They are still pressure gradient (capture) at close distances, and in this instance the measurement condition isn't that close (..it's not a true near-field condition). You can also *"calibrate" the result at any point/distance you want (..though I don't know if Joseph did that for that test).
Of course as you get close to the source the freq. response changes regardless..
*note though that calibration effects dynamic range which in turn effects the point where the capsule starts to clip relative to a given freq. (..and a standard 1kHz is likely to be the least corrected freq.).
Last edited:
Here is an interesting reference to mic's from E. Meriläinen's writeup on current drive:
The microphone used was of dynamic cardioid type (Sennheiser e815S), and an appropriate response compensation file was used to yield consistent response with a calibrated condenser mic. It was found that the choice of microphone is critical even at the moderate SPL:s it was subject to (about 100 dB). Budget electret microphones were found useless for this kind of testing due to their inherent distortion. The suitability of any microphone should first be proved with the full test signal by checking that the distortions detected are not inversely related to the measurement distance (as is the case with microphone distortions). The Sennheiser passed this test well and didn’t show any signs of contributing to the results presented.
The document can be found here: https://www.diyaudio.com/community/threads/help-to-understand-current-drive.390405/post-7127198
Seems to be a similar mic to the SM58, and the same price range too, so the SM58 might mot be such a bad choice.
The microphone used was of dynamic cardioid type (Sennheiser e815S), and an appropriate response compensation file was used to yield consistent response with a calibrated condenser mic. It was found that the choice of microphone is critical even at the moderate SPL:s it was subject to (about 100 dB). Budget electret microphones were found useless for this kind of testing due to their inherent distortion. The suitability of any microphone should first be proved with the full test signal by checking that the distortions detected are not inversely related to the measurement distance (as is the case with microphone distortions). The Sennheiser passed this test well and didn’t show any signs of contributing to the results presented.
The document can be found here: https://www.diyaudio.com/community/threads/help-to-understand-current-drive.390405/post-7127198
Seems to be a similar mic to the SM58, and the same price range too, so the SM58 might mot be such a bad choice.
I have only UMIK-1 electret mic from Minidsp and REW. I use provided calibration files, usually on-axis version. Here are noise floor from various measurements of same speaker. Variation obviously comes from room our wind noise. Single driver measurements done in nearfield will have high S/N ratio, but full speaker at 1-2 meters at 100dB and lower will have masking from mic noise. Like I said before, I have tried to do IMD too, but noise was masking too much.
Joseph said that he hasn' found test of audibility of IMD of loudspeakers, and me neither. Anecdotal reports can be ignored and exclusion of other more audible differences is obviously impossible.
Joseph said that he hasn' found test of audibility of IMD of loudspeakers, and me neither. Anecdotal reports can be ignored and exclusion of other more audible differences is obviously impossible.
I think the noise level should not be a problem if you measure close to the driver and use averages? But then again, distortion from the mic could be a problem instead when SPL gets higher when you get closer? How about using the 'harmonic covering multitone' discussed before to determine the noise level of the actual measurement?
-yes, the real problem is the extreme near-field testing typical of distortion measurements below 200 Hz or so which are done that way to reduce (ironically) the noise-floor of the room.
It's basically a "battle" of microphone noise-floor/dynamic range vs. room noise floor (or outside).
IF you had a giant anechoic chamber then it wouldn't be a problem because you would just measure the driver/loudspeaker at 1 meter(+) distance and the mic. wouldn't clip and you could do it with a full-dual-channel setup to remove everything but the mic. itself.
..note that I wouldn't trust anything but a full-dual-channel setup for distortion testing for pure accuracy. (..though it can be fine if you are doing it for comparative results (one-driver-vs.-another) when the test setup is the same.)
It's basically a "battle" of microphone noise-floor/dynamic range vs. room noise floor (or outside).
IF you had a giant anechoic chamber then it wouldn't be a problem because you would just measure the driver/loudspeaker at 1 meter(+) distance and the mic. wouldn't clip and you could do it with a full-dual-channel setup to remove everything but the mic. itself.
..note that I wouldn't trust anything but a full-dual-channel setup for distortion testing for pure accuracy. (..though it can be fine if you are doing it for comparative results (one-driver-vs.-another) when the test setup is the same.)
I just tried some multitone tests on my latest project ('full size 3-way') with the latest version of ARTA, it has more multitone variations now than before, and filters can be applied too. Now I just measured with the filters in the DSP. Tweeter is around -60dB, and mid has a peak at abt -50dB. Ok results I guess, but no wow-factor as with some of RC's measurements.
