A Comparison of Some Microphones, Pre-amplifiers and Calibration Files

Note, this is not unique to the Behringer ECM8000, which does not come with a calibration file. The Dayton Audio EMM-6 comes with a unique calibration file based on the specific mic’s serial number, but these have been reported to have unreliable calibration files on some forums. Whether that was limited to certain years or serial numbers or applies to EMM-6 mics in general, I do not know. I am not aware of rigorous studies on other mics and the reliability of their calibration files and I did not perform a comprehensive search to find out.

The Behringer and Dayton mics do not state the tolerance level the mics are manufactured to, or the tolerance of any calibration file. However, from the manufacturers marketing material (in italics) and spec sheets for some other mics:

• The Audix TM1 Plus is known for its linearity. The TM1 is very consistent and has a particularly accurate response with a frequency range of 20 Hz – 20 kHz. Per the spec sheet, this is +/- 2db from 20Hz to 20Khz.

• Beyerdynamic MM1. We manufacture these microphones within an extremely narrow tolerance with a deviation of +/- 1.5 dB between 50 Hz and 16 kHz. As with all microphones in the M series, the MM 1 is measured at the end of the production process. In addition, upon request we will provide the individual calibration data for 0° and 90° free-field measurement via E-Mail.

• The Line Audio OM1 is manufactured to +/- 1db from 20Hz to 20Khz and while it does not seem to be available in the US there is a Canadian distributor listed on their website.

Without a calibration file that you can trust, the true measurements from the Behringer and Dayton are uncertain. There have been no reports of poor calibration files for the Audix or Beyerdynamic mics that I am aware of, ; so I assume that those mics, with a calibration file, produce accurate SPL measurements for speaker design. however, such reports would have to come from either deliberate testing or people transitioning from one microphone to another with know reliability. (Again, I did not perform a rigorous search for other studies, if people are aware of comparisons of other mics, feel free to let me know and I will add links. 

After this write-up was nearly complete, I did find a study similar to this one with some nicely presented findings – see HERE.)

Note, also, that there are other advantages to higher quality microphones, specifically the max loudness they can handle and their ability to accurately measure distortion. Those other benefits, a thorough comparison of the mics and pre-amps concerned, and an analysis of the effects on an actual speaker design are covered in the rest of this study.

 Table of Contents

I. Overview of Mics, Pre-amps, and Software

A. USB Microphones

B. XLR Microphones

C. Audio Interface/pre-amp

D. Measurement Software

E. Summary

II. Testing Setup

A. Speaker Under Test

B. Pre-amplifiers Under Test

1. M-Audio M-Track Duo

2. Steinberg UR22mkII

3. Motu M2

C. Measurement Microphones Under Test

1. Behringer ECM8000

2. Dayton EMM-6

3. Earthworks M23R

D. Comparisons

E. SPL Measurements Using Different Pre-Amps and Different Microphones

F. Distortion – Different Pre-Amps and Different Microphones

III. Analysis – Microphones

A. Calibrated SPL Measurements

B. Calibrated vs. Uncalibrated Measurements

C. Distortion

IV. Implications for a Speaker Design

V. Conclusion 

To design a speaker, you need the following:

1. The ability to measure actual Thiel-Small (“T/S”) parameters. T/S parameters are used to model the speaker enclosure/box/cabinet but are not required for crossover design. Most manufacturers provide T/S parameters, but some are known to be more reliable than others. You may also be able to find reliable T/S parameters from other builders on the audio forums or third-party sites. Relying on incorrect manufacturer’s parameters can lead to a poorly designed box. A commercially available solution is the Dayton Audio Test System (“DATS”) which is now in Version 3, which runs around $130, whereas you can build a DIY impedance jig for a couple of dollars to $30 if you want to add some functionality and a nice enclosure. [See HERE for an article on building a DIY impedance jig and measuring impedance and T/S parameters using REW.]

2. The ability to measure impedance. Impedance measurements will be needed for accurate crossover design. An impedance jig or DATSv3 is the tool for this as well as measuring T/S parameters above. Measuring impedance in the actual box (i.e., a box designed from the actual T/S parameters measured in (1) above) is needed for the woofer and/or midrange that is in an enclosure. Tweeters and some midranges are sealed-back and the impedance can be measured without being installed in the box or possibly traced from a manufacturers data sheet.

