Review of the Sanken COS-11D-BP omnidirectional lavalier microphone (for linguists)
I would like to thank Mr. Martin Ucik, General Manager of Sanken/plus24, for providing the microphone for review and for offering valuable information for the purposes of the review.
Introduction: About Sanken microphones
Sanken is a family owned business in Japan dating back to 1925. Sanken has enjoyed a very fruitful relationship with NHK, a major Japanese TV and radio broadcast company, for whom it develops microphone solutions specially designed to meet the needs of modern professional audio production. Sanken engineers are known for innovative, physics-driven designs. For example, the rectangular diaphragm used in the COS-11D is an original Sanken design, only later implemented by other manufacturers. Most of the Sanken microphone designs are a direct result of a problem-solving engineering philosophy. They are unique tools, designed to address specific needs of audio production. Sanken's largest market is feature film, TV episode, and reality TV productions, where high audio quality is essential.
plus24, located in West Hollywood, CA, is Sanken's representative for the Western Hemisphere, Sanken's largest market. plus24 provides both marketing and consulting services. They are also instrumental in suggesting design solutions and testing microphone prototypes.
Sanken provides first-rate technical support and service. plus24 repairs many common electrical and mechanical issues, but whenever a capsule issue is involved (e.g., a diaphragm problem), the microphone is sent to Japan for specialized repair and testing in an anechoic chamber. All serviced microphones are refurbished to fully meet factory specifications, and are brought to a like-new condition.
Using lavalier microphones for research purposes
If you look through the microphone reviews available on this site, you will notice that most of them are omnidirectional head-mounted condenser microphones. One of the reasons why I typically recommend head-mounted microphones is their close placement to the talker's lips. The inverse-square law is the speech recordist's worst enemy. Positioning the microphone close to the talker's lips (1-5 cm), and keeping it at a constant distance are the prerequisites of rich and unbiased spectral detail. Regrettably, this is one of the most commonly flaunted rules of speech recording.
In addition to ensuring superior spectral detail, head mounting plays another, equally important role. Most of the professional head-mounted designs provide a shock mount for the microphone capsule, which helps isolate the microphone from its immediate environment. Microphones mounted on floor stands or table-top stands are notorious for picking up low-frequency rumble and handling noise. Recordings made with portable recorders with built-in microphones are perhaps the most common example of such noise, as all of us have come to realize at one point or another.
Finally, close microphone placement ensures no unwanted signal filtering (attenuation, head shadowing, etc.) due to obstacles, such as the chin, clothing, hair, etc.
I am often asked why I do not recommend lavalier microphones. In theory, there is no reason why a good lavalier microphone should not be able to make quality-critical speech recordings. I tend to recommend headworn microphones simply because they are much easier to use. Lavaliers have a steeper learning curve and require a bit more effort to achieve a comparable level of spectral detail and an equally high signal-to-noise ratio (SNR).
Of course, in addition to purely acoustic factors, one has to consider the practical issues related to microphone use in the field. There are many contexts in which the use of a head-mounted microphone might simply not be feasible. Sociolinguists are familiar with the Observer's Paradox and may, no doubt, have encountered situations where either they or the subjects did not feel comfortable using a headset microphone. My personal views are that more people are perfectly comfortable with headsets than we might think. I often recommend that you try to buy both a lavalier and a head-mounted microphone, especially if they are based on the same capsule (see, for example, my review of the Audio-Technica BP892 microphone) and use them interchangeably, if necessary.
Proper lavalier placement requires a bit of practice. In live theater, it is common to place the microphone out of sight, on the actor's chest, lapel, collar, tie, bra, or even hairline or wig. However, unless the recording situation involves video, most linguists are not concerned about making the microphone inconspicuous, though one should try to make sure that neither the microphone nor its cord or clip holder rub against the subject's clothing. You should try to place the microphone as close to the subject's lips as possible, say no more than 20 cm away. It is best to try to maintain the same distance across interviews, as much as possible. You can read more about microphone placement techniques in this article.
