It is difficult to find earphones that can serve as a reference transducer for evaluating loudspeakers. They are very useful. though, for evaluating the tonal qualities of a recording. Their high degree of spatial distortion and in-head localization renders them undesirable for me when I want to listen for enjoyment. But there are exceptions to this, like when trying to listen to something in a noisy environment or without disturbing others. There are 3 basic types of earphones, each with their set of benefits and problems.
What follows below was written several years ago.
When evaluating the accuracy of sound reproduction I rely on my own recordings and also on commercially available CDs. I have a mental memory and written notes from the live event that I recorded and I compare those to what I hear when I play back the recording over loudspeakers. There is no equalization applied to my own DAT recordings, unlike with almost all commercial CDs, and I use a very simple head-related microphone arrangement similar to a sphere microphone. When I purchase a CD I have no clear trail to any acoustic reference and I must rely on the judgments of the recording engineer, the mastering engineer and the producer for giving me a sonic experience that is traceable to familiar acoustic sounds. If there is a clear trace with a minimum of equalization and compression, then the recorded material can also be used to make judgments about the reproducing loudspeaker.
The problem lies in not knowing enough about the recording. If it sounds poorly over speakers, is it the recording's fault or the speaker's? To sort this out it is necessary to have a trustworthy reference sound reproducer. Headphones are often used, but have their own idiosyncrasies. They eliminate the effects of the listening room, but their frequency response is not necessarily flat or even smooth, and they heavily distort the distance of the auditory event. They also eliminate bone conducted sound and "chest pounding" bass. In the Seventies Russ Riley introduced me to the equalization of headphones. At that time he had found the Sennheiser HD 414 open-air, or supra-aural headphones to be sufficiently well behaved for equalization of their frequency response, unlike the then popular sealed, circum-aural phones. We used an audio signal generator that was smoothly tunable by hand to sweep the frequency range. It was not too difficult to hear where the headphone response had peaks, though we were not able to quantify their magnitudes. The next step was to build an equalizer that would notch the peaks and we continued adjusting it until we were satisfied with the smoothness of auditory response. The circuit and the voltage across the transducers ended up looking as follows when tuned for my ears:
Compared to typical loudspeaker response equalization these are huge changes in response. The results, though, when listening to program material were very convincing and easily demonstrated by switching the notch filters out of the circuit. We both had the same 2.8 kHz notch frequency for f1, but f2 was at 6.7 kHz for Russ. After further equalization to boost low frequencies, which was also sensitive to spacing of the transducer from the ear, some high frequency boost, and frequency dependent cross-feed to compensate for the unnaturalness of recordings with widely spaced microphones, Russ stayed mostly with headphone reproduction. I continued to pursue loudspeaker development because I found the other aspects of the headphone experience unsatisfactory. Nor could I really make useful comparative judgments between speakers and phones. The two systems seemed too different.
Recently, on the urging of two trusted friends, I acquired a pair of ear canal phones, the model ER-4S from Etymotic Research. Lyman Miller uses them for on-location monitoring during recording, and for playback listening. He works mostly with sphere microphones and always seeks to preserve accuracy of sound. Don Barringer was trying to compare them to the ORION to sort out some fine tuning of the speaker.
The earphones are pressed gently into the ear canals, after the white plastic ear tips have been moistened with water from a wetted finger. They seal and isolate effectively against external sounds. They also establish a well defined and repeatable acoustic environment. Based on my previous experience with headphones I was curious about these and decided to investigate their perceived frequency response. Somewhat to my surprise I found a similar response pattern as before. Therefore I proceeded to develop an equalizer and found the following corrections necessary to obtain a uniform perceived amplitude response to sinewaves.
The 7.5 kHz notch is about as deep as with the HD 414's, the 2.5 kHz one though is less pronounced.
Now the interesting part is how it sounds on program material. It would certainly be different, but an improvement? I used my own recordings, what material I have from Lyman and also many of my speaker evaluation CD tracks that I am intimately familiar with. In just a few words, it sounds superb and is very believable in terms of accuracy over the whole spectrum and dynamic range. Listening for a while and then switching the equalization out of the signal path reveals a very exaggerated high frequency emphasis of certain sounds and an overall coloration. This sudden emphasis is a known psychoacoustic effect, when a notched out portion of the spectrum is filled back in. After having heard the earphones with the two peaks in their response removed, it is difficult to go back to them un-equalized. Not surprisingly the bass response of the equalized phones seems stronger. Also the judgment and setting of appropriate loudness levels becomes easy. I consider this an important indicator of system accuracy and why I find remote volume control for loudspeakers absolutely essential once the speakers have reached a certain level of performance.