3. The ability to measure sound pressure levels (“SPL”) of speaker drivers in a cabinet. To measure SPL, you need a measurement microphone. There are many good crossover simulation programs available (VituixCAD, PCD/WinPCD, Xsim, etc.). These programs are highly accurate if given accurate inputs. They also will produce quite poor results if given poor inputs.

This study is concerned only with the measurement of SPL (item (3) above) and what microphone setup(s) are, or are not, appropriate. The typical measurement microphone is an omnidirectional electret condenser microphone. This microphone can be a USB microphone, simply connected to a computer via a USB cable, or an XLR microphone that connects to an audio interface/pre-amplifier.

I. Overview of Mics, Pre-amps, and Software

A. USB Microphones

USB microphones are practical and inexpensive. USB microphones simply connect to a computer with a single cable, thus are quite practical and avoid the cost of an additional pre-amp. Consider another advantage: no need to calibrate the SPL. However, USB mics do not measure actual phase. There are ways to process the measurements and extract minimum phase, and while these processes are not rocket-science, they are not necessarily simple or practical in some software, which offsets the first advantage to some degree. There are many threads about how important actual phase vs extracted minimum phase is, but that will not be covered here. The more practical point is that the additional and often confusing process of generating reliable minimum phase files adds a step that produces a good chance of introducing errors for a beginner.

Some inexpensive USB mics include:

• • Dayton Audio UMM-6 ($80)

  • miniDSP’s UMIK-1 ($110) and the new upgraded UMIK-2 ($230)

  • Dayton Audio’s OmniMic system ($300) includes a calibrated USB mic, and the included software simplifies some of the processing steps and reduces some of the issues; however, it is still a USB mic that does not measure actual phase. As already mentioned, how critical the severity of that issue is will be left for others smarter than me to argue about.

B. XLR Microphones

XLR microphones for speaker measurements are also omnidirectional electret condenser microphones but connect to an audio interface/pre-amp using a standard XLR microphone cable. In particular, a pre-amp with two inputs, one that measures the direct sound signal through the microphone and a second input used for loopback that records a timing reference and allows for measuring accurate phase. Thus, such a setup is often referred to as a “Dual channel” measurement setup as opposed to a “Single” channel USB microphone that connects with one input and no ability for a loopback signal.

Some XLR microphones include:

• Behringer ECM8000 ($35 (no calibration file)

• Dayton Audio EMM-6 ($55) (includes factory-provided calibration file)

• Sonar Works SoundID ($90) (includes factory-provided calibration file)

• Beyerdynamic MM-1 ($250) (includes factory-provided calibration file)

• Audix TM1 ($250 without calibration file, $300 with calibration file)

• Earthworks M23, M23R ($500, $600)

C. Audio Interface/pre-amp

As mentioned, one of the additional costs of adopting the dual-channel setup is the need for a pre-amp. There are many pre-amps that will work, but it MUST have two microphone inputs to allow for the critical loopback information and it must have 48-volt phantom power.

Also note that for a DIY impedance jig you will need a similar pre-amp unless you have a computer with dual mic inputs. If using a DIY impedance jig rather than DATS ($130), then it is quite possible to assemble a dual-channel setup for less than, or comparable to a USB mic + DATS setup. (Note, you do NOT need a pre-amp to use DATS.)

Some pre-amps include:

• M-Audio M-Track Duo ($70)

• Behringer U-Phoria UMC202HD ($100)

• Steinberg UR22MkII ($170)

• Motu M2 ($200)

• Focusrite Scarlett 2i2 ($200)

Cables: For the dual-channel setup you will need some additional cables.

