frequency response is heavenly (indeed, this is something that has happened to me numerous times). I assume I'm misunderstanding you here, and if so, apologies.

 

[amirm]

I worry that this statement is implying something you don't mean, can you clarify? It almost reads like you're saying "because Harman's research—limited by the measurement instruments available at the time—was limited to a bandwidth up to 8-10kHz, that this is the only band of audio that matters for how a listener would qualitatively judge their experience with an audio device."

My statement has nothing to do with Harman research. It comes from testing and equalizing hundreds of speakers, headphones and IEMs. Making adjustments at lower frequencies makes a big difference due to high energy in those bands in music. At higher frequencies, the differences are subtle and extremely content dependent. I routinely leave behind high frequency filters because their effect is too small and unreliable.

 

[MayaTlab]

n.b. what I'm referencing above:

View attachment 354433

from https://danishsoundcluster.dk/wp-content/uploads/2022/11/Olive_DSD_2022.pdf

Interestingly, the chart you posted shows the necessity of a sharper drop off below the peak, which mirrors my own findings in using EQ to level out my own personal resonances. It is a wonder why the Harman target does not echo this?

I think that the graph from Oksanen et al. doesn't show SPL variation at the eardrum for the test subjects, but the estimated transfer function between the inward facing microphone and the eardrum, based on a methodology proposed in the article and earlier ones from some of the same authors.

Just like Harman did a few years later, they used an inward facing microphone in the outlet port of an IEM :

You can visually represent it like this :

What they're doing is measuring the frequency response at the position of the inward facing microphone ("Mic" above). At low frequencies the SPL is going to be the same inside the "tube of air" that is formed when an IEM seals your ear canal whether it's at the driver, inward facing microphone, or eardrum position, so you can measure it directly. At higher frequencies, most importantly above 1kHz or so, that relationship breaks down : the SPL measured at the inward facing microphone no longer corresponds to the SPL at the driver or the eardrum. So have to apply transfer functions from one location to another (Gdm, Gme, Gde). Very annoyingly, that transfer function varies with fit / position / leakage or between individuals - and quite interestingly I've recently read a patent from Bose that suggests that the design of the IEM's front volume, eartip, inward facing microphone and protective mesh all play a role in determining exactly how that variation is expressed.

Oksanen at al have proposed a method to calculate the appropriate transfer function Gme for each of the test subjects, the result of these individualised transfer functions are in the graph above :

That method is a calculation, it is not a measurement nor a simulation.

@deadkrillin mentioned a more recent article from Samsung. This one is a simulation, after validating the simulation model in a 711. Figure 9 and 10 are closer in substance to the graph from Oksanen et al above, while figure 7 and 8 are simulations of the SPL at the eardrum itself. Important to note however that the values for the eardrum impedance in the simulation model for the human ear canals was kept a constant (while it varies across individuals) - so what you're mostly looking at here in terms of inter-individual variation between the human canals is variation in ear canals' shape, including volume, geometry and length -, and they used a different set of values than the ones more recently determined here.

I think this is an underrepresented viewpoint in many IEM reviews, unfortunately. Harman being the best publicly-available listener preference research we have doesn't mean we should ignore these very real sources of variability between listeners, these are simply questions that Harman's published IEM research—in their limited time and scope—didn't seek to answer. Unfortunately due to the IE research being unpublished for the most part, we don't have any information on how listener preferences vary in the area where anatomical/volume length variance would have the largest effects (treble).

Speaking further to the point of Harman's published work being scarce, I think it's unfortunate that Harman has a bit of a monopoly on the discussion around well-controlled and tested listener preference of insert earphones when firms like Apple probably have enough data in that regard to serve as a more statistically-significant foil to Harman's results. We'll just never see it posted or discussed publicly, unfortunately.


Indeed, and Sony, Sennheiser and Apple's IEMs all have something in common: The "target response" they aim for with their most popular TWS IEMs is meaningfully more elevated in the bass and low midrange than any of Harman's IEM targets dictates should be preferred.
Even Harman's own most-current premium TWS IEM offering (Galaxy Buds 2 Pro) deviates from the Harman IE target in this respect, shown below—normalized at a two-octave range centered at 800 Hz.

