Two REW SPL Meter Calibration Issues

Discussion in 'Official REW (Room EQ Wizard) Support Forum' started by markf2748, Jun 15, 2018.

  1. markf2748

    markf2748 New Member
    Thread Starter

    Joined:
    Jun 12, 2018
    Messages:
    6
    Likes Received:
    1
    I am interested in characterizing various component stereo 2.1 systems (no AVR), particularly room effects and subwoofer integration, with REW's SPL Meter. Setup:

    REW V5.19 Beta 9 on Win 10 64-bit fully updated on Dell laptop.
    New miniDSP UMIK-1 USB Microphone.
    Connect laptop audio output to desktop speakers through an ifi nano iOne USB DAC (mostly irrelevant to this topic).
    Cal. Reference: REED R8090 Sound Level Calibrator (purchased 10/2017, accuracy spec +/- 0.5dB, Selectable 1KHz tones at 94 dB & 114dB)
    REW settings:
    Preferences > Soundcard > Driver > Java.
    Preferences > Mic/Meter > Mic or Z weighted SPL meter File: download miniDSP cal. file for my serial number.

    Issue 1
    Preferences > Soundcard > Input > Direct input
    I insert the UMIK-1 into the REED R8090 and use the REW SPL Meter Calibrate Button. This works but the cal is lost when I close down and restart REW. My workaround, which is undesirable, is to find the one particular Windows Microphone Level setting (-4.1dB for my system) which effectively retains the calibration upon restart. For any other WIndows Level setting, I need to painfully recalibrate each time I restart REW.

    Issue 2
    Preferences > Soundcard > Input > MICROPHONE (Master Volume).
    In this case, REW uses the UMIK-1 calibration file to self-calibrate, independent of the WIndows Microphone Level setting. Works nicely, except when I check accuracy, by inserting the microphone into the REED R8090, the REW SPL Meter always displays 99dB instead of 94dB. So REW SPL reads about 5dB too high, assuming my cal source is correct. I have used this cal source with an inexpensive stand-alone SPL meter (American Recorder Model SPL-8810), and have no indication that the cal source is off by 5dB. Is there a setting that needs to be changed within the microphone, within REW, or should I modify the cal file sensitivity value by 5dB?

    Thanks for your attention.
     
  2. John Mulcahy

    John Mulcahy REW Author

    Joined:
    Apr 3, 2017
    Messages:
    990
    Likes Received:
    160
    Level is within a dB of 94 dB when I use my calibrator on several UMIKs. Has your UMIK's internal gain been changed? Default is 18 dB, usually see that figure in the device name.

    If you don't select the actual input device and input REW doesn't know whether whatever input is being used has been previously been calibrated or whether there should be a cal file.
     
  3. markf2748

    markf2748 New Member
    Thread Starter

    Joined:
    Jun 12, 2018
    Messages:
    6
    Likes Received:
    1
    I have not intentionally done anything I know of to change the Microphone internal gain.
    The microphone arrived a few days ago, ordered through Amazon and sold by the Recording Store.

    I always select the device name in the Input Device drop down, which shows:
    Microphone (Umik-1 Gain: 18dB )

    I always select one of the two input options:
    Default Input [raises Issue 1]
    or
    MICROPHONE (Master Volume) [raises Issue 2]

    In the latter case, I believe the miniDSP cal file is being read, because if I create a new cal file with a different Sens Factor value, then the REW SPL dB display value changes.
     
    #3 markf2748, Jun 15, 2018
    Last edited: Jun 15, 2018
  4. markf2748

    markf2748 New Member
    Thread Starter

    Joined:
    Jun 12, 2018
    Messages:
    6
    Likes Received:
    1
    I moved the microphone to another computer (Dell Workstation) with same OS and same REW version. The behavior described above is unchanged.
     
  5. Matthew J Poes

    Matthew J Poes Staff Writer
    Staff Member

    Joined:
    Oct 18, 2017
    Messages:
    1,461
    Likes Received:
    246
    Is the mic fitting tightly in your SPL calibrator? Other people have noted this discrepancy but found it was due to a poor fit. 5dB is a little extreme, not sure I've seen this.

