Earfonia IEM Measurement Setup & Methodology

Earfonia · Jan 2, 2022

One of the main objectives of In-Ear Monitor measurement (as well as other audio equipment measurements) is to understand the correlation of the perceived sound quality with the numbers and graphs. I’ve been doing IEM measurement for a while now, just as a hobby, to learn how to correlate what I hear and what is shown on the measurement results. I plan to share my IEM measurements with the community in the form of IEM reviews.

My objective in sharing measurement results is not to compare or verify the factory specifications as my measurement setup and equipment may not be good enough for it. The objective is as mentioned above to learn together with the community to get an understanding of the correlation of the perceived sound quality with the measurement result. Sharing measurement results is something that has to be done with care. There are so many things that could go wrong in measurement and sharing measurement results that were done without proper process and methodology is potentially misleading. Therefore, I try to be as transparent as possible as to how I obtain the measurement result. This post serves as a placeholder for information regarding my measurement equipment, setup, methodology, approach, and rating criteria in IEM measurement.

“Knowledge Is Not Understanding”
– Howard Pollick

Sections:

Measurement Setup

• Software
• Equipment
• Calibrations
• Settings & Conditions
• Measurement

Sound Quality

• Tonal Balance
• Earfonia IEM Target Curve - EITC-2021
• Frequency Response Analysis
• Perceived Tonal Balance
• Other Sonic Qualities - Liveliness

Engineering Quality

• Left-Right Match
• THD @ 94 dB SPL and 104 dB SPL
• IMD SMPTE and DIN
• Electrical Impedance
• Sensitivity
• Fit, Comfort, & Build Quality
• Metal shell to ground pin connection

Standards & References

Measurement Results (Examples)

• E610A-SN20229 and GRAS RA0045-S1
• EITC-2021 and Etymotic ER4XR, ER2XR, ER2SE

------------------------------------------------------------

Measurement Setup

Software:

REW on Windows 10 PC.

Equipment:

IEC 60318-4 occluded-ear simulator - Model: E610A SN:20229, consisting of:
- IEC 60318-4 Compliant Coupler
- 1/2" Prepolarized Pressure-field Microphone
- CCLD / ICP Microphone Preamplifier
B&K 1704-C-102 - ICP / CCLD Signal Conditioner / Amplifier
RME Babyface Pro FS - USB Audio Interface
ND9B SPL Calibrator
DIY noise isolation container made from Pelican 1500 case

Calibrations:

1. REW Soundcard calibration:
Setup:
RME Babyface Pro FS 3.5mm HP OUT (Left Ch) --> B&K 1704 IN1 > B&K 1704 OUT1 --> RME Babyface Pro FS XLR IN1

RME Babyface Pro FS 3.5mm HP OUT Analog 3(L) & 4(R) volume level: 0 dB
B&K 1704-C-102 gain: 0 dB (x1)
B&K 1704-C-102 filter: Linear
RME Babyface Pro FS XLR Input 1 gain: 6 dB
Calibrated to flat frequency response.

2. REW Microphone calibration:
IEC 60318-4 compliant coupler, E610A SN:20229
Calibrated to GRAS RA0045-S1.

3. REW SPL calibration:
ND9B SPL Calibrator set at 94 dB SPL (1 kHz)
REW SPL Meter reading calibrated to 93.6 dB SPL
Setup Dynamic Range: 135 dB SPL

Notes on SPL calibration:
Based on SPL measurement comparisons with Etymotic ER4XR, the ND9B unit that I use seems to output 94.5 dB SPL instead of exactly 94.0 dB SPL. With 0.9 dB correction factor based on measurement of the IEC 60318-4 Coupler SPL attenuation, the SPL calibration of my setup is set at: 94.5 dB SPL - 0.9 dB = 93.6 dB SPL.