View attachment 354643

Of course, none of this changes that the results of the research are what they are, but it does kind of explain why there may be an ever-growing "anti-Harman contingent" across our community.

If the measured results of the 5128 are indeed more human-like (and there are a few papers—including this one from Samsung—that seem to indicate this may be the case) then listeners using the Harman IE Target could be getting a more treble-focused sound signature than they would with eg. Harman's AE/OE 2018 Target. This is a bit odd to me given Harman's first published IEM study indicated that listeners preferred the opposite—a more bass-focused sound signature in their IEMs compared to AE/OE headphones.

It has less to do specifically with them aiming at a different "target response", but rather some of these IEMs having a feedback system, which will, intentionally or not, correct the response in the range where it effectively operates (and occasionally introduce errors in the transition range above, but that is another subject).

Think of it as an error control loop : the system expects a specific signal, and anything (not just noise) that isn't the signal will be seen as an "error", including a change in frequency response, for example caused by leakage, or here in the case of the 5128, a larger equivalent volume at lower frequencies (which will result, as you've observed, in a lower response in that range). Not all feedback control loops are created equal though, and some IEMs will perform better than others in terms of delivering a stable SPL across individuals and fit.

This also means that, generally speaking, these IEMs will have an inconsistent transfer function between 711 and 5128 measurements (and in all likelihood your own ears) compared to passive ones.

The inconsistent Gdm, Gme and Gde transfer functions mentioned above is one of the reasons why until recently the feedback control loop was limited to operating at lower frequencies, but Bose's CustomTune technology seems to be a genuine breakthrough that allows them to operate it up to several kHz without (presumably) introducing significant errors in terms of FR and noise reduction.

apparently the IE study didn't let them adjust treble preferences to begin with.

In an unpublished study some listeners were able to adjust both the bass and treble response :

E-library page - AES

AES E-Library ← Back to search

www.aes.org

But if the Q / frequency / magnitude of the filters they could play with was similar to the earlier articles, I think that no combination available would be able to compensate for how 711 couplers tend to deviate from the average ear canal impedance.

That is, find measurements of both the target IEM and your listening IEM (ideally 5128), calculate the response difference, and apply the difference as EQ to your music.

With active IEMs in particular I'd recommend being a bit careful about doing that, or at least not take it too much at face value. A simple example : this is the difference between the Truthear Zero Red, Anker A40 (ANC on), and the Bose QC Ultra, measured in the regular canal extension of a 711 coupler (compensated flat), and one longer by 5,1mm, normalised at 500Hz :

To be noted is that the CustomTune calibration was performed before running the sweeps in each canal extension. I've greyed out the 5kHz+ region because I'd prefer people to focus on what is happening lower and other variables pile up in that range.

If, let's say, the A40 and QC Ultra were EQed to the same target based on the measurements done in the regular canal extension, ceteris paribus, someone with longer ear canals would genuinely get an audibly different SPL at the eardrum. And we're only talking about canal length here, not other variables.

And what is Bose doing here ? Something I think quite clever and deliberate.

In future I expect there will be a single target as Amir and many of us hope for, but there will be a methodology for compensating for individual in-ear differences.

I rather expect more and more active IEMs to individualise their output. A company like Bose already is, in regards to their IEMs, in a post "single target on a single ear simulator" world, and if you want to know whether or not they've been doing a good job, you'll probably need something in the vein of a set of different ear simulators representative of a larger population, know what the response ought to be for each on of them, and see whether or not their earbuds come closer to these individualised target(s) or not.

 

[Art of sound]

Re: treble adjustment, there had to have been some adjustment since there is a difference between the 2016 and 2017 curves. Unfortunately we don’t know what exactly went into this, because the latter methodology wasn’t published. Frankly, it strikes me as a bit odd how little the absence of the methodology behind the curve is discussed as a legitimate issue with the output of the research, even in the “anti-Harman” circles.

