Article author: Jon Sank
Our previous article on microphone testing described methods that we anticipated using for microphone reviews.
When we wrote that article we may have had a stereotyped image of a moderately priced dynamic microphone for audiophile use with a cassette recorder.
Instead, we have been privileged to review a number of excellent microphones including ribbon, dynamic, electret and air-condenser types and have been pleasantly surprised to find that these high quality mikes are priced within the reach of audiophiles.
Our goal remains unchanged; we intend to acquire data as complete as time and budget allow, and follow applicable standards as closely as possible. To meet our goal, we are constantly improving our techniques to keep up with advancing technology in microphones.
This update will describe some of our "new wrinkles" in testing. Aspects of two new tests will be described, noise and frequency response testing under plane-wave conditions outdoors, and the setup and procedures that we're using for listening tests.
A recent review (2) showed results of an experimental noise test on an air-condenser microphone. The 1/3-octave noise levels of this high quality mike provide good baseline data. The overall measured levels were higher than specs, which inspired us to refine our methods. Our set up was similar to Fig. 1 but employed the same wide-band, solid-state RCA BA-31 used for frequency response testing [Fig. 1]. For noise testing only, we switched to our "standard of quiet," the RCA OP-5 portable amplifier, an old vacuum-tube, remote-broadcast amplifier of high quality, with a step attenuator and classic Weston 4-in. VU meter. The overall noise levels are -129.5 dBm (unweighted) and -131 dBm ("A" weighted) referred to the input. The dBV numbers are the same because the rated input source impedance is 250 ohms [1]. The 1/3 octave spectrum of the OP-6 noise is shown in Fig. 3, and the SPL scale is to be disregarded for the present. The 250 ohm impedance rating does not conform to the recent broadcast standard which is 150 ohms, but is ideal for testing today's low impedance microphones which vary from 150 to 600 ohms.
A little mathematics will show why a restricted frequency response amplifier is less noisy than a wide band unit. Random noise is generated by thermal agitation in a resistor or in the resistive component of a complex impedance. The total rms open-circuit voltage e in the frequency band (f2-f1) is [4]:
en = √( 4kTR(f2-f1))
Where k is Boltzmann's Constant, T the absolute temperature (°K), and R is the resistance value in ohms. If we reduce bandwidth from 150 kHz to 15 kHz, e will be reduced by 10 dB. Failing to band-limit the amplifier can result in a high noise reading that is not related to what you hear. Alternately, we could chill our device to near absolute zero and the noise will be very low, but this is only practical for specialized radio receivers.
If we assume a temperature of 20 degrees C and a bandwidth of 15 kHz, the noise voltage En in dB re 1 volt (dBV) is:
en = √(-156.2 + 10 Log10 R)
The 250-ohm resistor termination for our OP-6 produces -132.2 dBV of noise according to the equation. Thus the OP-6 adds -129.5 (-132.2) = 2.7 dB to the noise and is said to have a noise figure of 2.7 dB. preamp
Fig. 1-Test setup for noise measurement.
Fig. 2-Sound transmissions loss of microphone noise testing box.
This is a low N.F. for an audio amplifier, and a limited review of IC Op Amp specifications plus calculations have shown a higher N.F., particularly with unloaded input conditions where the input impedance is five to ten times nominal mike impedance. We would be pleased to receive test data on quiet ICs.
We've only considered the fundamental case of noise generated by a resistance. References 3 and 4 discuss other sources of noise in microphones and amplifiers.
A precision sound level meter (less microphone) is included in our noise test setup to indicate overall levels unweighted (flat response) and "A" weighted (bass rolloff). The "A" weighting is appropriate because microphone noise would be perceived by the ear at a very low sound level where "hiss" is more easily heard than "hum."
European manufacturers specify DIN weighting, but "A" weighting is commonly used in the U.S. for acoustic noise measurements and ratings and is specified by NAB for tape equipment testing. Our noise test equipment conforms to the standards listed in Table I, and these may be consulted for more detailed information.
Our sound-retardant box is made of 3/4-in. plywood, double thickness, braced by 2 x 3 in. boards and lined with several layers of 1-in. acoustical foam. It is purposely non-metallic and non-magnetic because microphone noise properly includes components resulting from ambient magnetic or electric fields. We position the box to minimize hum if it is encountered, because in practice one may so orient a mike. The sound transmission loss (TL) of the box has been measured by exposing it to a high level, calibrated, reverberant sound field. The results are shown in Fig. 2. The loss at low frequencies is remarkably high for an inexpensive box. Room ambient noise generally peaks in the 63 and 125 Hz region due to motors and lamp ballasts, then falls to a low SPL at 1000 Hz and above. This complements the TL curve, and the resulting ambient sound inside the box tends toward a flat spectrum, which minimizes interference to micro phone tests. Nevertheless, we only make tests when the room is very quiet, and continuously listen for ambient noise from the mike.