4. Calibrations for Electrical Impedance measurement:
Calibrations as per REW impedance measurement recommendations:
REW Impedance Measurement

Open circuit calibration, Short circuit calibration, and Reference resistor calibration (16 ohms).
Level / Headroom calibration:
- 3 ohms load: 33.3 dB Headroom
- 600 ohms load: 6.2 dB Headroom

Settings & Conditions:

General setting & conditions for frequency response and distortion measurements:
Ambient noise inside the noise isolation container: < 28 dBA SPL (measured)
Ear Tip: Stock medium size silicone ear tip is the default ear tip for measurement. If for any reason the stock ear tip is not available, other suitable silicone ear tip will be used as the replacement.
Headphone amplifier output impedance: 0.15 ohms (measured)
Audio driver and sample rate: ASIO – 96 kHz
IEC 60268-7 test level for most tests (except THD measurement @ 104 dB SPL): 94 dB SPL @ 500 Hz
Result: RAW (uncompensated)

RME Babyface Pro FS 3.5mm HP OUT Analog 3(L) & 4(R) volume level: 0 dB
(signal level set on REW generator)
B&K 1704-C-102 gain: 0 dB (x1)
B&K 1704-C-102 filter: Linear
RME Babyface Pro FS XLR Input 1 gain: 6 dB

Setting & conditions for Electrical Impedance measurement:
Test setup as per REW article for impedance measurement:
REW Impedance Measurement

Rsense: 75 ohms
Audio driver and sample rate: Java – 96 kHz

Headphone Output:
RME Babyface Pro FS 3.5mm HP OUT Analog 3(L) & 4(R) volume level: 0 dB
Signal level set on REW generator: -18.0 dBFS (approximately: 216 mV)

Soundcard Inputs:
RME Babyface Pro FS TS Input 3(L) & 4(R) gain: 9 dB (Windows driver input level set to 100%)

Measurement:

Frequency Response

Method: Continuous Sweep
Measurement Frequency Range: 16 Hz - 22 kHz
Observation Frequency Range: 20 Hz - 10 kHz
Level: 94 dB SPL @ 500 Hz
Sweep Length: 1 Million Samples
Repetitions: 4x
Smoothing: 1/48th of an octave smoothing

Left-Right Match

Range: 20 Hz - 7 kHz
(Data from frequency response measurement result)

Total Harmonic Distortion

Method: Stepped Frequency Sweep
FFT Length: 128k
Windows: Blackman-Harris 7
Average: 4
Max overlap: 93.75%
Frequency Range: 20 Hz - 20 kHz
Frequency Step (PPO): 6
Lock Frequency to the RTA FFT: enabled
Add Dither: enabled (24 bits)
Level: 94 and 104 dB SPL @ 500 Hz
Observation Frequency Range: 55 Hz – 7.1 kHz (THD through 6th harmonic)
Measurement Points: 43

Intermodulation Distortion (IMD)

SMPTE & DIN
IMD measured at 94 dB SPL @ 500 Hz

Electrical Impedance

Method: Continuous Sweep
Measurement Frequency Range: 20 Hz - 20 kHz
Level: -18.0 dBFS (approximately: 0.216 Vrms)
Sweep Length: 1 Million Samples
Repetitions: 1x
Noise Filter: Medium
Observation Frequency Range: 20 Hz - 20 kHz
Resolution: 96 PPO (points per octave)
Smoothing: 1/48th of an octave smoothing
Measurement Points: 957 points

Sensitivity

30 seconds of Loudness Equivalent (LAeq) of 50mV full range Pink noise playback. The 50mV level is measured using 500Hz tone.
Expressions: ... dBA SPL / 50mV of Pink noise (50mV set at 500Hz)

Metal shell to ground pin connection

Earfonia · Jan 2, 2022

Sound Quality

Tonal Balance

Earfonia IEM Target Curve - EITC-2021

As we know frequency response target curve for In-Ear Monitor (IEM) is not a flat line as commonly expected for speakers. This is due to what is called the head-related transfer function (HRTF). There are several target curves for IEM frequency response that are being used in the industry, but none has been considered as a common international standard/consensus for the In-Ear Monitor frequency response target curve. The Harman Target Curve for IEM 2019 is probably one of the most popular among those target curves and the one that is probably the closest to my target curve.