Measuring the woofer with multitone was not successful, I was not able to get the FFT resolution needed with the max length, with sample frequency set to 44k,so instead I did a two-tone, maybe not with the smartest selection of frequencies though, now that I come to think of it..
The level is what I would call high listening level at abt 3m, maybe normal to others 🙂
Measuring the woofer with multitone was not successful, I was not able to get the FFT resolution needed with the max length, with sample frequency set to 44k,so instead I did a two-tone, maybe not with the smartest selection of frequencies though, now that I come to think of it..
The level is what I would call high listening level at abt 3m, maybe normal to others 🙂
@Rallyfinnen Please check out my post #55 where I provide the settings for low tones. I have to divide my testing into the following groups...
Low Tones (50Hz-5kHz) (ARTA setup with 8000Hz sampling and FFT sequence length of 65536)
85dB
90dB
95dB
High Tones (1kHz - 20kHz) (ARTA setup with 44.1kHz sampling and FFT sequence length of 65536)
85dB
90dB
95dB
For smaller fullrange drivers I lower the test SPL to what is needed to get into the linear range of the driver. For example the Markaudio CHR70 needed 70dB, 75dB, 85dB.
If you want to test full bandwidth from 50Hz to 15kHz then you can set the sample FS to 32000Hz and the sequence length FFT to 131,072.
The room has to be very quiet. I use my laptop for this reason so there are no computer fans in the room. Generally speaking the mic distance should be around 10 to 20cm. (Test signal SPL is at 1m)
Always do a self noise check to confirm your environment and test setup is quiet enough for the test. You do this by changing the test type from multitone to 'external' and pressing play. You can then observe the spectrum noise real time for your test setup. I sometimes do a screen grab just for evidence if anybody questions it. You can confirm noise products are from the speaker if distortion performance does not change if you double the mic distance from the speaker.
Low Tones (50Hz-5kHz) (ARTA setup with 8000Hz sampling and FFT sequence length of 65536)
85dB
90dB
95dB
High Tones (1kHz - 20kHz) (ARTA setup with 44.1kHz sampling and FFT sequence length of 65536)
85dB
90dB
95dB
For smaller fullrange drivers I lower the test SPL to what is needed to get into the linear range of the driver. For example the Markaudio CHR70 needed 70dB, 75dB, 85dB.
If you want to test full bandwidth from 50Hz to 15kHz then you can set the sample FS to 32000Hz and the sequence length FFT to 131,072.
The room has to be very quiet. I use my laptop for this reason so there are no computer fans in the room. Generally speaking the mic distance should be around 10 to 20cm. (Test signal SPL is at 1m)
Always do a self noise check to confirm your environment and test setup is quiet enough for the test. You do this by changing the test type from multitone to 'external' and pressing play. You can then observe the spectrum noise real time for your test setup. I sometimes do a screen grab just for evidence if anybody questions it. You can confirm noise products are from the speaker if distortion performance does not change if you double the mic distance from the speaker.
Sorry, I forgot about the settings you mentioned earlier. I got some strange artefacts when I just tried to lower the FFT frequency before, but now I set the sound card (simple UCA222) to11kHz which was the minimum, and it worked better. Still got some 'odd' measurements (bug?) where the floor jumped up, and the average was destroyed. Seems to do it at the first measurement, if I activate averages with the measurement running, it seems to work. I have no calibrated SPL diff, so I just go by ear, and these were pretty loud, and the noise/sound is disturbing to listen to.
One is abt 10dB louder than the other, I turned the knob on the preamp, so level in the graph is almost the same. There is a lot of 'grass' coming in around 50Hzand then falling toward higher frequencies, I'm guessing overtones from the lowest tones with a lot of excursion? Same on both levels basically.. Any insight on that?
One is abt 10dB louder than the other, I turned the knob on the preamp, so level in the graph is almost the same. There is a lot of 'grass' coming in around 50Hzand then falling toward higher frequencies, I'm guessing overtones from the lowest tones with a lot of excursion? Same on both levels basically.. Any insight on that?
The Lamar M2-16 field coil compression driver (based on the Wester Electric 555) is one of the best compression drivers I've heard. I found it odd that H2 and H3 were moderately high while H4 was significantly low.
H2: -45dB
H3: -57dB
H4: -78dB
Intermodulation was also at least at -65dB limited by my microphone's performance.
So, there's nothing conclusive about this data, but it shows me that there is weak correlation between low harmonic distortion and subjective sound quality, at least when it comes to H2 and H3. It might be possible that some forms of H2 and H3 might be preferable, but I really don't know. There was no lack of clarity or resolution, and the driver has a beautiful sound that is palpable and engaging, and the correlation with H2 and H3 seems intriguing. But again, nothing conclusive.