• 20-25ft XLR microphone cable ($15-$25)

• Short ¼” TRS to ¼” TRS (or TS to TS) cable for loopback. (1.5ft cable approx. $10-$15)

• Cable from pre-amp to power amplifier. This depends on the actual setup but could be TRS/TS-to-RCA or RCA-to-RCA and whatever length you require (approx. $10-$15)

D. Measurement Software

I’m sure there are others out there, but I am familiar with ARTA and REW, both of which are free. ARTA requires a $100 license to save files and if you are running multiple off-axis measurements (for example, 0 to 180 degrees in 10-degree increments), you will need this feature. (ARTA is no longer being maintained as of spring 2024, so I don't know the status of needing a license to access the ability to save measurements. Also note, if you build an impedance jig, either ARTA or REW can measure impedance and generate T/S parameters.

E. Summary

Based on the discussion so far, some reasonable options are:

USB microphones (single channel):

• Dayton UMM-6 and DATSv3: $210

• UMIK-1 and DATSv3: $240

• UMIK-2 and DATSv3: $360

*Note, a DIY impedance jig is not included as an alternative to DATS above because you would need a two-input pre-amp unless you have a computer with two mic inputs.

Dual channel setups:

• Behringer ECM8000 (uncalibrated), M-Audio Duo (open box) and Cables: $128

• Dayton Audio EMM-6, Behringer UMC202HD and Cables: $200

• Beyerdynamic or Audix mic, and Steinberg, Motu, or Focusrite pre-amp, and Cables ($465-$545)

• Earthworks mic and Steinberg, Motu, or Focusrite pre-amp, and Cables ($715-$845) 

*Note, the prices above assume a DIY impedance jig. The DATS is very convenient and may be worth $130 to some people, but for the beginner that is budget-constrained, a modest mic and pre-amp along with a DIY impedance jig is reasonable starting point.

So, what is the difference between a $70 pre-amp and a $200 pre-amp? (1) The accuracy of the reproduction of the relevant bandwidth, (2) distortion, (3) build quality, (4) features, etc. The inaccuracies inherent in measuring SPL with the pre-amp can be removed via a calibration file. Unlike the microphone calibration file, which requires an accurate base measurement in order to create a calibration file, the pre-amp will generate its own baseline via a loopback reference. The distortion of a pre-amp is well below that of a loudspeaker, so while lower distortion matters in many instances (e.g., mixing and recording) it won’t be significant to the measurement of a speaker for crossover design.

For measuring speaker SPL, an inexpensive $70 pre-amp can work just as well as a $200+ pre-amp.

What is the difference between a $35-$80 microphone and a $600 microphone when used for speaker building? (1) The accuracy of the microphone in measuring SPL, (2) the upper limit of SPL that can be measured, (3) the accuracy of the microphone to measure distortion, (4) probably some stuff I don’t worry about. However, while measuring distortion may be important in some circumstances, it is not necessary for crossover design. And, while an accurate SPL measurement is necessary, an uncalibrated or poorly calibrated microphone may be okay, if it can produce accurate measurements with an accurate calibration file. (Although of course, you have to be able to get the accurate calibration file.)

For measuring speaker SPL, an inexpensive $35 microphone can work as well as a $600 microphone, if it has a good calibration file.

In addition to examining the accuracy of SPL measurements, we will look at speaker distortion measurements across different microphones and speakers. This is the where the higher quality mics show their true colors. If you want accurate distortion measurements, you need a higher quality microphone. However, the distortion of the pre-amp is so low relative to the speaker that the pre-amp does not make a material difference in the speaker distortion measurements.

II. Testing Setup

A. Speaker Under Test

Some notes on the testing setup. The tweeter of the test speaker was about 54 inches from the floor. The speaker is 7-1/2” wide by 13” tall with ¾” roundovers on all 4 of the baffle edges, using a 1” soft dome tweeter in a 3D-printed waveguide and a 5” midwoofer. I only measured the tweeter for the pre-amp and mic tests. 

The power amplifier used is a Dayton Audio APA150. For the first measurement I adjusted the amplifier volume and the Steinberg pre-amp output until my external SPL meter read 87db. (This was an inexpensive SPL meter. I was not concerned with absolute SPL, but rather setting a reasonable level and then staying consistent across measurements.) (Consider naming make/model of SPL meter- what is the precision and tolerance of this meter eg. 87.0 dB +/- 0.1dB or 87dB +/- 1dB)   REW was calibrated to this 87db level. I did not touch the amplifier volume after this point. When changing to the M-Audio and Motu pre-amps for the first time, the pre-amp output was adjusted until my external SPL meter read the same 87db, . After this, the pre-amp output levels were never changed. The REW sweeps were taken at -10dbfs for farfield measurements.