-----------
Table I ANSI (American National Standards Institute) and NAB (National Association of Broadcasters) standards related to audio noise measurement.
ANSI 51.11-1966 (R1976) octave, half-octave, and third-octave filter sets, specifications.
ANSI 51.4-1971 (R1976) sound level meters, specifications.
NAB Standard, April 1965 magnetic tape recording and reproducing (reel-to-reel).
-----------
We have shown that a simple resistor makes noise, and thus condenser mikes having integral amplifiers are not singular in their ability to generate noise. In fact, all types of mikes make noise. Figure 3 shows some "baseline" data on our lab reference BK-5B ribbon mike. Below 1000 Hz, hum and harmonics predominate above the OP-6 amplifier noise even though the box was oriented to minimize hum. Incidentally, our sound room does not have strong fields, and the BK-5B has low hum sensitivity. Above 1000 Hz, the mike generates no more noise than a resistor. The result is that the noise does not exceed 25 dB unweighted and 19 dB "A" weighted equivalent sound pressure level in dB re 20 microPascals. How is this rating determined? We calibrate our chart paper scale as well as the sound level meter scale in dBV by the insert voltage technique described in Ref. 1. Then we remove this calibrating voltage and measure overall noise levels with the SLM, and record the spectrum on the chart. In order to compare with manufacturer's specifications, the dBV numbers must be converted to equivalent SPL (Lp) as follows:
L_pn= E – R_t + 94.
R, is the microphone acoustic sensitivity (previously measured) expressed in dBV/Pascal, as described in Ref. 1. Obviously, if E is "A" weighted, Lp, will be an "A" weighted sound level.
The L number may be easily utilized in a practical recording situation. Let's assume we have a rather noisy mike with an "A" weighted Lp of 35 dB and a Nagra IV SD recorder with an ("A" weighted?) S/N of 72 dB (6). In order that the mike noise will be below tape noise, the sound level at the mike must exceed Lp by 72 dB (plus a safety margin). Thus our minimum SPL must be an astounding 107 dB! With this mike we can record high-level rock music or pavement breakers but preferably not classical music or normal speech. If we use our BK-5B and OP-6 combination (which is a very quiet system), our minimum SPL is 19+72 = 91 dB.
We can record classical music successfully because the recorder gain will be adjusted for 0 VU on the loud passages, which should exceed 91 dB. A high quality, high S/N tape recorder requires a low noise, high sensitivity microphone.
The mike preamp in your recorder or mixer can degrade S/N if a low sensitivity or mismatched impedance mike is used or the preamp can be just plain noisy.
Now that we've attracted your attention to microphone noise, we think it only fair to mention a saving grace in realistic audiophile recording sessions and that is ambient acoustical noise.
The church in which we record has an ambient noise level of 25 dB "A" weighted, with no people. During a concert, the background sound must obviously be higher. Thus, any mike having a noise rating of less than 25 dB "A" is adequate.
What is the least noisy mike we've measured? It is (surprise!) a condenser. Our RCA MI-10006A vari-directional condenser microphone (7) measures approximately 7 dB unweighted and 0 dB "A" weighted! The MI-10006A was designed for motion picture applications on very quiet sound stages, but surprisingly was not favored because of infrequent "pop" noises which could spoil a long scene. This mike has a lower "threshold of hearing" at low frequencies, than the human ear, and listening to ambient sound with this mike reveals otherwise inaudible noises that sound like distant thunderstorms.
Listening Tests
In our 1977 article in Audio, we described the principles of AB comparisons. We did not show specific methods because of space limitations, but think it a good idea to describe the details, both for the record and for audiophiles who are interested in making more productive listening tests.
Figure 4 shows our test setup. It may appear complicated, but the audiophile could construct a rough equivalent with a pair of mike mixers, a switch, and a headset.
We always use the same reference mike as "A," our BK-5B that we built with T.L.C. about 10 years ago. Periodic checks on the frequency response have shown it to be a stable mike. Since it is a cardioid, it can properly serve only for comparison with cardioid microphones, but these comprise about 80 percent of mikes reviewed to date.
Each mike is connected to the same load, a balanced, 150-ohm unloaded input. The step attenuators in each channel are adjusted for equal audio levels as indicated on the VU meter.