Re: the new Harman IE target, if we’re putting in guesses, my guess is it will look a lot more like the equivalent Harman AE/OE curve on the 5128 (ie. take the 5128 into the IEC room and record the flattened Revel F208 with the -30/0/+30 angle average for a HATS measurement —> apply Harman AE/OE filters). This means on 711, it will likely have meaningfully more low-midrange… but maybe the same amount of bass, as the IE curve already has more than the OE curve in a roughly equivalent amount to the average difference between IEMs measured on the two systems.

interesting. The way its portrayed here is that the users were allowed to adjust below the ridiculous 12db or so treble boost. Also, my understanding is that most users prefer different bass levels +3 to + 6 db below 200Hz and very few prefer a variation of slightly lower 5-6k i.e 1-2db below OE 2018 which is to say a very good curve for adherence and brilliant research. In iem's its all over the place if you look at individual preference curves of those that publish their curves.

 

[Chromatischism]

I'd be happy to help.

Could you please check if you own any IEM that's listed in Crinacle's 5128 database?

IEM Graph Comparison Tools | Hangout.Audio

The world's largest public database of IEM measurements.

crinacle.com

If the tool doesn't respond properly, try refreshing the window.
You need to pass Cloudflare for it to work properly, and sometimes the Cloudflare check doesn't load correctly upon first open. A refresh fixes that.

You can also check csglinux's 5128 database:

What are the best in-ear monitors?

IEM database

www.hypethesonics.com

If you have something from Moondrop or Truthear, that'd be ideal as those have shown excellent quality control so far.

Yes, I'm running the Truthear Zero.

Also, it's worth mentioning that when I listened to a recording of the Zero with my equalized Zero, it sounded like an unequalized Zero, which is exactly what you'd expect if I'm using a neutral IEM. The sound was instantly recognizable.

 

[staticV3]

Yes, I'm running the Truthear Zero.

Attached below is an EQ preset that'll turn your Truthear Zero Blue into a 7Hz Zero 2, tonally speaking:

 

[Chromatischism]

Attached below is an EQ preset that'll turn your Truthear Zero Blue into a 7Hz Zero 2, tonally speaking:
View attachment 354840

Thanks. While I could use this to "evaluate" the 7Hz Zero 2 (and others using the same method) in isolation using my own music, can you quickly switch to another EQ?

With my neutralized Truthear Zero I can quickly switch between the same program material from various other IEMs, and it translates—including listening to recordings of the stock Zero, which sounds like a stock Zero. This wouldn't work if I were not using my EQ, because then you'd be stacking the two stock curves on top of each other. So it's proof that this setup isn't as bad as you think.

 

[MacClintock]

Thanks. While I could use this to "evaluate" the 7Hz Zero 2 (and others using the same method) in isolation using my own music, can you quickly switch to another EQ?

With my neutralized Truthear Zero I can quickly switch between the same program material from various other IEMs, and it translates—including listening to recordings of the stock Zero, which sounds like a stock Zero. This wouldn't work if I were not using my EQ, because then you'd be stacking the two stock curves on top of each other.

Sure you can, at least I can. I use SoundSource on a Mac. Having multiple EQ settings uploaded (for example mimicking different other IEMs), you can switch smoothly between them while listening to the same music.

 

[taotone]

Indeed, and Sony, Sennheiser and Apple's IEMs all have something in common: The "target response" they aim for with their most popular TWS IEMs is meaningfully more elevated in the bass and low midrange than any of Harman's IEM targets dictates should be preferred.

A standardised parameter that somehow gets taken for granted and very infrequently mentioned is the SPL of the playback device. I'm not sure what this is normalised to?

This might well mean that standard playback volume (e.g. at 80% max on most devices) means that SPL is below the standard Harman tested level - necessitating an additional "loudness" adjustment as the default.

In Europe bureaucratic madness in Brussels has dictated that no music should ever be played loud enough so that some idiot somewhere might be able to damage their hearing if they left it on 24/7. Sony (perhaps having been the victim of a hefty fine at some point) is particularly subject to making sure that devices shipped to the EU (alas still including the UK) cannot play loud enough for my taste. ANC helps to make their TW range marginally usable.