I did not calibrate SPL in REW with every change, so the SPL levels indicated in the REW measurements differed quite a bit across the microphones and pre-amps, even though they all measured 87db on my SPL meter. For all of the figures, the absolute SPL levels are adjusted in REW using the SPL Offset settings in Overlays so that they are equal at 1Khz. (Consider expanding on how this was done

Although the mic setup and output levels were generally unchanged, I had to swap mics and pre-amps several times. I probably was able to keep the setup within ¼-inch of the original setup but some of the smallest “squiggles” are probably likely the result of minor differences in the setup and not actual differences in the mics or pre-amps.

B. Pre-amplifiers Under Test

This study will include three different amplifiers. The M-Audio Duo, the Steinberg UR22MkII, and the Motu M2. Why these? Because I already had the Steinberg and Motu, and the M-Audio was cheap enough for me to be willing to spend a few dollars on this project.

1. M-Audio M-Track Duo

I basically searched Sweetwater and Amazon for inexpensive USB audio interfaces, and this came up. Because it is one of the less expensive interfaces, Amazon sells a lot of them. Thus, there were several “Used – Acceptable” ones for sale by Amazon Warehouse, which are basically open-box with some scratches or such on them, available for $47. Mine was described as “Cosmetic imperfection(s) bigger than 1" on top, front or sides of item. Cosmetic imperfection(s) bigger than 1" on bottom or back of item. Item will come repackaged.” Mine actually had no scratches and came in the original M-Audio box. I chose this one because I am genuinely curious if it, possibly the cheapest pre-amp I found, will work acceptably for speaker measurements.

It is cheap plastic and looks pretty cheap. It has the largest fonts, bright lights, and large controls on the top which are actually quite convenient. The other two pre-amps have small knobs (especially the Steinberg) and tiny fonts (especially the Steinberg). The only thing vanishing faster than my ability to hear above 10Khz is my eyesight. The M-Audio has 2 outputs, 2 mic inputs and 48v phantom power and will work for speaker measurement. The highest sampling rate is 48Khz.

2. Steinberg UR22mkII

Built like a tank. Tiny knobs and tiny fonts. It has 2 outputs, 2 mic inputs and 48v phantom power and will work for speaker measurement. (I believe the UR22mkII has been replaced by the UR22C which runs about $195.)

3. Motu M2

Build quality is good; it has some LED displays on the front; it has RCA outputs in addition to ¼” outputs on the back; and it has an on/off switch on the back (whereas the M-Audio and Steinberg are always on when the USB cable is plugged into the PC). Small knobs and small fonts, but the Output knob is large which is the only one I’m usually adjusting. It has 2 outputs, 2 mic inputs and 48v phantom power and will work for speaker measurement. I did not think the LED meters would make much difference, but they are convenient when trying to measure distortion at high SPL levels if clipping is an issue.

C. Measurement Microphones Under Test

Digression on microphone calibration:

When someone asks about measurement microphones, the typical response is to get a calibrated measurement microphone, and then the discussion turns to the single channel vs. dual channel debate. The astute reader will have noticed that twice above it was noted that the Behringer ECM8000 microphone, for $35 (!!), does not have a calibration file. A key question this study is intended to investigate is the importance / implications of an accurate calibration file.

I have seen a few people who do not advocate that a calibrated microphone is that important. I don’t know why, but I can speculate the reason…Most of these omnidirectional electret condenser microphones use the same, or similar condensers with similar properties. These condensers have some roll-off on the top and bottom frequencies, so an uncalibrated measurement may be off by 1 to 2db, but this can be fixed when “voicing” the crossover. However, this assumes a pretty well-trained ear. And if a person is testing the waters of DIY speaker building and asking about how to get started, there is a good chance that this does not apply to them.