The difference in attenuator settings in decibels should equal the measured difference in microphone sensitivities. The master attenuator is adjusted so that the overall gain (in the reference "A" channel, at least) is unity, so a ribbon microphone level signal is present at the OP-6 input. The A-B switch is a console-type key switch. It is followed by the matching transformer which presents a balanced Circuit to the OP-6 amplifier which functions to boost the signal to a normal program level of +8 VU and provides a large VU meter for indicating levels.
Fig. 4--Equipment setup for comparative listening tests.
It is important to use headphones which reproduce the entire audio frequency range with low distortion. We have tried a number of high quality phones, but in general, changing phones does not change the conclusions of an A-B comparison of microphones.
We try to follow an orderly and consistent test procedure.
First, we position the two mikes close together, make a talk test, balance the levels, and check for correlation of attenuator settings and known differences in sensitivities. This is followed by several comparative tests, some of which are more objective than subjective:
1. Evaluate sound quality of voice.
2. Shake the mike and listen for noises; check operation of on off or voice-music switches.
3. Tap the stands and mikes and observe outputs.
4. Check sensitivities to breath "pops," with and without accessory windscreens.
5. Orient each mike for maximum output at a specific distance from a coil carrying 60 Hz current, with a field of approximately one gauss, and note relative hum levels.
6. Check clipping level on calibrated scope with loud speech.
The voltage value is converted to SPL by calculation using the measured RI value.
Pending results from these tests, we set up the microphones in our rectangular listening room which is approximately 400 sq. ft. in volume, with 15-in. studio monitor speakers on one end and 15-in. thick sound absorbing material at the other. Real time analysis at selected locations throughout the room reveals a uniform acoustic response with no need for "room" equalization.
The monitor speakers have uniform response from 40 to 15,000 Hz, so that a full range of musical instrument sounds can be reproduced from master tapes of concerts, and this is satisfactory source material for most A-B comparisons of microphones. Then we usually make a quick test using available live instruments, to see if the results correlate.
Further listening tests may be made as circumstances dictate, including stereo taping of live concerts, which necessitates obtaining a second mike. Products for review are submitted in unit quantity, which is usually one microphone. We have not yet attempted A-B comparisons between pairs of microphones.
Response and Directivity Tests--Outdoors
The relevant ANSI and EIA standards (see April 1977, Audio article) require that a plane wave be used for testing frequency response and directional characteristics of microphones. This requirement is essentially met at high frequencies where the microphone is one or more wavelengths from the source. At lower frequencies where the distance is less than about 1/3 wavelength from the source, pressure-gradient microphones, such as cardioid or figure-eight pattern, exhibit a bass boost or proximity effect as illustrated in Fig. 8 of our 1977 article. A cardioid will exhibit a varying effect depending upon angle of incidence. The 0-degree response follows the curves of the graph, the 90-degree response is not boosted, and the 180 degree response is boosted more than the 0 degree.
Normally, we measure microphones indoors on the small spherical sound source [1] at distances from 6 in. to 2 ft. These data are published along with the plane wave response which is usually calculated from the 6 and 12 in. curves, using the proximity effect graph. This is fine when you are using a source whose dimension is small compared to the microphone distance, in other words, the mike is in the far field of the source.
When the mike is closer than twice the source dimension (for this purpose use the diaphragm diameter), the distance to the (effective) point source is undefined, probably varies with frequency, and plane wave response cannot be calculated.
Fig. 5--The large spherical sound source in the foreground with the SPX
ribbon microphone enclosed in a large windscreen.
We have discovered that few, if any, manufacturers actually publish low frequency response curves for plane wave conditions. Frequently, the distance to the source is not mentioned, and the source dimension is rarely stated. Most overseas manufacturers use a 1-meter distance. We discovered one overseas company testing at 50 cm, and one domestic company testing at 12 in. (30.5 cm). The latter was using a 12-in. speaker. The 50-cm distance is all right if the source is less than 25 cm in diameter and the data appropriately corrected to plane wave conditions. Testing at 12 in. from a 12-in. speaker is improper unless, of course, you only test omnidirectional (pressure) microphones.
L. J. Anderson (for whom we worked at RCA) is quoted by Ref. 5, stating that essentially plane wave conditions are obtained at a distance d from a (point) source, above a lower limiting frequency f according to f=350+d, where d is in ft. and f in Hz.
We have calculated f for the distances mentioned in this article:
d meters | d feet | Hz
_________________
0.15 0.5 700
0.31 1.0 350
0.50 1.6 219
0.61 2.0 175
1.00 3.3 106
1.22 4.0 88
1.82 6.0 58
Strictly speaking, data corrections should be made when the microphone is closer than six ft. from the source.