 

[taotone]

I rather expect more and more active IEMs to individualise their output. A company like Bose already is, in regards to their IEMs, in a post "single target on a single ear simulator" world, and if you want to know whether or not they've been doing a good job, you'll probably need something in the vein of a set of different ear simulators representative of a larger population, know what the response ought to be for each on of them, and see whether or not their earbuds come closer to these individualised target(s) or not.

Agreed. Also I don't understand the argument for having a single target curve at all. Dr Olive's recent Canjam talk and followup (easily found on Youtube) make it clear that there are 3 main clusters from factor analysis in his research. The talks are very well worth while.

I do agree that it would be nice for manufacturers to aim at publishing their own measurement curves for an IEM relative to Harmon as that would allow baselining against a known good standard to make it easier to decide if one is likely to enjoy a particular tuning.

But these are consumer goods, and at the very least one might accept a baked in Bass, Treble and 3Khz adjustment from a particular IEM because those will line up with individual users' in-ear anatomy and preferences across all music. I am quite sensitive to the 3KHz region and have chosen speakers that have a "BBC dip" for that reason. Do I really want a -2dB adjustment for ALL music at 3KHz? Actually yes I do, otherwise that region in general and female vocals in particular are unbearable to me, and that preference follows through into headphones and IEMs too. The preference for boosted bass, boosted treble and a compensating 3Khz dip is a minority preference but certainly not very uncommon. This is very much the FR of the Sennheiser IE900 for example.

A lot of consumer goods have extensive research on consumer preferences and make a lot of effort to make sure they address a wide range of preferences. Imagine research showing that 60% of consumers preferred vanilla ice-cream, 20% chocolate and 10% strawberry, with a margin who liked other flavours. Would one immediately leap to the conclusion that vanilla ice cream should be the target going forward, that people who didn't love vanilla were objectively wrong? That the correct approach for all ice cream production should be vanilla with drizzle-on sauces for those deviants who didn't like vanilla? That ice cream should be recommended on the basis only of it being vanilla and that the more it deviates from vanilla to gain a grudging recommendation only if it could have sauce added to make it taste more like vanilla?

 

[Blorg]

In Europe bureaucratic madness in Brussels has dictated that no music should ever be played loud enough so that some idiot somewhere might be able to damage their hearing if they left it on 24/7. Sony (perhaps having been the victim of a hefty fine at some point) is particularly subject to making sure that devices shipped to the EU (alas still including the UK) cannot play loud enough for my taste. ANC helps to make their TW range marginally usable.

I get this is an issue for Sony DAPs where they don't know and as a result these devices are kneecapped with less sensitive IEMs. Less of an issue on a TWS where Sony control both the amp and the transducer and so can know the actual SPL.

I believe the EU limit is meant to be 100dB after you click up past the warning at 85dB. I get 95dB on my SPL meter with EU-limited Samsung Galaxy Buds 2 Pro, and I don't go beyond about 2/3 up with them.

I also have the Sony XM4, but not EU limited. No WAY could I listen to these anywhere near that, I don't go above 50% with them. A kid listening to these at 100% will be wrecking their hearing, I can understand why the EU did this.

This is with a 1kHz sine tone, holding the bud at the end of the meter with the tip inverted. This meter is A-weighted, so it will discount bass. Max volume on my phone, no EQ in the Sony app. This is very back of an envelope, I'm not trying to hold this out as some absolutely correct measurement, but it is probably indicative of relative levels.

I can hear the Sony buds pretty loud with them sitting on the desk at this volume. Can't imagine putting them in my ears.

You do need to bear in mind that music is usually significantly boosted in the bass, 1kHz will not be at 0dB in an actual recording, so it's not 113dB average A-weighted with music into your ears. But I would venture the EU did not set these limits too high and listening to these at 100% volume is almost certainly still doing you damage. The upper 100dB limit is already taking into account lower recording levels, bass weighting, etc, NIOSH has a 15 minute limit for 100dB, for example. The all-day occupational limits are A-weighted and much lower than this (87dB is a hard limit in the EU, with employers mandated to provide hearing protection over 80dB).