But the absence of a calibration file is a little deceiving. The simple solution would be to skip the Behringer and get the Dayton Audio EMM-6 or something similar. But, while these microphones that are “one step up” do have a calibration file (usually) specific to that microphone (based on serial #), is the calibration file accurate? There have been many reported instances of the Dayton microphones calibration file not being very accurate. Cross Spectrum Laboratories used to sell “calibrated” Dayton, Behringer and miniDSP UMIK-1 microphones with a more reliable calibration file at a small premium over retail price, but it is not clear what offerings they still have.

So, a particular aspect of microphones that will be considered is what the measurements look like without a calibration file, with a stock (manufacturer provided) calibration file, and with a more reliable calibration file. We will also look at the difference in the raw driver measurements, as well as what it implies to a competed speaker relying on poorly calibrated vs. accurately calibrated measurements.

1. Behringer ECM8000

These microphones were on back order for quite some time during the supply-chain issues during 2020-21. Coming out of 2021, Sweetwater had them listed for pre-order at $29 so I ordered one to have a cheap “spare” microphone. They are now $35. This microphone does not have a calibration file.

2. Dayton EMM-6

I bought this microphone used from the classified on one of the DIY speaker forums. This may be the most popular of the low-priced alternatives because it is less expensive than anything except the Behringer and does come with a mic-specific calibration file. It currently lists for $80, but Parts Express has sales of 10-12% many times throughout the year. However, the accuracy of the calibration files has been questioned on different audio forums. The prior owner of this specific mic provided the stock Dayton Audio calibration file along with a file created from a more reliable mic.

3. Earthworks M23R

The Earthworks microphones are top quality and unfortunately top price. The M23 mic has has a calibration files that provides “flat” SPL measurements from 3Hz to 23Khz and is $500. The M23R is a reference microphone that provides +/- 0.5db from 3Hz to 23Khz without a calibration file. The M23R is $600. 

D. Comparison of Pre-amps

I am sure for other audio-related work, the differences between pre-amps can matter. But all of these have 2 outputs, 2 mic inputs and 48v phantom power and will work, at least in theory, for speaker measurements. The question to be examined is whether the differences in quality (flatness of SPL and distortion) will matter for speaker design.

Flat frequency response

Note, the measurements below are using the loopback measurement of the pre-amp and do not include any microphone at this point. Figure 4 are the frequency responses, following the Calibration instructions in REW. As the left and right channels were nearly identical for all three pre-amps, only the left channel results will be shown.

• The Steinberg is the flattest. It is down 0.08db at 20Hz and 0.10db at 10Khz and 0.45db at 20Khz. Its -3db points are 1.8Hz and 24Khz.

• The Motu M2 is a little disappointing at the low end. It is down 0.80db at 20Hz and 0.05db at 10Khz and 0.20db at 20Khz. Its -3db points are 8.8Hz and 23Khz.

• The M-Audio Duo is pretty bad compared to the other two. It is down 1.90db at 20Hz and 0.50db at 10Khz and 2.40db at 20Khz. Its -3db points are 14.6Hz and 21.7Khz. 

For speaker measurements and crossover design this generally doesn’t concern me. (1) You can generate a Soundcard Calibration file in REW or ARTA and it will account for these effects. (2) At least on the low end, I’m not terribly concerned about 1db accuracy at 20Hz. What is concerning is if someone is starting out and does not realize the need for the calibration file. (I’m not sure when I realized this, but luckily I would have been using the Steinberg at the time). The M-Audio Duo is down 0.60db at 40Hz which is probably not perceptible but understating the SPL by 2.4db at 20Khz could easily result in a “hot tweeter” if designed to what appeared to be a flat frequency response. (So, if you read this, make sure you calibrate for your pre-amp!)

As seen in Figure 5 above, the Steinberg exhibits the lowest distortion. The Motu is a little disappointing – at a small price premium over the Steinberg it performs worse with respect to flat frequency response below 40Hz and rising distortion above 1Khz. The M-Audio’s distortion rises consistently from 20Hz to 20Khz, exceeding 0.01% from about 1.2Khz to 20Khz. However, speakers are the high-distortion device in an audio system and these differences aren’t going to matter for crossover design. The examination of the effects on the speaker distortion measurements will be seen in Section III.C and IV.C.