The primary reason for the lack of consensus for the IEM frequency response target curve could probably be the consensus of the methodology of how the target curve should be measured. Besides that, HRTF varies from person to person. Another reason is the fact that frequency response measurement results of an IEM may differ between different measurement equipment. Because of those reasons I have come up with my own target curve that suits my listening preference and is compatible with my measurement equipment. This target curve is subjectively my ‘perceived neutral’ target.

My approach is simple, I look for IEMs with a tonal balance that is close to my perceived neutral. Perceived neutrality is highly dependent on the recording material we use for observation. Besides ‘audiophile recordings’ I also use my own recordings raw tracks for the observation (mostly piano and vocals). The selection process took more than 2 years. Initially, there were around 10 IEMs selected, but in the end, I chose 5 of them. After that, I measured those IEMs and created the average frequency response from the measurements and reconstructed the estimated target curve to create the ‘smooth’ version of the target curve. I’ve made several drafts and finally, I’m quite satisfied with the latest draft that I think is good enough to be my target reference for IEM frequency response analysis. I named it ‘Earfonia IEM Target Curve 2021’ or ‘EITC-2021’ in short.

None of the 5 IEMs are perfectly matched to my target, but all have a high degree of naturalness to my ears. Obviously, they don’t sound the same, one may sound a bit more mid centric and other may sound a bit more V shape in tonality. What makes them unique to me are they are within a close range to my target. So, the average frequency response of the 5 IEMs is the base for Earfornia IEM Target Curve.

The 5 selected IEMs are:
1. A prototype IEM given to me by a local IEM brand for testing. I cannot disclose any further detail about it. It just sounds right to my ears, balanced and pretty close to my perceived neutral. Just for referencing, I named it STP.
2. SeeAudio Yume. Perceived tonal balance is pretty neutral to me and at the time of writing, it is among the top 5 IEMs closest to Harman IE Target 2019 in the HypeTheSonics database. Mids may sometimes sound a tad lean but overall, it is a good neutral reference IMHO.

What are the best in-ear monitors?

IEM database

www.hypethesonics.com

3. IKKO OH1. Mildly V shape tuning. It just sounds right with practically all recordings I played on it. Lively and engaging. One of my all-time favourite IEM.
4. TFZ King Pro. Mildly V shape tuning. Often perceived as neutral with classical, acoustic instrumental, and orchestral works. One of my top favourite IEM for classical recordings. Represent the maximum V shape tuning that for certain recording can still be perceived as neutral.
5. Moondrop KXXS. Neutral with a tad mid centric tonality. The mids of KXXS sounds perfectly neutral to my ears. I prefer to have a tad more sub bass and upper treble presence from KXXS. KXXS is a good IEM to represent my acceptable limit of a slightly mid centric tuning.

I applied a custom weighting for averaging the graphs. I doubled the weighting for STP and SeeAudio Yume as they represent the more balanced tuning among the 5. Then I averaged the 7 frequency response graphs. The following is the (custom weighted) averaged measurement curve from the selected 5 IEMs (Red curve) and the EITC-2021 smoothed reconstructed curve (Blue curve):

EITC-2021 is more suitable as a target curve for IEMs designed for regular insertion depth (~ 20mm from eardrum / IEC coupler microphone membrane). For the deep insertion (< 16mm from eardrum / IEC coupler microphone membrane) IEMs such as the Etymotic ER series and some micro drivers IEMs the treble dip and resonance peak frequencies will be higher.

EITC-2021 is not meant to be a static target curve, but it is more of an estimated target of a dynamic target. A dynamic target as the value of frequency peaks and dips are not static values but a range of preferred values. For example, the midrange peak shown in the graph is at 2.7 kHz, but it is actually a range from 2.3 kHz - 3.1 kHz. So, a midrange peak measured in that range is considered acceptable and within the range. For the same reason, I prefer to present all frequency response graphs in raw without any compensation. I think there is still a lot to learn about how to meaningfully interpret IEM frequency response graphs, therefore simple frequency response compensation to a static target curve potentially might cause misinterpretation. I will elaborate more on my preferred way of dynamic interpretation of IEM frequency response graphs in the frequency response analysis section.