The alternative is to convert a lavalier to a headset configuration by mounting it on an third-party microphone headband. This is my preferred option. I make my own headbands, so I can use the same microphone capsule in different configurations for most flexibility. You can read more about making a simple DIY headset microphone in this article. However, not all lavalier microphones are suitable to head-mounted use. Omnidirectional condenser microphones of medium sensitivity and high maximum SPL work best.
The Sanken COS-11D-BP design and specifications
Figure 1. (courtesy of Sanken)
The COS-11D microphone was announced in 2009 as the latest version of the popular COS-11 series. The capsule is based on a new designed, developed specifically for improved hybrid wireless transmission. The COS-11D can be terminated for use with many popular wireless systems, such as Lectrosonics, Sennheiser, and Sony. Because I am going to focus exclusively on wired applications, I am going to leave the wireless features out, but you can read Sanken's full press release here.
Sanken designed the COS-11D microphone with speech in mind, advertising the unit as "optimized for human voice." The microphone is engineered to withstand the rigors of both stage and field use, including improved resistance to moisture and temperature, as well as resistance to handling noise. Detailed specifications are listed in Table 1.
Table 1. Sanken COS-11D Specifications
50Hz - 20kHz
6mV/Pa (-44.5dB,0dB=1V/Pa) with battery power
17.8mV/Pa (-35dB) with phantom power
Equivalent noise level (A-weighted)
Max SPL (1% THD)
Output impedance at 1KHz
AA battery , +12V to +52V
less than 0.8mA (AA battery)
122g , 139g (with AA battery)
16.1mm X 4.0mm (diameter)
The Sanken COS-11D has all the acoustic and electrical properties required of a good speech recording microphone. The frequency response is wide and flat (Figure 2). There is a presence peak at around 8,000 - 10,000 Hz, in ways similar to many competitor microphones, such as the Audio-Technica BP892 (review). It has medium sensitivity (6mV/Pa) and a high maximum SPL value (120 dB SPL). It can be powered either by a single AA battery or standard phantom power with the provided power supply unit (Figure 3). It is worth noting that, as expected, the microphone's sensitivity increases to about 17.8 mV/Pa (by approx. 9 dB) when powered by standard 48 V phantom power.
Figure 2. Frequency response of the Sanken COS-11D BP omnidirectional lavalier microphone, according to the manufacturer (courtesy of Sanken)
The battery power adapter
The Sanken COS-11D BP microphone reviewed here comes with a hard-wired power adapter (Figure 3). As mentioned above, the adapter operates on one AA battery or standard 48 V phantom power. Battery life is one of the field recordists' most serious concerns. Many speech recording condenser microphones (e.g., the Beyerdynamic Opus 55 Mk II) need only a small voltage of 1.5 - 9 V for proper operation. However, standard phantom power supplies available on many field mixers and recorders supply as much as 48 V of DC voltage, which puts a significant strain on the device's battery life. Most microphone manufacturers provide a phantom power adapter, such as the Sanken COS-11D BP unit, in order to make the microphone compatible with 48 V phantom power. While this works perfectly with power supplies operating on A/C voltage, it is a rather inefficient system for battery-powered devices. Sanken clearly understands the importance of providing the microphone with a very economical power source (i.e., a AA battery) and should certainly be commended for that. Also, as you can also see in Figure 3, the adapter comes a convenient, removable belt clip.
Figure 3. The Sanken COS-11x BP phantom power supply adapter
One disappointing fact about the Sanken power supply adapter is that it has neither a mute switch, nor a low-cut filter (a.k.a, "high-pass filter" or "bass roll-off filter"). From my point of view, this is not a problem. I rarely use low-cut filters and do not generally recommend using them, either. Besides, dedicated field microphone pre-amplifiers and mixers (e.g., Sound Devices MixPre, reviewed here) have a much more robust implementation of high-pass filters, so Sanken's decision to have no high-pass filter available on the power adapter might be considered perfectly justifiable. The absence of a mute switch is perhaps a bit more troubling, but by no means a deal-breaker. Finally, the adapter is terminated with an XLR-3M connector, which ensures compatibility with most standard professional audio equipment (Figure 4).