It would be difficult to explain the equalizer response by referring to head-related-transfer-functions, HRTF, because in this case the entrance to the ear canal is sealed off by the earphones and all the effects of the outer ear and body are eliminated. The ER-4S transducers are already equalized in some form to generate the same sound pressure at the ear drum as would be caused by a diffuse sound field. The added equalization is not simply an inversion of a Fletcher-Munson equal loudness contour, because it changes much more gradually with frequency. The 7.5 kHz peak is due to the acoustic impedance mismatch between transducer, ear canal and ear drum causing a half wavelength resonance in the canal. The 2.5 kHz peak appears to be designed into the earphones.
I find two observations remarkable:
I encourage you to experiment and to draw your own conclusions.
Realize that you need to find your own, individualized equalization, which also
may be different for left and right ears. Take my filter as a starting
Already the earphones have become very useful in confirming the accuracy of the ORION. Material which sounded marginally recorded, or had unfamiliar artifacts when heard through the speakers, revealed the same misbehavior through the equalized ER-4S. Indeed the two systems are highly comparable in their frequency response, resolution and dynamics. Being isolated from extraneous noise and any room effects gives an immediate connection to the recorded material with the ER-4S. I even find that the distortion of acoustic distance is not nearly as disconcerting as I remember from earlier days. I have spent hours of listening without getting tired. If you want to withdraw into your own world and disconnect from your surroundings these earphones are it. At 10% of the material cost of the ORION they seem like immensely affordable reference transducers, whether you are designing speakers or just want to hear how accurate your system really is.
The equalizer is formed by inserting a passive network between the earphone amplifier output and the connector to the ER-4S. The headphone amplifier's open circuit output voltage is Vs and its output impedance is R1. The earphone impedance is 100 ohm and symbolized by R2 in the schematic below.
The insertion loss of the network will be 1.6 dB when a value of 20 ohm is assumed for R1. The inductor L must have sufficiently low dc resistance R3 compared to R1 to achieve the necessary notch depth. The inductance value determines the width of the notch. Larger values make it narrower. The notch frequency is then adjusted by C. The optimum values for L, C, and R3 are determined empirically by adjusting them for a constant amplitude sound while a sinewave generator's frequency is changed. The generator must have the same output impedance as the headphone amplifier output in order to determine the necessary values for the series-tuned circuit. If the impedance is 20 ohm, then a power amplifier may be needed. By adding a resistor to its output the total impedance can be brought to 20 ohm. The generator voltage Vs must not exceed 400 mVrms to keep the sound pressure level below 100 dB and avoid damage to hearing.
Be extremely careful with signal levels when performing your perceived frequency response tests!
I would expect the 7.5 kHz notch frequency to be somewhat different for different individuals. The notch depth does not appear to vary much. Thus, as a first approximation, you may need to change only the value of C relative to what I used for myself and not R3. A 10% increase in C decreases f0 by 5%.
Before attempting to build the equalizer circuit it is absolutely necessary to check that the chosen headphone amplifier or the output port into which the circuit is plugged, is capable of driving the network and phones properly. The output impedance Rs must be known. It is easily measured in a few steps.
1 - Play a 1 kHz sinewave signal from a Test CD.
Rs = 100 (Vs - V) / V
For example, with Vs = 530 mV and V = 490 mV measured,
Next, with the 100 ohm resistors still in place, measure the maximum output voltage V by increasing the volume setting. For 2 Vrms the SPL from the ER-4S will be 114 dB. The maximum safe continuous input voltage to the transducers is specified as 3 Vrms, but this is definitely not safe for your hearing on a continuous basis.
Be very careful, the ER-4S are capable of 122 dB SPL!
If you have test frequencies from 10 Hz to 20 kHz on your CD you can also check the frequency response of your amplifier when driving 100 ohm loads. Be sure the DVM has a flat response.
The ER-4S equalization response curve shown above was measured with R1 = 100 ohm driving impedance and is the voltage across the transducers. The rise at high frequencies is due to an increasing transducer impedance from 105 ohm at dc. The notch depth changes by less than 0.5 dB if R1 stays within a range of 80 ohm to 120 ohm and the shape of the curve changes insignificantly. The major effect is on the insertion loss. Some headphone outputs may have very low impedance and I modeled a circuit with essentially the same response when R1 = 20 ohm. It applies over a range from 18 ohm to 22 ohm. If the measured headphone amplifier output impedance falls outside these ranges, then add external resistors R0. A measured Rs above 120 ohm requires different circuit values from what I provide. It also means >7 dB insertion loss and the amplifier may not be able to deliver sufficient maximum output voltage.