My observations so far from my own measurement:
● There are IEMs that sound very good to my ears (not necessarily close to perceived neutral) with FR graphs that are not in agreement with the EITC-2021 target curve.
● So far there is no IEM with FR graph that is closely matched to EITC-2021 that sounds unbalanced in the tonality.

For the reasons above, the rating for Tonal Balance will be a combination of objective comparison to EITC-2021 and my personal subjective listening test. EITC-2021 is a work in progress. I might revise it in the future if necessary. It is not a perfect target; therefore, the listening test is still my highest priority in sound quality evaluation. But the target curve is a necessity for objectivity.

Comparison to Harman IE Target Curve - 2019

I know that I shouldn’t compare The Harman Target Curve for IEM 2019 with my EITC-2021 because of different measurement equipment and methodology. But some (including myself) might be curious about how they compare so here is the comparison:

The preferred bass level is surprisingly very close. Bass is a personal preference, so there won't be any consensus on how much should be the exact bass level for a target. My personal preference for bass is more or less in agreement with the Harman Target. The higher level at around the 60-400 Hz area could probably be due to the different measurement equipment or my personal preference for a slightly fuller sounding midrange. FYI I’m not a fan of a recessed-sounding midrange.

The biggest difference between the 2 target curves is on the treble area around 5 - 10 kHz. We ignore the area above 10 kHz because it is beyond the accuracy range of the IEC 60318-4 Ear Simulator. My guess the difference is due to the different method of formulating the target curve. We can read the measurement methodology of Harman Target Curve from the following document:
PERCEPTION & MEASUREMENT OF HEADPHONE SOUND QUALITY: IS THERE A PREFERRED TARGET RESPONSE?

From what I read the Harman Target Curve for IEM (IE Headphones) seems to be an estimated target curve based on the listeners’ preference and may not be an estimated average response from real IEM measurements. That's why the treble response between 5-10 kHz doesn’t seem to be a realistic target for measurement since it has smoothed out (ignored) the treble dip and the resonance peak that show in most IEMs measurement using the standard IEC 60318-4 Ear Simulator.

The resonance peak frequency depends on the insertion depth of the IEM to the IEC 60318-4 coupler, and it is usually in the range of 8 - 11 kHz due to the λ/2 distance between the IEM and IEC coupler mic membrane. And the treble dip follows right before the resonance peak, usually between 6.5 - 9 kHz, most often at around 7 kHz. I observe that the treble dip is a good indicator of treble sharpness and sibilance, therefore, to ignore it makes the treble part of the Harman Target Curve an unrealistic target for measurement. From my observation, my preferred level of the treble dip is at around the same level as the midrange dip with ±3 dB of tolerance. While on the Harman Target Curve that area shows a much higher level than the midrange dip. Meaning if my observation is correct, if an IEM frequency response at around 7kHz area matches the Harman Target Curve, it will have an unnatural sharp treble that is prone to sibilance. But again, my observation is based on different measurement equipment, so my conclusion here may not be accurate. And that’s also one good reason why I cannot use Harman Target Curve for my IEM measurement (even after my best effort to calibrate my measurement setup to G.R.A.S. RA0045 Ear Simulator, the same model of IEC coupler that was used to measure the Harman Target Curve) because the treble area is proven to be an unrealistic target for my measurement setup.

Frequency Response Analysis

The objective of analyzing the IEM frequency response graph is to estimate the perceived tonal balance of the IEM. As mentioned earlier EITC-2021 is a dynamic target. There are several observation points on the target curve, and the value of the frequency and level of the observation points are a preferred range of value rather than a static value. The suggested observation points on the target curve are as the following:

Observation Points:

B60: 60 Hz is at the border between sub-bass and bass area. Although the sub-bass level is sometimes measured higher than the 60 Hz level, our hearing sensitivity for frequency below 60 Hz is lower than for frequency above 60 Hz. Therefore, I think 60 Hz is a good point to observe the overall perceived bass level. Bass level is a personal preference, my preference is B60 (bass level @ 60 Hz) at around 2.5 dB (±3 dB) below the Midrange Peak.