    Maybe I missed this, but did you try setting a speaker as a source and used your SPL meter (after it was calibrated) to compare to the mic? That would help ensure the calibrator isn't the issue.
     
    markf2748 likes this.
  6. markf2748

    markf2748 New Member
    Thread Starter

    Joined:
    Jun 12, 2018
    Messages:
    6
    Likes Received:
    1
    Hi Matthew,

    Thanks for your suggestions.

    The REED R8090 Sound Calibrator is specified to accommodate 1/2 inch diameter sound meters and has an o-ring to assure a snug fit.

    My American Recorder SPL-8810 measures 0.51" dia. and is indeed a snug sliding fit into the R8090. Due to low resolution of the SPL-8810 cal. screw, the closest I can set it to 94.0dB on the meter's LO scale is 95.9, so I then subtract 1.9dB from any of its LO scale readings. For the 114dB calibrator tone, I can adjust the meter to exactly 114dB, so there is no HI scale external correction.

    The UMIK-1 measures 0.47" dia. which is about 1mm undersized. Consequently its fit is slightly loose in the calibrator. If I wiggle the R8090 around at different small angles, it reads 99.1 +/- 0.1 dB. If I pull it out from the microphone by a few mm, the reading varies between 99 and 101 dB. So it is consistently high inside the calibrator, by at least 5dB. This is found for both 94dB and 114dB tone levels.

    I then wrapped black electrical tape, with its edge 1/2 inch from the UMIK-1 tip. As I slid the calibrator up against the edge of the tape, the REW SPL Meter reading dropped from 99.1dB to 95.6dB.

    Next I compare the UMIK-1 and SPL-8810 in "free space" with 1KHz tones from the REW Generator under three scenarios: 1" from a small speaker, 2" from a small speaker, and 3 feet from a separated pair of small speakers. I find the UMIK-1 reads between 2.0 and 2.7 dB higher than the corrected LO scale SPL-8810. For pink noise 3 feet from separated speakers, the UMIK-1 reads about 1.9dB higher than the corrected SPL-8810.

    So for real speaker measurements, the UMIK-1 is perhaps a little better than I expected based on my original 1KHz calibrator tones alone. Given that the SPL-8810 is itself an inexpensive IEC 651 Type II meter, with an accuracy spec of +/- 1.5dB under reference conditions, perhaps the comparison is about as good as I can hope for. I also started a tech support case with miniDSP to see what they say, especially regarding the assumption of calibration at 18dB gain.
     
    #6 markf2748, Jun 16, 2018
    Last edited: Jun 16, 2018
  7. Matthew J Poes

    Matthew J Poes Staff Writer
    Staff Member

    Joined:
    Oct 18, 2017
    Messages:
    1,461
    Likes Received:
    246
    Yeah I only knew that the mic was not a standard 1/2” and that others had bad readings. I use a Dayton USB mic with calibration from across-spectrum and in my case the spl calibration was close to correct. They all have the fit problem so I racked up differences to a poor fit. The 5dB difference seems odd but the <2dB difference in free space seems close enough.

    Unfortunately to have absolute accuracy I think you need to switch to a higher quality microphone and separate audio interface. I’ve had and used both good and cheap mics and these USB mics are fine for their intent but they don’t equal the good mics. A friend of mine has an Earthworks M23 and has been teaching me the CEA-2010 protocol. We have seen some decent differences between the mics and actually hope to do more careful testing in the future. I sold an M30 I had thinking it was a waste of money and now need one again for issues just like this.
     
  8. markf2748

    markf2748 New Member
    Thread Starter

    Joined:
    Jun 12, 2018
    Messages:
    6
    Likes Received:
    1
    Out of insatiable curiosity, I measured the UMIK-1 in the REED R8090 Calibrator again, this time with a snug fitting black rubber o-ring (available in hardware stores) slipped onto the microphone. I push the calibrator up against the o-ring to seal and tilt it such that I barely hear any sound escaping.