Figure 4. The Sanken COS-11x BP phantom power supply adapter: the XLR-3M connector compatible with most professional audio equipment
The microphone comes in four color choices and it includes a number of useful accessories, such as a metal windscreen (Figure 5), a holder clip (Figure 5), a rubber mount, and a vinyl case. The holder clip is very well designed and is a pleasure to use. It allows quick placement with very little fuss. It is among the best I have ever used. The metal wire used to hold the capsule provides a sort of mini shock mount for the microphone, which further reduces handling noise.
Figure 5. The metal windscreen and holder clip that come with the Sanken COS-11D BP microphone
The microphone has a very slim profile (Figure 6), which should facilitate inconspicuous and unobtrusive placement. The microphone capsule has a unique vertical diaphragm design and comes with a coated wire mesh, which provides additional water resistance. The coating of the capsule as well as the cord does reduce handling noise. In fact, the Sanken COS11-D is rather resistant to handling noise, certainly on par with or better than the competition.
Figure 6. The Sanken COS11-D microphone capsule
Low frequency response
Low frequency response is one of the critical features of a speech recording microphone. I must admit that I am somewhat biased in this regard because the bulk of my own acoustic analysis and synthesis lives below 500 Hz. However, if you are interested in the analysis of pitch, phonation, spectral tilt, F1, nasalization, breathiness, laryngalization, lateralization, and other articulatory features with low-frequency correlates, you will need a neutral low-end. I tested the Sanken COS-11D microphone with my usual setup. I generated a waveform with peaks of equal amplitude at the frequencies of 50, 100, 200, 300, 400, and 500 Hz. I then played the signal out of a flat-response loudspeaker and recorded it with the Sanken COS-11D microphone about five inches from the sound source (a typical lavalier distance). The spectrum in Figure 7 shows the low frequency response of the Sanken COS-11D microphone. The microphone reproduces each center frequency really well, with no appreciable attenuation or amplification below 100 Hz. There is, however, a slight dip around 50 Hz. The dip is also evident in the speech sample analysis in Figures 10, 13, and 15 below. The question is, of course, whether this slight attenuation below 100 Hz will influence your analysis. My answer is that, at least in the frequency domain, you should be safe, but you have little room for error. Lavalier microphones do require perfect technique to achieve quality-critical recordings of speech.
From my point of view, the Sanken COS-11D passes the low-frequency test. If used with a flat-response microphone pre-amplifier, the Sanken COS-11D will deliver rather respectable low-frequency response. I would argue that the slight dip below 100 Hz is much less of a problem compared to the overall signal attenuation and decreased spectral detail that might occur as a result of lavalier placement.
Figure 7. Sanken COS-11D's low frequency response showing a flat spectrum, with a slight dip around 50 Hz
Since we record speech at relatively low sound pressure levels (50-60 dB SPL), we require the entire recording chain to be as quiet as possible (to have a low noise floor). Microphones of medium sensitivity (such as the Sanken COS-11D) may require your recorder's pre-amplifier gain to be turned up to perhaps 50-75% of its range, which, depending on the recorder, may or may not generate appreciable levels of self-noise. You can read more about testing equipment for self-noise in this post. The idea is to measure real-world inherent noise in the recording chain. I use the same methodology in all of my microphone reviews, so you can compare the present results with those obtained in the other tests.
Again, the Sanken COS-11D does not disappoint and delivers very respectable self-noise performance. It is unavoidable for a lavalier microphone (placed about 20 cm from the talker's mouth) to have a slightly decreased signal-to-noise ratio. One needs to increase pre-amplifier gain by at least 6 dB in order to capture a signal level comparable to that of a head-mounted microphone. This is going to both slightly increase pre-amplifier noise and allow the microphone to record a higher level of ambient noise. The key, is, of course, proper placement. You should try to place the lavalier as close to the talker's mouth as possible (e.g., on a shirt collar). The further away you place it, the higher the level of inherent noise. Figure 8 shows a spectrum of self-noise generated by a recording chain consisting of the Sanken COS-11D and the Sound Devices USBPre (reviewed here), calibrated to the peak level of -12 dB SPL, and normalized to the RMS of 70 dB SPL.