MRP: Measurement Reference Point, 500 Hz at 94 dB SPL. The loudness reference for setting the output level of the headphone output. Although frequency response measurement is less affected by the loudness level, this level is critical for distortion measurement.

Md: I didn’t set any preference for the Midrange Dip frequency, typically midrange lowest dip point can be between 200 - 900 Hz. Midrange Peak and Treble Dip level are measured in reference to the midrange dip level.

Mp: Midrange Peak is one of the most important measurement points. This is the natural ear resonance of the ear canal and eardrum and is responsible for the perceived midrange presence. The very important peak for vocal presentation. From several articles that I read; this primary peak of ear resonance is approximately +17 dB at 2,700 Hertz. My preferred frequency range for Mp is between 2300 - 3100 Hz, with a level around +10 dB ±3 dB above the level of Midrange Dip. I’m not sure why my personal preference for the level of the primary resonance peak is far lower from +17 dB. I found that beyond +13 dB vocal may sound a bit shouty and too in your face kind of presentation.

T1p: Sometimes called the secondary peak, this is the resonance frequency of the concha of the ear. I call it the 1st Treble Peak. T1p is sometimes not shown on some IEM frequency graphs, or less emphasized. I found that T1p is important for perceived clarity, therefore I prefer the T1p to have a certain level of emphasis. My preferred frequency range for T1p is between 4000 - 5000 Hz, with level around 1.5 dB (±3 dB) below the Midrange Peak.

Td: I observed that Treble Dip is a good indicator for treble sharpness or the level of sibilance. Therefore, I call it Sibilance Dip. I don’t have any preference for Td frequency, but it is typically around 7kHz. It gets higher with deeper insertion IEMs. My preferred level for Td is around 0 dB (±3 dB) from the Midrange Dip. So more or less on the same level as the Midrange Dip.

T2p: When the ear canal is closed by an IEM, it acts as a closed tube. And the length of the closed tube is the half lambda for the 1st resonance frequency of the tube. That’s the reason that this 2nd treble peak indicates the insertion depth of the IEM into the IEC coupler because when an IEM is inserted into the IEC coupler it creates a kind of closed tube with the length around the half lambda of the T2p frequency. The deeper the IEM inserted into the coupler, the higher the resonance frequency because the length of the tube is shorter. I don’t use this point as a point for tonal balance observation because the frequency and level are highly influenced by the insertion depth of the IEM into the IEC coupler, therefore it is not a consistent observation point for tonal balance analysis. I use this peak as a measurement point to observe the insertion depth of the IEM, and during measurement, I try to match this resonance peak between the left and the right channels for a more consistent insertion depth between the 2 channels. I usually insert the left channel first into the coupler with a moderate insertion force, until the IEM is inserted properly and won’t fall out by itself from the coupler. After the left channel measurement, then when measuring the right channel, I adjust the insertion depth until the T2p frequency of the right channel matches the T2p frequency of the left channel. This process of matching the T2p of the right channel sometimes takes more than 15 minutes. I use a mounting putty to keep the IEM stable in its place. Otherwise, it is very difficult to keep the insertion depth consistent during measurement. Although T2p frequency varies with insertion depth of the IEM into the coupler, the level could give us some indication of the perceived treble level around this 8-10 kHz area. I cannot set an accurate target level for T2p, but I prefer the T2p level to be around the same level as the T1p. When the T2p shoots up higher than T1p usually we can hear some emphasis on treble sharpness on that 8-10 kHz area.

The following are my steps to analyze the IEM frequency response graph:

1. Midrange dip (Md) normalization. For measurement we normalize the SPL level at the 94 dB SPL at 500 Hz. But for frequency response analysis I propose to normalize the midrange dip to the same level of the target curve midrange dip, which in this case also at 94 dB SPL.