Our indoor test setup with a small spherical source is satisfactory for omni, cardioid, and figure-8 microphones. The 180 degree response of cardioids at two ft. is a little in doubt below 175 Hz because of uncertain correction factors, but this is a minor difficulty.
Our test procedures had been routine until the Editor (in his infinite wisdom) submitted a shotgun type of ultra-directional microphone. Our extensive experience with this kind of mike has shown that plane wave test conditions are definitely required. It is intuitively evident that a microphone with sound entrances distributed along 40 cm of length should not be tested at, for example, 40 cm from a source, because the SPL at the rear of the mike will be 6 dB less than the SPL at the front. Generally, these mikes are intended to reject reverberant sound that would measure the same SPL at any point in space to be occupied by the mike.
In response to this challenge, we produced our large spherical sound source (Fig. 5), which like the Son of Godzilla, is not smaller than its father, the small spherical sound source! We are sorry to disappoint you with a description of this device, which is in no way as magnificent as its predecessor. The large source consists of a fiberglass sphere, about 18 in. in diameter,1/4-in. thick, coated on the inside with 1 1/4 in. of vibration damping compound, and filled with absorbing material. The transducer is a selected ALTEC 755C (8 in.) speaker. Our principal use for this source is for reverberation, response, and sound transmission tests in buildings. Devotees of plumbing will note the stand which is made of 1 1/4-in. pipe and contains an esoteric fitting known as a "side arm cross." This is actually a railing fitting it's hard to find but is the key ingredient of this very useful portable stand.
Figure 5 may lead you to believe that our beloved SPX laboratory ribbon microphone has had a monstrous offspring, but alas, this is not so. The microphone shown with the source is the SPX enclosed in a large windscreen. This screen is sufficiently good that wind noise from trees is generally heard before the velocity becomes great enough to generate noise in the SPX. The screen is a silver-solderer's masterpiece, made of thin brass rod covered with open-weave fabric. It was made to permit outdoor tests of large horn-type loudspeakers.
Fig. 6--Calibration of the large spherical sound source.
Those of you who have tried to test loudspeakers outdoors using conventional condenser pressure microphones know that you must either bury the speaker flush with the ground or position it and the mike very high above the ground to avoid wiggles from ground reflections. The ribbon velocity microphone produces accurate, smooth calibration curves of our source with the more practical setup of Fig. 5 where the source is six ft. from the mike, six ft. above grade. The figure-eight directional pattern of the velocity microphone rejects ground reflections, and of course at a six-ft. distance the mike has negligible bass boost due to the proximity effect [8]. It follows that our outdoor setup is best suited to frequency response tests of microphones having a null or at least some significant rejection of sound from the 90-degree direction.
These microphones include figure-eight, ultra-directional shotguns, and super cardioids. Of course, most types of mikes, other than the shotgun, can be tested indoors. On high quality cardioids, the outdoor test will show a greater front-to-back ratio, particularly below 200 Hz, than the indoor test.
Figure 6 shows the calibration of the source. The large source is not ideally suited to frequency response tests of microphones having very smooth response, as the reduced date contain wiggles that are not present on curves from indoor tests. Presumably this is because the response of the small sound source is much smoother [1, Fig. 5]. The various schemes of source flattening we described enable more precise tests of microphones using commercial loudspeakers as sources, but it is doubtful that an artificially smoothed cone speaker can even equal the naturally smooth small source in making accurate microphone measurements.
We hope that this explanation of our "new wrinkles" in microphone testing has given you food for thought, both natural and artificial, of course.
References
1. Sank, "The Compleat Microphone Evaluation," Audio, Vol. 61, No. 4; pp. 48-58, (April, 1977).
2. Sank, "Neumann Models 83, 84, & 85 fet-80 Series Microphone System," Audio, Vol. 62, No. 1; pp. 80-82, (Jan., 1978).
3. Olson, Acoustical Engineering, pp. 335-339, Van Nostrand, 1957.
4. Beranek, Acoustic Measurements, pp. 189-192, J. Wiley, 1967.
5. ibid., pp. 642
6. Equipment Directory, Audio, Vol. 61, No. 10; pp. 166-168, (Oct., 1977).
7. Tremaine, Audio Cyclopedia, pp. 209-210, Howard W. Sams & Co., 1977.
8. Yes, we realize that your test equipment store is fresh out of SPXs. If you have an application (and a budget), we'd like to hear about it.
Also see: A Guide to COINCIDENT MIKES (Nov. 1978)
Nakamichi Model CM-700 Electret Condenser Microphone Systems (Sept. 1978)
(Source: Audio magazine, Sept. 1978; )
= = = =