Figure 8. A spectrum of self-noise generated by the chain consisting of the Sanken COS11-D microphone and the Sound Devices USBPre, calibrated to peak conversational levels of about -12 dBFS, calibrated to the RMS of 70 dB SPL
One word of caution: watch out for low-frequency hum. As you can see in Figure 8, there is a 60 Hz peak of around 50 dB. Low-frequency rumble is virtually unavoidable in any industrialized urban environment. In addition, your own equipment may generate 60 Hz (or 50 Hz outside of the US) hum due to ground loops or induction from power lines. This is especially true when using any of the popular USB recording interfaces with a laptop computer. The hum may come both form the interface and the laptop itself. Check your laptop's A/C adapter, as many of them are "dirty," generating high levels of noise that leak into your precious recordings. You may want to try the Ebtech Hum X hum eliminating device (reviewed here). It does work wonders for hum originating from ground loops.
Perhaps it is not fair to assess the microphone's performance in a head-mounted configuration because the capsule is only currently available as a lavalier. However, some researchers feel comfortable converting lavaliers to head-mounted microphones with third-party headsets (see Figure 16 below). This is precisely what I did to see whether there would be any appreciable difference in performance. The answer, as expected, is "yes, there is a difference, and it is significant." As you can see in Figure 9, self-noise performance dramatically improves in a headset configuration. It is primarily due to the fact that closer placement requires less pre-amplifier gain, which helps reduce self-noise and improve SNR. The addition of the Tucker-Davis pre-amplifier helps reduce noise even further.
Figure 9. A spectrum of self-noise generated by the chain consisting of the Sanken COS11-D microphone, a Tucker-Davis microphone pre-amp, and the Sound Devices USBPre, calibrated to peak conversational levels of about -12 dBFS, calibrated to the RMS of 70 dB SPL. The spectrum shows marked improvement over lavalier placement.
While the Sanken COS-11D is a truly superb omnidirectional condenser microphone, it does not provide the same amount of spectral detail as head-mounted microphones. It has got nothing to do with the microphone capsule per se, and everything to do with farther placement and the resulting signal attenuation. There is just enough detail in the first and second resonant frequencies, but above F2, spectral detail diminishes rather abruptly.
Figure 10 shows superimposed FFT and LPC graphs illustrating the loss of spectral detail in frequencies roughly above F2. The frequency envelope is unnaturally steep above 1500 Hz, as a result of placement below the talkers mouth (about 20 cm away in this case).
Figure 10. FFT and LPC graphs showing a loss of detail in frequencies above F2
One of my favorite practical tests of spectral detail is LPC analysis/re-synthesis. For the synthesis to be successful, the recording must have exceptionally high level of spectral detail and a very favorable SNR. In the case of the Sanken COS-11D, the detail is simply not adequate for successful analysis and re-synthesis. Figure 11 shows formant tracks of F1, F2, and F3, before (black) and after (red) re-synthesis. The F2 of the Polish word "bordo" was increased by 100 Hz, and, while F1 and F2 appear to be relatively well resolved, F3 shows a significant degree of frame drop-out, which leads to audible artifacts. Simply put, head-mounted microphones are the only viable method of recording speech signals for LPC analysis/re-synthesis.
Figure 11. Ten frames of LPC analysis/re-synthesis of the Polish word "bordo" showing frame drop-outs in F3, resulting in audible artifacts
Sanken COS-11D's performance on the LPC analysis/re-synthesis test dramatically improves when the microphone is used in a headset configuration. As you can see in Figure 12, every analysis frame is perfectly resolved, with no frame drop-outs and no audible artifacts.