2. Midrange peak (Mp) level calculation. Mp level is calculated in reference to the level of midrange dip (Md):
Mp - Md ≈ 10 dB ±3 dB.

3. B60, T1p, Td, & T2p level calculation. Bass (B60) and the 1st Treble peak levels are to be compared to the Midrange peak and not the midrange dip. This is probably the key difference between my method and the usual way of frequency response analysis. For example, we often see the bass level in reference to the midrange dip, but in my observation, it is more important to see the bass level in reference to the midrange peak and treble peak. The reason is the bass level reading from the midrange dip alone is not very meaningful without the relation to the midrange peak and treble peak. We might see a bass level like 12 dB above the midrange dip and we might think that the IEM will sound warm and bassy. But if the midrange peak and treble peaks let’s say 15 dB above the midrange dip, the IEM will not sound bassy, but it could probably sound a little bright because Mp and T1p are higher than the bass level.
There is no exact target level for T2p but my preference is for the T2p level to be more or less equal to the T1p level.

The calculations in this step base on my preference:
B60 - Mp ≈ -2.5 dB ±3 dB
T1p - Mp ≈ -1.5 dB ±3 dB
Td - Md ≈ 0 dB ±3 dB
T2p ≈ T1p

Perceived Tonal Balance

The listening test is always the most important part of the IEM review process. Here is where I give the final rating for the overall Tonal Balance. Observation from the target curve observation points doesn’t tell us the whole story. There are other aspects of the Tonal Balance that are best described by words. I will try my best to be consistent with my target curve, but sometimes I have to deviate from it a little bit if my ears tell me otherwise. Below is the tonal balance rating guideline.

Other Sonic Qualities - Liveliness

Some said it is all in the frequency response measurement results. It is probably true, but other than the tonal balance, sonic qualities such as perceived detail, clarity, transient, attack, dynamic, tonal density, holographic spaciousness, etc. are hard to read from the frequency response graph. Human subjective interpretation of the overall sound quality I think is still essential in any review of audio equipment, especially transducers such as In-Ear Monitors, headphones, and speakers.

The objective for the ‘Other Sonic Quality’ evaluation is to share my subjective perception of the sonic qualities of the IEM that make the IEM sounds livelier and more lifelike. As mentioned earlier, some of those sonic qualities are perceived detail, clarity, transient, attack, dynamic, tonal density, and holographic spaciousness. It is quite subjective until we have a better understanding of how to quantify those qualities on measurement. Some may call it the ‘Technicalities’ aspect of the sound quality but I prefer not to use that term because ‘Technicalities’ may give the impression that it is an aspect that is objectively measurable and quantifiable. Well hopefully in the future it will be measurable and quantifiable, but for now, it is an aspect from the sound quality that can only be subjectively evaluated. My objective for this evaluation is the more lifelike and livelier the better. For the lack of a better word, I just call it ‘Liveliness’. It is an overall impression of liveliness by the sonic qualities mentioned above.

Rating is used to make simple comparisons between ‘The Perceived Liveliness’ quality of one IEM to the other. Rating is done by subjective listening test and comparison to other IEMs. Please be reminded that this rating is only a subjective rough estimation of the ‘Liveliness’ aspect of the sound quality, and the interpretation of ‘Liveliness’ may be different from person to person.

The following is my guideline for the sound quality rating criteria:

Earfonia · Jan 2, 2022

Engineering Quality

• Left-Right Match
• THD @ 94 dB SPL and 104 dB SPL
• IMD SMPTE and DIN
• Electrical Impedance
• Sensitivity
• Fit, Comfort, & Build Quality
• Metal shell to ground pin connection

Ratings for the engineering quality are based on my own measurement statistic data. That means the targets are all realistic values and practically achievable. It also factors in the limitation of my current measurement setup. But since most of the time we are limited to checking only one sample, this quality check doesn’t represent the overall manufacturing quality control of the factory.