E. SPL Measurements Using Different Pre-Amps and Different Microphones

Figures 6 and 7 show the Behringer and Dayton mics at 1 meter from the test speaker, on axis, using each of the three pre-amps. As can be seen, the pre-amps are not making a material difference in the SPL measurement, although the M-Audio pre-amp clearly has some noticeable “squiggles” that are absent when using the Steinberg and Motu. This is likely due to the 48Khz sampling limit. SPL curves have no smoothing applied. Applying 1/24 octave smoothing (which is the default in the VituixCAD REW measurement guide) almost completely removes these squiggles.

Figure 8 below shows the SPL measurements of the Earthworks mic using all three pre-amps. In this case the squiggles when using the M-Audio pre-amp are not minor. Even if I used 1/12th octave smoothing (not shown) it did not make the SPL curve using the M-Audio pre-amp roughly the same as when using the Steinberg and Motu pre-amps.

Looking more closely (Figure 9) there appears to be a reflection at the 6ms point (which is a 3ms window from the initial t = 5ms starting point). This reflection does not appear in any other measurement of any of the mics or any of the pre-amps (in particular either of the other mics but still using the M-Audio pre-amp). I thought maybe I bumped something, or my kids were wrestling upstairs or something, so I repeated this measurement (Earthworks with M-Audio) and got the same result. When I was finished and wrapping up the write-up, about two weeks later, I re-measured it again and it was exactly the same. If this was the actual measurement that I had for crossover design, and without any other measurements for comparison, I would see this reflection and not know that it was some kind of artifact. (Although I might realize something was amiss given the 3ms reflection doesn’t make sense in my measurement setup.) Thus, I would set my window to 3ms instead of 4.75ms and I would get the SPL response as in Figure 10. This looks okay, so I have changed the scaling in Figure 10 to emphasize that while the SPL curve from 400Hz and up are nearly identical, the narrower window provides useful data only down to 350Hz rather than the 200Hz. Merging nearfield and farfield measurements takes place right in this 200Hz to 400Hz range, thus this error would create problems, especially in a 3-way speaker that might crossover around 300Hz to 350hz.

F. Distortion – Different Pre-Amps and Different Microphones

Figures 11 to 13 show THD for the Behringer mic, Dayton mic, and Earthworks mic, using each of the three pre-amps. 

Figures 14 to 16 show the same for third-order harmonic distortion. The pre-amps do not seem to be making a significant difference. It is clear, however, if you pay attention to the y-axis, that the levels of distortion across mics, rather than pre-amps, differs substantially. This is shown in Figure 17, which shows THD for each of the three mics using just the Motu pre-amp.

III. Analysis – Microphones

A. Calibrated SPL Measurements

The testing setup is described in Section II.A above. However, as the focus turns to analyzing the differences between microphones and we leave the study of pre-amps behind, the following measurements are all performed using the Motu M2 pre-amp. To assess the capability of the microphones to measure accurately on-axis, as well as off-axis, angles, measurements were taken at 0, 15, 30, 45 and 60 degrees with each microphone.

The critical importance of the calibration file will be examined in Section IV.B and V. In this section, the Dayton mic measurements reflect a calibration file provided by the original owner (not the stock Dayton calibration file), the Behringer mic reflects a calibration file I created using the Earthworks mic as the reference (created several days earlier using farfield and nearfield measurements from a speaker different than the one being used for these tests). The point of analysis here is, assuming one has a reliable calibration file, inexpensive microphones are capable of basic SPL measurements needed for crossover design. (I don’t think this is a controversial claim, this section just confirms the conventional wisdom.)

Figure 18 shows the three mics measuring the same speaker on axis. 

Figure 19 shows the measurements at 15 degrees off axis. While I measured the remaining angles, the story is the same, the SPL measurements are quite close. 

Thus, I omit the 30- and 45-degree measurements and wrap this section up with Figure 20, measurements 60 degrees off axis.