Figure 12. Ten frames of LPC analysis/re-synthesis of the English word "love" showing no frame drop-outs, resulting in no audible artifacts - a marked improvement over lavalier placement
Figure 13 shows another example of LPC anlysis/re-synthesis done on samples recorded with the Sanken COS-11D microphone. This time, I modified the F1 and F2 of /ae/ to sound more like the Northern Cities Chain Shift /ae/, to emulate the so-called "ae-raising." Again, the tracks are resolved perfectly, even with the small time-step increment of 0.001 s.
Listen to MP3 at 128 kbps (the first token original; the second re-synthesized):
Figure 13. Ten frames of LPC analysis/re-synthesis of the English word "ad" showing no frame drop-outs, resulting in no audible artifacts - a marked improvement over lavalier placement
As I mentioned earlier, the Sanken COS11-D produces really respectable speech recordings. As you can see in the spectrogram of the Polish phrase "czarna krowa" (Figure 14), there is quite a bit of properly reproduced detail in the lower frequencies (approximately below 1,500 Hz), but due to the far placement below the talker's mouth, there quite a bit of attenuation resulting in the loss of detail.
Figure 14. A spectrogram of a Polish phrase "czarna krowa" showing adequate detail below F2, but some degree of loss in spectral detail in higher frequencies
However, if your analysis involves primarily the measurements of formant frequencies and bandwidths, you should be able to obtain reliable recordings, as evident in the really nicely rendered LPC of /a/ in the Polish word "krowa" (Figure 15).
Figure 15. An LPC of the vowel /a/ in "krowa" showing adequate resolved F1 and F2 peaks
Spectral detail improves dramatically when the microphone is used in a headset configuration, about 2 cm from the talker's mouth (see Figure 17 below). Figure 16 shows a spectrogram of an English phrase "buy a large barrel of good beer." The detail is truly spectacular. If you look closely towards the bottom of the frequency scale, you will notice a gentle low-frequency roll-off, as predicted from the low-frequency test in Figure 7. The roll-off occurs around 50 Hz, and should, therefore, have no appreciable effect on most types of acoustic analysis of speech.
Figure 16. A spectrogram of the English phrase "buy a large barrel of good beer" obtained with the Sanken COS-11D microphone in a headset configuration. The spectrogram shows a significant improvement in spectral detail over lavalier placement.
I am really impressed with the Sanken COS-11D microphone. It sounds fantastic. Speech sounds clear and natural. Such signals are a dream to work with, be it with real-time outboard processors (e.g., EQ, compressors, limiters, and gates) or in post processing. The COS-11D is a perfect interview microphone. It would be excellent for the purposes of oral history and conversational analysis. It is very easy to use, comes with a well-designed holder clip, is resistant to moisture, temperature, and handling noise. Finally, it is very power-efficient and will save your field recorder's precious battery power. The Sanken COS-11D is one of the best lavalier microphones I have ever used. As such, it is a fantastic tool. If you are thinking of buying a lavalier microphone for field recordings of speech, look no further. I am certainly considering replacing my trusty Audio-Technica AT831b with the COS-11D BP.
Having said that, I am not sure I would recommend the Sanken COS-11D BP as the ultimate speech recording microphone for the purposes of speech and hearing research. It simply does not deliver the amount of spectral detail necessary for critical research.
In essence, this review of the Sanken COS-11D BP shows that lavalier microphones, no matter how good they are, are, in principle, not as effective as head-mounted microphones when it comes to capturing rich and unbiased spectral detail of speech. If you mount the Sanken COS-11D BP on a third-party headband, it performs on par with or better than many head-worn microphones, such as the DPA4066 (review). I am convinced, however, that if Sanken provided the same capsule in a head-mounted design, I would not hesitate to use it as my research microphone of choice.
And here's the good news. When I spoke with Mr. Martin Ucik I inquired whether Sanken is planning to release a head-mounted microphone based on the same capsule. It is, after all, a fairly common approach taken by Audio-Technica, DPA, Sennehiser, and others. I found that Sanken does, indeed, have a head-mounted prototype that is currently being tested and might begin shipping later this year. I am really looking forward to this microphone. It's likely to be superb.
Figure 17. The Sanken COS-11D mounted on my headband for head-mounted use