In search of low distortion omnidirectional microphones for DIYers

Member
Joined 2005
Paid Member
GOAL:
To find a moderate cost microphone that's capable of measuring the lowest distortion drivers, in an indoor environment.

PROBLEM:
Measuring loudspeaker drivers in an indoor environment can be a challenge, with limited space and tools that diyAudio'ers have at our disposal. In this small study I will endeavour to see how well one can measure drivers, in box, not a large IEC baffle, without an anechoic chamber, or laboratory grade instrumentation.

As usual, this a collaboration with other diyAudio users. Thanks in advance to @Hörnli,@5th element and @DcibeL for their direction and feedback. Special thanks to @IamJF for your patient replies and professional advice.

As usual, all errors are mine and mine alone.

I welcome any queries, feedback and contribution.

METHOD:
To start I chose to look at a 6.5” midwoofer, the PTT6.5W04 the first driver released from Purifi's catalogue. As a 2021 release, there may be a few in the field, so it can serve as a kind of reference for a recently released and available driver.

Here's the frequency response, along with H2 and H3, from Purifi's datasheet


1714021214006.png


Comment: Being a 4 ohm driver, the SPL is between 87-91dB through much of the range. As usual, because the driver is not mounted in an enclosure (standard procedure for manufacturers) there's a roll off below 100Hz. In practice, the bass response is determined by the T/S parameters and the enclosure design.

In the first instance, we trace these curves with an auto tracing program, and export into a measurement viewer, so we can work directly with the numbers:
1714791033050.png


In the original datasheet, the 2nd and 3rd harmonics are obscured by the Text Box in the lower left, and the 3rd harmonic (H3) falls below 0dB at 6KHz. Hence, the harmonic data is accurate only between ~60Hz and ~6KHz.

I use software to calculate H2 and H3 in relative terms, compared to the fundamental:
1714791576539.png

Comment: It appears that H2 ~-60dB between 100Hz and 3 KHz and H3 ~-70dB between 200Hz and 3KHz

There are numerous 3rd parties who have reviewed/taken measurements of this driver. Some have been more successful in measuring H2 and H3 eg. Erin's Audio Corner and HifiCompass
1714021736160.png
1714021876721.png 1714798134049.png
Others have not been as successful*
1714022174813.png
*In this measurement, the microphone is 10cm from the driver, and thus tasked with observing up to 121dB.
The microphone's own distortion is 1% at 126dB.

With thanks to @IamJF, it has been shown that a microphones have their own distortion profile . With such a wide variance in maximum SPL and distortion ratings by different manufacturer's, DIYers and enthusiasts might be left wondering "What microphone can give me an honest assessment of a driver?"

For my study, I measure the PTT6.5W04 mounted in a 14L enclosure of dimensions 40x20x30cm (LxWxD), with a drive level of 2.83V.

For far field measurements, I place the microphone 40cm from the baffle. Using gating of 10ms, reasonably accurate frequency response information can be taken from ~ 600Hz and above.
To calculate the 1m SPL, subtract 8dB.

The graph below shows the readings from 4 microphones:
MiniDSP Umik-1 (2013 model)
Sonarworks Xref20 (2015 model)
Earthworks M23 (2024 model)
Sennheiser MD42 (?model year- purchased second hand)

1714638899691.png

Comment: All microphones show similar responses up to about 5KHz, with larger differences starting to show up above, notably Umik-1.

Here is the 2nd harmonic measurement, for each microphone.
1714752564458.png

Comment: There are differences of up to 10dB between microphones in measurement of the 2nd harmonic when observing 98-101dB

The third harmonic for 4 microphones:
1714752645433.png

Comment: There are minimal differences between the microphones in measurement of the 3rd harmonic when observing 98-101 dB

Now we take measurements taken with the microphones in the near field (1cm) to look at low frequency response:

1714786005805.png



1714786103554.png

Comment: There is a difference of up to 10dB between the microphones in measurement of the 2nd harmonic when observing 110-116 dB (50Hz to 300Hz)

1714786113970.png

Comment: There is a difference of up to 6dB in between the microphones in measurement of the 3rd harmonic when observing 110-116dB (between 50Hz and 200Hz)

For a comparison to Purifi's datasheet, I split the graph into 2 parts- the near field for a look at the bass to lower mids…

1714796352731.png


Here is the measured indoors at a mic distance of 1cm with Sennheiser dynamic mic MD42 (“S-mic”)
1714796571426.png

COMMENT: Measurements indoor with mic distance of 1cm give good correlation above 150Hz. Below 150Hz the effects of the enclosure of 1/2 cu ft, there is a difference compared to the infinite baffle measurement by Purifi. It's unclear whether this difference can be reconciled.