Note: ER-4P phones have lower impedance and a rolled off high frequency response. By adding 75 ohm resistors in series with each channel they are effectively turned into ER-4S phones and the high frequency roll-off is greatly reduced. The resistors can be added to the output side of the network below.
The actual circuit diagram is shown below with values for
the components in Table 1. A selection of useful components is given in Table 2.
The resistor value includes the dc resistance (DCR) of the inductor in the
signal path. Add external resistors to obtain the required values for R10, R20
and R30, if necessary. Rs+R0 corresponds to R1 in the general notch filter
network above. Determine R0 by subtracting the measured Rs from the Table 1
value for Rs+R0.
Table 2. List of useful parts from www.mouser.com, (800) 346-6873
Toggle switch S30 allows for a small amount of cross-feed between channels. This feature reduces some of the unnatural feeling when a source signal is only presented to one ear. At higher frequencies the shadowing of the head would allow this to occur naturally and the cross-feed frequency response is therefore rolled off.
Experiment to find your own, individual equalization for the ER-4S. You have here a performance potential that circum-aural and supra-aural headphones have difficulty to match, because of the many physical variables in their use.
SONY Fontopia® Bud Style Headphones
Recently I evaluated Sony MDR-EX71SL ear buds and was surprised by their accuracy of sound reproduction. They are quite comfortable to wear, yet their seal is good enough and repeatable for excellent bass reproduction. The very soft cups conform well to the surroundings of the ear canal entrance. The subjective frequency response of the EX71SL is remarkably smooth with only a resonance due to the mismatched acoustic termination of the ear canal. Since the buds are not inserted into the ear canal, but rest in front of it, the length of the acoustic transmission line is longer than for the ER-4S and thus its resonance frequency becomes lower, i.e. 5.2 kHz instead of 7.5 kHz.
The resonance peak was equalized experimentally by adjusting the notch filter parameters, while observing the change in volume level as the frequency of a sinewave generator was changed. Switching equalization on and off is clearly audible on pink noise. It becomes much less so on program material, where its effect only becomes noticeable when there is a lot of high frequency content, like in recordings of massed strings. The low frequency response is articulate and reaches very far down. Its level seems slightly excessive.
The Sony ear buds present an attractive low cost alternative to the Etymotic ear canal phones when maximum sound isolation is not a concern. The equalized ER-4S phones work exceptionally well when monitoring a recording while close to the musicians. Either one of them belongs in the tool box of every loudspeaker designer, in order to have a reliable reference for hearing the minute details of what was actually stored on a given CD.
SHURE E2c Earphones
In my ongoing investigation of reference quality transducers I came across the Shure E2c earphones. The box and documentation that I received from www.samedaymusic.com was labeled E2. Upon questioning Shure Inc. wrote to me: "There is no difference between the E2 and the E2C earphones. The actual product is exactly the same only the packaging is different. The E2's are packaged more towards musician who use in ear monitors and the E2C's are packaged for all audio users."
November 2008 update:
Listening tests to commercial recordings and my own DAT head-related recordings revealed a very realistic tonal balance, frequency extension, superb detail and dynamics without using any equalization. When I actually listened to a slow sinewave sweep, I found a fairly broad peak centered at 3.8 kHz, a mild peak at 5.9 kHz and another peak at 9.2 kHz. I made no attempt at equalization, because the realism of the E2 is so convincing. Also, at all frequencies the sound seemed fuller than with the other models, though I cannot explain why it should.
There are definite differences between the three earphones that I tested, though not so much in their sound, if you equalize two of them. The ER-4S definitely need equalization, otherwise their sound is quite colored, which is easy to spot on most recordings. The MDR-EX71SL could almost be used without equalization since the canal resonance shows only on certain program material. This might be acceptable for general listening but not for use as a sound reference. The E2 meets my sonic requirements right out of the box. These are the sturdiest phones of the three in all aspects. Vibration transfer from the cable is low, on par with the Sony and far better than the Etymotic. Sound isolation against ambient noise is not quite as high as for the ER-4S, but higher than with the EX71SL. The $100 price tag of the E2 falls between the EX71SL at $50 and the ER-4S at $270.
I want to emphasize that anyone who makes critical evaluations of loudspeakers needs to know the quality of his source material. Any one of the three earphones that I investigated can become a useful reference transducer.
Circuit design, analysis and documentation for the "Direct Coupled Stereo Headphone Amplifier" by John Conover should be mentioned here because it is exemplary. Even if you do not need such amplifier, a study of this project could be highly educational. He also developed a "Spatial Distortion Reduction Headphone Amplifier" and a functionally related software alternative.
You might also be interested in the dc-servo design of an
amplifier by Kevin Gilmore.