Quick Summary:  This Article describes a device and procedure for measuring the sensitivity of earphone elements.  Its purpose is to provide a quantitative method for comparing elements.  Actual measurements of an assortment of elements is provided.

The Transduction power loss of a headphone element can be defined as the ratio of its output acoustical power to input electrical power.  We will call it HPEL and express it in dB.  A convenient way to measure HPEL is to use one element of a pair of identical headphone elements as a speaker and the other as a microphone, acoustically couple them together and then measure the input electrical power to the speaker element and the output electrical power from other element.  Ten times the ratio of the log of the ratio of output to input power is the transduction power loss of the combination of the two elements, in dB.  If the two elements are identical, the power loss of each is one-half that figure.

Here is a step by step procedure:

1. Know the average impedance, Zh of the headphone elements.  If you don't know it, measure it by means of a FILVORA (See Article #2 of this series), or estimate it as 6 times the DC resistance of the element.
2. Couple two identical elements A and B together with an appropriate acoustic coupler and hold everything in place with several heavy rubber bands.
3. See Fig. 1.  Connect a white noise generator through a 0.3-3.3 kHz bandpass filter and a source resistor Rh of value Zh to element A. Connect element B to an output load resistor Rh of value equal to Zh.  The filter is necessary to limit the bandwidth of the white noise signal to the audible range of interest.  If this were not done, the reading of e1 would be too high, since the noise at that point covers a wider band than that at the output.. 
4. We will measure the HPEL by the insertion loss method.  See the section on "Maximum Available Power" in Article #0 and the next-to-last section of Article #5 for information on this method.  Measure the input voltage e1 at point P1 and output voltage e2 at point P2.  The HPEL  = 5*log (4*((e2/e1)^2)) dB.  The 5 is there instead of the usual 10 because only half the measured loss can be attributed to one element, and we are actually measuring the sum of the two losses.  Note: It is usually recommended that elements A and B be pressed together with a force of 1 to 2 pounds so that no air leak occurs between the elements and the coupler.  Actually, if squeezing elements A and B together more tightly than the rubber bands do does not increase the value of voltage e2, the rubber bands are OK to use by themselves.

Figure 2 shows the circuit of a bandpass filter having -3 dB points of 0.3 and 3.36 kHz with a loss of 0.4 dB at 1.0 kHz.  It's powered by a common 9 V. battery.  Since the typical 1000 DC Ohms magnetic earphone element has an impedance of 6000 Ohms and the typical balanced armature sound powered magnetic element has an impedance of 600 Ohms, approximations of these two resistance values are included in the switched resistive output controlled by S1.  An IC suitable for the circuit is an LF353 or an MC34002.
 

A convenient source for white noise is the headphone output of a small FM/AM transistor receiver, switched to FM and tuned to a point on the dial where it receives no signal, just noise.  The noise density is rolled off at 6 dB per octave above 2.1 kHz by the de-emphasis filter in the receiver, but this should make little difference in the results.  The acoustic coupler used to couple the two elements under test is based on the ANSI 9-A earphone coupler.  The approximation to an ANSI 9-A coupler I use is made up of a 0.492 inch long piece of copper tubing that has an inside diameter of 0.815 inches.  This piece of tubing has a neoprene O ring glued on each end.  The ID of the O ring is 0.735 and it is 1/8 inches thick.  The volume is about 5.6 cubic cm.  An alternative coupler that gives almost the same results is a stack of eight or nine 1/8 inch thick garden hose washers having an ID of about 5/8 inches.  The ANSI 9-A coupler is (was?) the standard coupler used in Audiometry when calibrating an earphone element with a standard microphone.  It is a greatly simplified version of a model of the human ear canal with an earphone cushion pressing on it.
 

Comments:

The DVM should preferably be an RMS responding instrument.  The typical DVM responds to the full wave rectified average signal and will be satisfactory.  Don't use a meter that responds to the peak or peak-to-peak value of the AC signal. 

1. A pink-noise generator can be substituted for the white-noise generator, but it is hard to use.  It has a larger low-frequency output than does the white-noise generator and therefor will show a greater fluctuation in the output as read on the DVM.
2. The noise output voltage of the white noise generator will probably have to be amplified (or increased with an audio transformer) when measuring headphone elements having a high HPEL, in order to overcome ambient noise and hum pickup.
3. The measurement method described here does not include the effect of the usual air leak between the ear pinna and the headphone element cap.  This leak rolls off the low frequency response below 500 Hz and results in a somewhat greater actual power power loss than is shown below.
4. It's a good idea to make sure that the two elements used have about the same sensitivity, otherwise the result will be the average of the good and not-so-good elements. The result will be a higher loss than if two good elements were used.