Incidentally, both Sony and Apple change the frequency response curve depending on volume, their TWS don't have one fixed curve. Samsung, it's the same FR response regardless of volume.

 

[taotone]

I get this is an issue for Sony DAPs where they don't know and as a result these devices are kneecapped with less sensitive IEMs. Less of an issue on a TWS where Sony control both the amp and the transducer and so can know the actual SPL.

I believe the EU limit is meant to be 100dB after you click up past the warning at 85dB. I get 95dB on my SPL meter with EU-limited Samsung Galaxy Buds 2 Pro, and I don't go beyond about 2/3 up with them.

I also have the Sony XM4, but not EU limited. No WAY could I listen to these anywhere near that, I don't go above 50% with them. A kid listening to these at 100% will be wrecking their hearing, I can understand why the EU did this.
View attachment 357589

This is with a 1kHz sine tone, holding the bud at the end of the meter with the tip inverted. This meter is A-weighted, so it will discount bass. Max volume on my phone, no EQ in the Sony app. This is very back of an envelope, I'm not trying to hold this out as some absolutely correct measurement, but it is probably indicative of relative levels.

I can hear the Sony buds pretty loud with them sitting on the desk at this volume. Can't imagine putting them in my ears.

You do need to bear in mind that music is usually significantly boosted in the bass, 1kHz will not be at 0dB in an actual recording, so it's not 113dB average A-weighted with music into your ears. But I would venture the EU did not set these limits too high and listening to these at 100% volume is almost certainly still doing you damage. The upper 100dB limit is already taking into account lower recording levels, bass weighting, etc, NIOSH has a 15 minute limit for 100dB, for example. The all-day occupational limits are A-weighted and much lower than this (87dB is a hard limit in the EU, with employers mandated to provide hearing protection over 80dB).

Incidentally, both Sony and Apple change the frequency response curve depending on volume, their TWS don't have one fixed curve. Samsung, it's the same FR response regardless of volume.

Hmm, I have the XM4 EU version and they are soft on public transport at 100% on my Samsung Galaxy. I would guess an equivalent loudness of around 65dB comparison to sitting at home in my living room on my hifi. I don't have a sound meter like yours.

They do seem to have gotten softer, so perhaps they are just not working very well and I need to replace them. (It's not my hearing.) As for the Sony DAPS unfortunately it makes them unusable for me. My XDuuo or Fiio devices can play way louder than I need on IEMs, and drive most headphones just fine.

On the London Underground ANC and passive isolation is extremely useful as some lines can have painfully high peak background noise - actually can be quite fatiguing - ANC makes quite a difference to the ability to complete a 1-hour commute without being exhausted.

 

[MacClintock]

A lot of consumer goods have extensive research on consumer preferences and make a lot of effort to make sure they address a wide range of preferences. Imagine research showing that 60% of consumers preferred vanilla ice-cream, 20% chocolate and 10% strawberry, with a margin who liked other flavours. Would one immediately leap to the conclusion that vanilla ice cream should be the target going forward, that people who didn't love vanilla were objectively wrong? That the correct approach for all ice cream production should be vanilla with drizzle-on sauces for those deviants who didn't like vanilla? That ice cream should be recommended on the basis only of it being vanilla and that the more it deviates from vanilla to gain a grudging recommendation only if it could have sauce added to make it taste more like vanilla?

True, there exists segmentation, as was well elaborated on by the Harman research. But, the difference is just in a litte bass boost or reduction or a little more or less treble, all with a shelf. So the analogy to ice cream does not work at all, if you could transform vanilla ice cream by some filters (EQ) readily into chocolate, than it would be fine, but you can´t.
What the research has shown is that NOBODY, in numbers 0 persons, likes huge resonances or swings in the frequency response and the like, as still offered by a lot of headphones. The differences in preferred target are small (a few dB) and exclude crazy tunings like many even expensive headphones still have, for example from Meze or ZMF. Also for example the SoundGuys target, which came out quite favourably in the last Sean Olive presentation, is not far off their own Harman targets and has in principle the same shape.