    Four trials: O-Ring dimensions in inches, calibrator tones at 94dB and 114dB, REW SPL readings:

    O-Ring # 44 9/16 x 7/16 x 1/16 95.5dB 115.5dB
    O-Ring #9 5/8 x 7/16 x 3/32 95.0dB 115.0dB
    O-Ring # 91 11/16 x 7/16 x 1/8 94.6dB 114.6dB
    O-Ring #1 21/32 x 13/32 x 1/8 94.6dB 114.6dB

    Results for o-rings # 91 and #1, both with the larger 1/8" cross section, read within 1dB of the calibrator values.

    Puzzler: Why does a sealed calibration chamber result in lower, presumably more accurate SPL readings?
    Maybe there is a pressure transducer inside that attempts to keep the sound pressure at a predetermined constant value. If the chamber is open, a feedback loop causes the sound wave generator to undergo larger excursions as it tries to reach the predetermined pressure. Just guessing.
     
    #8 markf2748, Jun 17, 2018
    Last edited: Jun 18, 2018
  9. akl

    akl Member

    Joined:
    Nov 16, 2017
    Messages:
    10
    Likes Received:
    0
    Faulty UMIK cal file
    ---
    My umik shows approx 3 dB too high SPL values. I modified the UMIK to fit into the calibrator. A leakage would result in LOW values not HIGH ones. A faulty cal file allows NO absolute SPL reading, no value to establish a correct LoudSpeaker sensitivity value ( 1 m at 1 W 8 Ohms == +9 dBV ).

    I id not want to change the internal gain setting, so I was not able to use my 124 dB calibrator, but my 93.8...94 dB calibrator. I controlled that cailbrator against several corrected 124 dB calibrators. I asked minidsp for a solution ( to modify the cal file ) but got NO answer yet to this problem (4 weeks) .

    The reference mic and calibrator are Bruel&Kjaer or Gefell class 1 .

    I do NOT understand why minidsp uses such a high standard gain ( 18 dB ) allowing such a low max SPL ( 115 dB SPL ) . Is the internal AD so noisy or ??? But it is 24 bit (approx 133 dB dynamic range) and the noise floor will be not lower than 20 dB SPL.

    Measuring a subwoofer nearfield, a high max SPL (without opening Hte MIC) may be needed.

    The same problem arises with EARS ( the headphone MIC )
     
  10. akl

    akl Member

    Joined:
    Nov 16, 2017
    Messages:
    10
    Likes Received:
    0
    Bruel&Kjaer 4230 notes 93.6 for 1" and 93.8 dB SPL for 1/2" , I think the acoustic impedance ( chamber volume changes and the piezo transducer reacts to this. The pistonphones 4220 124 dB SPL are NOT sensitive to MIC , even to my 1/8" . but they have to be corrected for weather == air pressure ( ... -0,3 dB , at Mount Everest a little more)
     
    #10 akl, Jun 29, 2018
    Last edited: Jun 29, 2018
  11. akl

    akl Member

    Joined:
    Nov 16, 2017
    Messages:
    10
    Likes Received:
    0
    You
    For sensitivity issues you can not use pink noise (FULL range). If you do, you compare frequency response. (The UMIK showing a lower difference may be due to the rising low and high end of the UMIK, it is NOT a sensitivity issue !) You may use band filtered pink noise.
    First exclude problematic parts of the frequency response (of the mics). Also the room and your reference speaker should be not too bad. Near field , you will have no precise comparable positions. You may use 1 m , sometimes probably 0.5 m. Do not disturbe the space around the mic, this will alter the SPL at the mic.

    In a room you also have the influence of the directivity of the NO-directivity-mic. This gets problematic at the high end.

    I use band pass fitlered noise like REW 500--2k, or 250--2k or 125--6k3. If you use periodic noise, you get a steady figure. REW uses brickwall-filters for periodic noise (PN). This is very suitable to compare the sensitivity. Since noise has a higher CREST factor than sinus, true random noise higher than optimised digital periodic noise, watch for measuring well below your max. SPL ( UMIK standard setting max. 115 dB SPL ).
    Measuring the power delivered to the speaker, you have to use a good RMS meter, a standard DVM will not do. It has to measure RMS voltage at high crest values.
     