B. Calibrated vs. Uncalibrated Measurements

This section goes into a little more detail, but in general comes to the same conclusion as Figures 1 and 2 at the beginning of the study - with a reliable calibration file we can get accurate measurements for crossover design using any of the mics. Without a reliable calibration file, you simply do not know whether you have an accurate enough representation of the true SPL to make an accurate simulation. You may be getting within the ballpark of the true SPL. And if you have well trained ears, you may be able to voice the speakers to where they need to be. But for the goal of using the mic as a tool to provide accurate inputs into the process, you simply don’t know unless you have a good idea regarding the quality of the mic and/or the calibration file.

Figure 21 shows the Behringer mic with, and without, a calibration file relative to the Earthworks mic. The uncalibrated SPL is 1 to 2db higher than the Earthworks mic from 4Khz to 14Khz. This implies the Behringer mic is overstating the true SPL measurement. The good news is that this is not a 5 to 6db difference, which would lead to the end speaker being drastically different than what a simulation/microphone would lead the listener to believe. The bad news is that a 1 to 2db difference over almost 2 octaves will be an audible difference, so it cannot just be dismissed flippantly that it is “close enough”.

Figure 22 shows the Dayton mic with, and without, a calibration file relative to the Earthworks mic. The uncalibrated SPL (magenta) is 1 to 2db higher than the Earthworks mic from 6Khz to 20Khz. The green line that is quite close to the Earthworks mic uses the calibration file provided by the prior owner. The cyan line uses the calibration file provided by Dayton Audio. It produces an SPL measurement that is actually worse than the uncalibrated measurement.

I have “heard” that the Dayton and Behringer, and possibly a large number of microphones, use the same condenser. Comparing the uncalibrated measurements of the Dayton and Behringer mics in Figure 23, they are quite close. We’ll see when we get to the distortion measurements, though, that this isn’t true for distortion.

Figure 24 shows the SPL response across a large number of Behringer ECM8000 mics. I found this on the internet, so can’t vouch for its accuracy. But I have seen people apply a “generic” calibration file to uncalibrated, or questionably calibrated, microphones and get reasonable results. The general inaccuracy in Figure 24 is below 300Hz and above 2.5Khz, so applying a generic calibration file and getting reasonable results is not surprising. However, the problem is you only know whether they are reasonable or not if you have an accurate measurement to start with. If you buy an uncalibrated Behringer or even a calibrated Dayton and apply a generic calibration file, you do not know whether your results are “close” to accurate or if you have a mic that is an outlier and you still have severely inaccurate results at the low and high ends of the 20Hz to 20Khz range.

C. Distortion

Distortion tests were measured with the mic 31.5cm from the tweeter at a level of 100db (per external SPL meter) which corresponds to 90db at 1m. I would normally test distortion at a level of 96db at 1m but the M-Audio pre-amp could not reach this level due to clipping, so I maintained the 90db level for all three pre-amps to keep them comparable.

Figure 25 shows the THD measurements for each microphone, using each of the pre-amps. The Dayton mic using the three pre-amps are shown in different shades of green, the Behringer in different shades of blue, and the Earthworks in different shades of red. As can be seen, the pre-amps lead to small changes (roughly 1 to 3db differences) while the mics perform drastically differently – the Behringer measures 10db more distortion than the Earthworks around 3Khz while the Dayton measures 20db (20db!!) more distortion. Figure 26 shows similar results for just third-order harmonic distortion.

To clearly show the results across the mics, Figures 27 and 28 show the THD and HD3 measures from above for only the case of the Motu pre-amp.

IV. Implications for a Speaker Design

So far, this analysis is consistent with other comparisons I have seen and conventional wisdom regarding objective measures of pre-amps and mics. However, the real question is, “does it actually matter for designing a crossover?” This section takes the actual data used in my most recent two-way speaker, which all of the tweeter measurements so far in the analysis have come from. Note, the frequency response, listening window, in-room response, polars, etc. are not some ideal for testing – this is a small budget build using a 5-inch midwoofer and 1-inch soft dome tweeter in a 3D printed waveguide speaker. There is no objective reason for choosing these measurements other than they were the most recent measurements and simulation that I had available, so this was simply the most convenient.

For the actual speaker design, the drivers were measured with the Earthworks mic. Measurements were done horizontally and vertically, zero to 180 degrees in 10-degree increments (76 measurements total). While the analysis up to this point has relied on actual measurements of the tweeter using different mics and pre-amps, the calibration files for the mics do cover 5hz to 40Khz. So, we can now see some of the small effects below 1Khz which weren’t available in the prior figures.