* One of the RCA/GE units was about 6 dB less sensitive than the other. Thanks to Dieter Billinger (sky_wave_99), I knew that some RCA/GE elements having low sensitivity could be improved by sticking a small neodymium magnet to the outside of the case.  It worked in this case, increasing sensitivity of the weak element by 6 dB, so it was somewhat more sensitive than the other element.  BTW, a magnet could not increase the sensitivity of the other originally more sensitive element. These two units appear to be of somewhat different construction.  To easily compare the power sensitivity of any two elements, even if they differ widely in impedance, see Article #3.
** These two elements were selected for having strong magnets, and their air gaps were optimized.  Run-of-the-mill Brandes elements may not be as sensitive.

To help in understanding these charts, consider that an eight dB (6.3 times) change of power is usually perceived as a two times subjective change in loudness.

In order to compare the sensitivity of headphone elements that are used flat against the ear, as well as those that are not, (but are inserted into the ear canal (tips) or outer ear (buds)), I decided to make one of my best elements a "standard" and compare the others to it using a DFLVORA (see Article #3).  That "standard" is a Western Electric Sound Powered receiver element # D173012 held flat against one ear.  In the results shown below, two Mouser elements were tested and found to be of equal sensitivity.  (That was after I found one Mouser to be weak until I whacked it several times.)   The two Radio Shack units also tested equal.   Note that the DFLVORA can be used to easily compare the power sensitivity of any two earphone elements even if they differ greatly in impedance.
 

These comparisons were made using a voice signal from a small transistor radio fed into a DFLVORA with the radio volume set to a level at which I estimated I could understand about 50% of the words.  Results may differ for people who are not old and do not have poor high frequency hearing, such as myself.  The differences in tone quality between the standard element and a DUT will have a different effect on intelligibility for different people.

The sensitivity values for the piezo electric elements can only be attained if the elements do not have a resistor placed across them to supply a DC path for diode detector current.  The resistor adds loss (although this is a low cost approach to provide a DC path for the diode current).  Also, if the resistor is made large in order to reduce its loss contribution, audio distortion will oftentimes take its place.  The best way to drive the elements is to use an audio transformer to impedance match the diode detector output resistance to the average impedance of the element. The transformer supplies supply a DC return path for the diode.  A further advantage of using a transformer is that no DC voltage can get across the piezo element.  Sometimes, if a strong signal is tuned in and it produces a large rectified DC voltage on the element, the element will "freeze" and its sensitivity will drop.  See Article #5 for info on transformer coupling and diode DC resistance loading. (The value of the DC load resistance on the diode should equal the average value of the AC audio load impedance.)

*  The comparison of the sensitivity of an element in a series connected element pair DUT with the "standard element" was made in the following manner:  The full (two element) headphones DUT was connected to the J1 output of the DFLVORA. The DFLVORA was fed by a weak voice signal. and the source resistance switch adjusted for the greatest volume and intelligibility.  The "standard element" was then connected to the J2 output and the 3, 6, and/or 12 dB attenuators were adjusted so that the intelligibility of the voice in the "standard element" was equal to that in one element of the headphones DUT .(The other element was left dangling.)  The amount of attenuation placed in the circuit feeding the standard element is a measure of the difference in sensitivity between the standard and the DUT.  Since 1/2 the power going into the full headphones DUT goes into each element, one element of the DUT headphones being listened to receives 1/2 the power (3 dB less power) than that delivered to the full headphones, giving the reading for a single element a 3 dB handicap.  Thus, the sensitivity of one element of the headphones DUT is 3 dB better than the sum of the readings of the attenuators.  This 3 dB correction is made in the figures for the DUT in the table above. When doing a comparison of this type (comparing one element of the pair in a  full headphones, to a single "standard element"), first check the volume in each of the two elements of the pair.  If they are not equal, error will result.  If the volumes are not too far apart, perform the measurement for each element of the pair and average the result.  There is some error introduced by the procedure given above because the acoustic loading on each earphone of the pair is not the same.

Summary:
The Western Electric #509W headphones tested 6 db less sensitive than the "standard".
The Baldwin type C headphones tested 9 dB less sensitive than the "standard".
The Brandes Superior Marched Tone headphones and the two Radio Shack Piezoelectric speakers tested 12 dB less sensitive than the "standard".
The each of two Mouser Ceramic Earpieces tested 20 dB less sensitive than the "standard".

In all cases it is assumed that the source resistance driving an element is equal to the average impedance of the element over the audio frequency range of interest.  This is the closest that we can get to an impedance matched condition.

Last item:  Remember that headphone sensitivity can vary a lot from unit to unit.  The figures given above are not gospel for all units of a particular model.  Diaphragms warp, magnets weaken and air gaps may get changed.  All affect the sensitivity.

Published: 08/30/00;  Last revision: 07/12/01

 Return to Home Page