  12. CherylJosie

    CherylJosie New Member

    Joined:
    Feb 26, 2018
    Messages:
    1
    Likes Received:
    0
    Could be, but somehow I doubt it. Too complicated. The manual trim for atmospheric air pressure (additional quote below) is a clue. If the calibrator was generating a precision pressure level off an acoustic feedback loop it would self-correct for atmospheric pressure when driving a properly mated omnidirectional calibration mike that has a sealed capsule, and there wouldn't be a need for that atmospheric pressure correction factor.

    The calibrator is probably just a precision peizo transducer running open-loop off a precision sinusoidal voltage source. The transmitter only has to generate a single frequency so its fidelity is strictly limited to that frequency and as long as the transducer hasn't aged or been damaged significantly it should remain stable because it's easier to do so at a single frequency than across a broad spectrum, particularly when it's just a resonating crystal that might have been tuned to resonate at 1KHz.

    Acoustic feedback would probably just degrade its performance with control loop error. The feedback will most likely be in the voltage reference that is controlling the frequency and amplitude of the precision sine voltage source because that's simpler and easier. Bandgap references are predictable and stable.

    I have not designed any acoustic references however so I really can't say with authority how it's done.

    That depends on the electro-acoustic properties of the transducers that are generating and receiving the signal, but more importantly (possibly), it depends on what the leak is doing to the SPL inside that chamber as it leaks. Assuming it is all SPL loss ignores several effects of the leak.

    A sealed chamber controls for resonances by eliminating any port action, and every sealed omni mic that is inserted will preserve that seal provided there is a correct fit. As soon as the seal is broken, the port action of that leak can cause resonances and also depending on the size and shape of the leak it could also induce port turbulence. Both of these effects could result in additional SPL being generated off that precision voltage source by drawing extra current with lower transducer input impedance as the Helmholtz resonance bucks the excursion of the transducer, or by altering the efficiency of the transducer at 1KHz with modified acoustic coupling, and also by exaggerated SPL output from the attached mike at '1KHz' plus the additional SPL of whatever random port turbulence frequency component is added across the audio spectrum due to the chaotic airflow at the leak.

    I would expect that the coupling between the receptacle on this calibrator and the microphone body of the SPL meter/measurement microphone should have an inner shoulder to control the length of the insertion, but I don't know because i have never done any of this. If it were my decision I'd add a shoulder to control the depth and preserve the volume of the coupling chamber, that is to at least preserve the volume on the transmitter side of the coupler, for repeatability and accuracy across multiple microphone types and multiple insertions.

    I've only used my Onkyo AVR's 75dB white noise subwoofer input level adjust signal/meter function to 'calibrate' my UMIK-1 in prior measurements, but now that the Linux version of REW supports the sensitivity factor in the UMIK-1 cal file, I am looking forward to comparing the UMIK's internal calibration with the sensitivity provided by MiniDSP against what the AVR is saying is 75dB worth of subwoofer. I am assuming that for simplicity the AVR is also using Z-weighting and that seems a safe assumption but how accurate the reading from the AVR is, well, I don't really care because I am just tuning a home theater and I don't use any nonlinear sound mode functions such as dynamic compression or loudness compensation that rely on a well-controlled reference level.

    I have noted that the UMIK1 input channel gain setting on Ubuntu 16.04 has a center--detent mouse drag behavior at the 'unamplified' and '100%' values, plus adjustment range beyond both detents. The 'unamplified' seems to work well with the 18dB internal DIP settings but I just recently installed my system to a new location and I have not yet tried measuring with the internal sensitivity factor of the cal file controlling the REW calibration gain. In my haste to 'house curve' the frequency response with the manual graphic and a fresh./upgraded Ubuntu/REW install on a new PC, I neglected to add the sensitivity factor to the cal file that Cross Spectrum provides, so REW demanded an SPL calibration before beginning.

    Last time I calibrated, the sensitivity factor was not recognized by that older version of REW on older version of Linux, but then I have been using the Ubuntu distro Java instead of Oracle so maybe that has something to do with the bugs I'm seeing, such as waterfall plots not working and radio buttons not refreshing. Sometimes the audio output from the PC is distorted too, requiring a driver reset, and I worry that if there is also any such corruption of the input signal it might lead to disastrously erroneous measurements that I wouldn't be able to explain and wouldn't otherwise be aware of without distortion analysis because I don't listen to the input from the UMIK-1 before taking a measurement.