The counter-factual presented below is that all measurements were taken without knowing how accurate the mic or calibration file were. The speaker was designed and completed and taking final measurements with the same mic would just confirm the final design. However, if none of those measurements were accurate measures of the true SPL, then the realized speaker would not sound the way it was intended. To restate the point: the raw drivers would be measured with an inaccurate microphone, those inaccurate measurements would be used for simulation, a crossover would be assembled and the speaker would measure consistently with the simulation - but if you measured with an accurate microphone (and/or listened with your ears) it would NOT look the same the simulation or the inaccurate measurement. The counter-factual to be applied is that the actual measurements were incorrect to the extent that the un-calibrated Behringer mic differs from the Earthworks mic. So, all 76 measurements of the speaker are adjusted by the calibration file. Recall that the uncalibrated Behringer mic inaccurately measures about 1db to 2.5db too high from 4Khz to 14Khz (Figure 21). Thus, designing a speaker based on those measurements will result in an actual speaker that is roughly 1db to 2.5db too low in that range. Here we drop the designed crossover into the speaker with the “true” measurements and compare it to the simulated design.

Figure 29 shows the on-axis frequency response as intended (red-dashed line) and the realized speaker (blue solid line) as measured by an accurate mic or, more importantly, our ears. Note the slightly high output in the 6Khz to 8Khz area of the red-dashed line was intended in the original design because it smoothes out off-axis.

The differences between Figures 29 through 32 are subtle. But it should not be understated that the difference between a bad and good (or preferably good and great) speaker can be subtle. Also, if we go back to Figure 24, we see that most microphones can measure from 300Hz to 2Khz fairly reliably - which is good news considering that covers some of frequencies that are ears are most sensitive to. But without a reliable calibration file, you just do not know what is going on at the lowest and highest frequencies. If you have very well trained ears, you may be able to "voice" this without a reliable microphone. If you are designing your first speakers and you are satisfied with "okay" speakers, then you may be able to achieve this without a reliable microphone. But if you want to design an excellent speaker, based on what you hear with your ears but also confirmed with objective measurements, then you need a microphone with a reliable calibration file. 

V. Conclusion

I have tried to show objective data to this point, but I’ll conclude with some subjective thoughts. To the extent we are more sensitive to peaks rather than dips, if the typical error is that an inexpensive mic produces overstated SPL measurements, then the end-result would be a speaker with that is 1 to 2db less than the simulations suggest, which would at least be preferred to 1db to 2db higher peaks than the simulation. But the problem is that we simply do not know whether our inexpensive or poorly calibrated mic is too high or too low, or off by 1db or 5db. The differences shown above in Section IV don’t look all that different. You could probably hear the difference in an A/B test, but the “realized” speaker probably wouldn’t sound horrible. But again, this is based on one specific speaker and one specific mic that wasn’t too far off.

Microphones are a valuable tool for designing a speaker. As in most things, the better the tools you use, the better the outcome. For cross-over design, to build the best speaker that you can, you need either an accurate microphone or at least a microphone with an accurate calibration file. This does not mean it has to be expensive. If you want to take accurate distortion measurements, then you probably need a higher-quality and more expensive mic. Also, like most tools, you have to buy what is affordable to you. I often hear the advice to buy better quality tools because you will end up paying more in the end if you buy poorer quality tools. This is 100% true. I started with a $16 Dayton iMM-6 microphone, then a UMIK-1, then a Dayton EMM-6, and now an Earthworks M23R. But when I started, the choice wasn’t between a Dayton and an Earthworks mic, it was a Dayton mic or no mic. The idea of paying $100 for a mic to build speakers was pretty hard to swallow at the time. Speaker building is a great hobby. People often refer to “going down the rabbit hole” when getting started. So, buy what you can afford and get started. Yeah, you’ll probably be getting something better down the road. The good news is that mics and pre-amps actually have decent re-sale value.

Lastly, the best tool you have for speaker building is your ears. Train your ears. Look for online programs, YouTube videos, etc.