Just in regards to test setup:
If using USB audio out of your computer, you may have to use the blue USB 3.0 ports. The issue is that USB 2.0 sometimes cannot handle the bandwidth and so it buffers the audio. The same might be true for the USB microphone input (if that is what you are using). Generally I don't see anything suspect or wrong with the spectrum results you've provided.
If you don't have a decibel meter, you can try using the Decibel X app on your smart phone.
Thank you for the tips! I was not aware that there would be any problems with audio in USB 2,I'mnot even sure if the laptop I use for measurements has USB 3..
I use different sound card and setups from time to time. Now I have ben using the minidsp as DAC and the UCA222 as ADC. Mic is the B&K to a dedicated mic preamp. I use a Focusrite 2i2 when I measure amps, but it's a bit bulkier and expensive, the UCA222 is ok for speaker measurements I think.
I downloaded the app, seems pretty neat. I guess you use the Z weighting (full range) to determine SPL for measurements? I have no clue how (well) they can work, guess it depends on the phone manufacturers 'calibration'? Maybe I could check that at work, they have calibrated SPL meters (automotive sound lab). There are some pretty well damped chambers as well, but with concrete floors (and rollers in the floor).
I use different sound card and setups from time to time. Now I have ben using the minidsp as DAC and the UCA222 as ADC. Mic is the B&K to a dedicated mic preamp. I use a Focusrite 2i2 when I measure amps, but it's a bit bulkier and expensive, the UCA222 is ok for speaker measurements I think.
I downloaded the app, seems pretty neat. I guess you use the Z weighting (full range) to determine SPL for measurements? I have no clue how (well) they can work, guess it depends on the phone manufacturers 'calibration'? Maybe I could check that at work, they have calibrated SPL meters (automotive sound lab). There are some pretty well damped chambers as well, but with concrete floors (and rollers in the floor).
I just did a test comparing my Samsung phone with the app, and in the midrange, the diff was more than 10dB (lower on the phone), but running lower frequencies, it was within 1dB. I had to do the comparison with A-weighting, because the meter did not have a flat setting, and the free version of the app has limited choices on weighting, so the only common for the two was A.
I was just doing multitone tests at home for the first time, so I thought I would paste a relatively clean graph to keep the thread alive. I measured tweeter and mid separately, just overlaid the tweeter and then measured mid.
Background noise was a bit lower here, but I think the disturbance around 100-500Hz is still noise.. it's pretty similar for tweeter and mid.
The app was showing around 60dB@ 1m (non weighted curve) for the filtered midrange, whatever that is worth. This is normal background listening level for me, like now when I sit and do stuff on the computer.
Background noise was a bit lower here, but I think the disturbance around 100-500Hz is still noise.. it's pretty similar for tweeter and mid.
The app was showing around 60dB@ 1m (non weighted curve) for the filtered midrange, whatever that is worth. This is normal background listening level for me, like now when I sit and do stuff on the computer.
I was trying to set a higher SPL now, so I held the phone at 1m, increasing the volume. I could se it going up to 80dB, and stopped increasing the volume, and it slowly went down to 70.. Seems like there is some AGC on the mic or something, now I totally lost faith in using the phone for SPL..
I’ve conducted further testing which further highlights the various distortion levels of four different mics.
https://josephcrowe.com/blogs/news/microphone-distortion-comparison
https://josephcrowe.com/blogs/news/microphone-distortion-comparison
Interesting results.
One thing I don’t understand about the test setup is gain. All mics show about the same amplitude, but they should have different sensitivities. Did you change the gain at the mic input for each?
One thing I don’t understand about the test setup is gain. All mics show about the same amplitude, but they should have different sensitivities. Did you change the gain at the mic input for each?
The mic preamp gain was set to the lowest setting based on extensive testing of different levels. The preamp had the lowest distortion when set to the lowest gain, which makes sense.
Looking at the charts at the website I see that the Dayton EMM-6 goes up to about -60 dBFS, whereas the ACO goes up to -50 dBFS. Not sure about the Shure (no pun intended...), where the plot looks like the data might be cut off at -50 dBFS. Overall I can see a difference of the max. signal of about 10 dB between those mics. So, if the electronic gain setting was the same with all measurements, I'd guess the mic sensitivities were different by about 10 dB. Would have to check their datasheets to see if that makes sense.
There's something else that bothers me, though. Looking at the low-frequency data (say below 300 Hz or so), the frequency resolution is much lower for the Dayton UMM-6 and the Shure than for the other mics. This means the "grass" on the baseline is smoothed out more for the Dayton UMM-6 and the Shure than for the other mics. I suspect the difference between the peak signals and the grass (i.e., the IM distortion level) might be affected by this different data treatment.