    Over at AVS Forum the experts are pointing me to the Cross Spectrum web site that now seems to have language about how operating systems that mix the sound in the sound card driver upset the sensitivity cal with user-controlled mixers plus other processing (including perhaps surround expansion and sample rate adjustment). Cross Spectrum don't warrant the accuracy of that MiniDSP sensitivity factor anyway due to these incidental audio processing steps.

    I'd just like it to be good enough to run with for my home theater cal so I can measure my max SPL capability and long-term dynamic compression of my modest subwoofers in this open floor plan, but for people taking precise speaker sensitivity measurements this internal processing of the OS (regardless of whichever OS is being used) is obviously something to be aware of and control for.

    I hope my musings are of value. Good luck with your measurements.
     
  13. AudiocRaver

    AudiocRaver Senior Admin
    Staff Member

    Joined:
    Nov 21, 2016
    Messages:
    673
    Likes Received:
    158
    Location:
    Lincoln, NE, USA
    Where an electrical measurement is called for having to do with sensitivity, I use an averaging Fluke DVM with REW's brickwall-filtered Pink Periodic Noise, using a band 250 Hz to 500 Hz. The Fluke's response is only down. 0.1 dB at 500 Hz. Results are stable, repeatable, and make sense accuracy wise.
     
  14. akl

    akl Member

    Joined:
    Nov 16, 2017
    Messages:
    10
    Likes Received:
    0
    Averaging
    An averaging meter will show errors, as long as it is not pure sinus and the meter compensates for the difference of averaging to RMS. pink periodic noise will show lower errors than true random noise. Real (every day) noise measurements require very accurate (true) RMS Meters. Allmost NO DVM will do, you need the required frequency range and the ability to handle (very) high CREST factors (and to use defined time constants).

    LEAKAGE
    The Bruel&Kjaer 4230 does NOT have any feedback. I do not know the exact physics, but it does not have to be compensated for air pressure. I tried with different leakages and increased chamber volumes, the level DEcreases.
    Other calibrators behave differently.
    The Bruel&Kjaer 4220 employs two pistons in anti-action. The 4220 has to be corrected for air density/atmospheric pressure.
    All B&K use adaptors with 'shoulders' to control insertion depth/chamber volume.

    UMIK-1 error
    I did some high accuracy measurements (comparing with B&K and Gefell) using 7 pistonphones and two piezos. My two UMIKs show an error of over 3 dB to high ! That is bad and definitely should NOT be. I will propose a different way to set the calibration in the cal - file. The easiest way is to state the max SPL, ie the SPL at 0 dBFS.

    The advantage of USB-mics
    USB mics can be easily calibrated.
    1. there is not additional error of the amp and AD.
    2. it is easy to calibrate the absolute SPL level. (It sould be possible to do SPL-measurement right out of the box, no costly calibrator required ! )
    BUT ... the cal file should be correct, both UMIK-files had been definitely wrong. With an erroneous or no cal-file an USB-MIC is (just) easy to handle but not flexible. The UMIK-1 misses the max SPL to be calibrated with high level pistonphones (117 and 124 dB SPL). An ECM8000 can handle this case (I use RME MIC-Amps-AD-USB-interfaces, easy to calibrate).

    diameter
    Important is a suitable diameter. A MIC with the wrong diameter can NOT used with a calibrator. Some MICs may be modified using thin layers of tape (possibly also by O-rings). Do NOT buy a MIC with the wrong diameter and no adaption. Always (!) think of a way to calibrate absolute level. HIGH accuracy required !


    After (!) correction the UMIK-1 is very easy to handle. With the cal-file correction measurementdata is consistent to Precision MICs/SoundLevelMeters. Though for high SPL and/or 40 kHz/80 kHz ot low noise I use my B&K and Gefell MICs (200 V pol.).
     
  15. Matthew J Poes

    Matthew J Poes Staff Writer
    Staff Member

    Joined:
    Oct 18, 2017
    Messages:
    1,461
    Likes Received:
    246
    This is some helpful information. Thanks for reporting back. Since none of the USB mics fit a standard calibrator, I’m not sure you could call any of them suitable for high accuracy work. Thankfully, what most people are doing relies on relative spl, not absolute level. The accuracy of the absolute level is unimportant.

    Do you think it’s possible that further inaccuracies could come from other variances that you can’t do easily fix. What if the distance between the mic diaphragm and pistaphobe transducer is different than its supposed to be? Pistaphones were designed around mic capsules built to an international standard. These were not. I’ve always wondered how much of an issue this could be.
     
  16. akl

    akl Member

    Joined:
    Nov 16, 2017
    Messages:
    10
    Likes Received:
    0
    1. ABSOLUTE Level ( 94 dB SPL == 1 Pascal )
    For me absolute LEVEL is important ... and should easily be done, like the frequency response calibration. NO .. both require experience, BUT >3 dB error is much to high !
    The power of POWERamplifiers is quite expensive. 1 dB --> 100 W ... 127 W ... 159 W . Building Loudspeakers requires comparisons to external results (sensitivity).
    Also a manufacturer should not state that the cal file calibrates level if it does not. And ... the level calibration is really the most important feature of USB-MICs, at least for me.
    If I measure the noise of a LED bulb, it is important to state the absolute level !
    If I measure the noise of the fans in the PC ...

    2. Usage of calibrators:
    Little NEGATIVE differences in outer diameter of the shaft, may be corrected by some layers of tape. I did this UMIK-1 ECM8000 . For a BEYER MM-1 I bought an adaptor (but the surface of MM-1 is very rough, leaky). I calculate 0.1 dB additional error/uncertainty. Some MICS seem to be very difficult to adapt. I still recommend: do not buy a mic without opportunity to calibrate (at moderate cost or almost NO cost {accurate cal file}).

    standard is 1" 1/2" 1/4" 1/8" not exactly but : .... 13.2 7.0 3.5 mm !
    You have to adapt to these standards.

    easy to adapt:
    UMIK-1 12.0 mm (with several layers of tape 13.4)
    ECM8000 (and similar) 1/2" = 12.7 mm (easily adaptable to 13.2..)

    adapter availability , cost:
    MM-1 9 mm but very rough (with adapter)

    dificult to adapt:
    UMM-6 is 8 mm
    (try 1 or 2 O-rings ??)

    3. PISTONphones
    are very precise and usually longtime stable. Level ist high 117 124 or ... dB SPL. Frequency is NOT 1kHz but 250 or even lower (motor driven 4-cam). Correction required to weather and altitude (absolute air pressure).

    Pistonphones rely on a certain volume. To my experience leakages or increased volume give a little smaller mic level. If you move the mic the level should alter just a little. Otherwise a calibration may be erroneous.

    please refer to uptodate info:
    https://www.bksv.com/en/products/transducers/acoustic/calibrators/4228
    https://www.bksv.com/-/media/literature/Product-Data/bp0881.ashx
    or
    https://www.gras.dk/files/p/d/pd_42aa_ver_07_december_2011.pdf
    https://www.gras.dk/files/m/a/man_42AA_42AA-S1.pdf
    or
    a complicate example how to adapt a special mic
    https://en-de.neumann.com/product_files/1577/download


    4. cal-files
    cal-files have to be improved. the MIC-sensitivity is NOT transparent ... and somrtimes the cal statements do NOT work properly (also REW).
    There should be transparent rules to allow for comments. Sometimes special settings of the MIC-AD-USB interface apply.

    The optimal way to calibrate is to state the max. SPL , i.e. the soundpressure level for 0 dB FS.
    Software and also the human brain may easily use this value. REW shows # dB FS.
     
  17. Matthew J Poes

    Matthew J Poes Staff Writer
    Staff Member

    Joined:
    Oct 18, 2017
    Messages:
    1,461
    Likes Received:
    246
    I think your claims around absolute level really don’t apply to most of the people buying this. They aren’t doing peak dynamic testing. They are using it for eq. That means relative accuracy is fine. If you want absolute precision, that’s is fine.

    After this thread got started I actually bought a lab grade measurement mic and use that. I have a pistaphobe for it and it’s a 1/2” standard. My CSL UMM-6 calibration from CS turned out to be about 2dB’s high.

    ***edited to remove an inaccurate statement.
     

Share This Page