Now the second part of Purifi's datasheet for 600Hz to 6KHz…
1714796388579.png


Sennheiser omnidirectional dynamic mic @ 40cm:
1714796762929.png

Compared to an indoor measurement at 0.4m using the S-mic, above 600Hz there is indeed a similar correlation between the S-mic and Purifi's datasheet.


Appendix:
PART 1b:
Measurement imprecision
When we take a measurement indoors, we are prone to all kind of reflections, because sound can act like a wave. If you've ever taken a measurement of your speakers are your listening position, you may have noticed how the frequency response looks nothing like the straight lines from the manufacturer's data. As a countermeasure, we measure at a closer distance, and "block" all those reflections from affecting our measurement. One way to do this is to use a "gate"- to stop the reflections from being interpreted.
With a "gate" of 7 milliseconds, we can block the first echo, which is usually the floor (or ceiling) from affecting our measurement. Perfect right? Well, not so fast. Since we've used a gate of 7ms, the measurement precision is limited to about 150Hz. (142Hz to be exact, since frequency = 1/Time period; T being 0.007 seconds)

This means that EVERYTHING below 142Hz is invalid.
In the octave between, 2-4 KHz, which spans 2000Hz, our resolution is 142 / 2000 = ~1/14th of an octave. Between 500Hz and 2000 Hz, which is 2 octaves and having a numerical span of 1500Hz, we have a measurement resolution of 142/1500= ~1/10 over 2 octaves. And between 142- 425Hz, we're stuck with 1/3 octave resolution. So even though the curve looks nice and smooth on the left side of the graph, there is less accuracy in those lower frequencies.

Let's go back to look at the original graph:

1714100967290.png

What's the immediate consequence of this? The measurement below 600Hz is affected by room reflections (floor bounce)

(NB. To get a complete frequency response measurement, one option is to merge or blend the two parts together. First we, take near field measurements, a technique just 50 years new, by Don B. Keele . In short, we move the microphone up again the cone and take measurements. This way we know what the woofer or port is doing. This method largely ignores what the room is doing. Later, we can even apply some correction factors later to see what it might look like at 1m, or 3m, or 10m, and then blend the low frequency and high frequency measurements together. Here the late Jeff Bagby explains his near and far field blending process)

We must have strategies to measure, even if we had a Klippel Near Field Scanner or anechoic chamber. This is what Wolfgang Klippel does need to take a measurement in his office, which doesn't have space for his Near Field Scanner. He uses a big table for a ground plane measurement of a small consumer speaker:
1714103363037.png

Reference: Acoustical Measurement of Sound System Equipment according IEC 60268-21, Session #4: Simulated standard condition at an evaluation point, slide 8.

So one method is to measure the high frequencies in the "far field", and the low frequencies in the "nearfield". Please note that this term is NOT the same as the same term as that used in pro-audio use eg. nearfield OR midfield monitors

EDIT 04/05/24 : Big clean up, improved precision in tracing Purifi's datasheet (previously 1/12 smoothing, now 1/48 parts per octave without smoothing), separated near field and far field measurements, corrected erroneous graphs, moved explanations to Appendix, added 4th microphone- Sennheiser MD42, nicknamed “s-mic"

To Be Continued… (live article)
-Different microphones to come...
-Implications...
 
Last edited:
  • Like
  • Thank You
Reactions: 8 users
Thanks for continuing with this topic and digging deeper.

As you can see, H3 can be as be low as 10dB. NB. A typical quiet bedroom at night is 30dB, and a TV studio 20dB.
That list is pretty wonky. No chainsaw is 110dBSpl in 1m, that's simply forbidden in Europe. And a PA speaker in the disco is way louder - it's at least 100dBSpl(A) in the audience. Low level noise (and noise measurements for ruleas and goverment cause they don't know better) is mostly A weighted. "20dB" is a nice number but not a SPL level :geek:
With an average level in your home of 50dBSpl(A) ... you will not be very happy ... and I would for sure not be happy with 30dBSpl(A) in my bedroom.


About the topic - I'm still wondering about some of the M23 THD, e.g. H2 at 500Hz is higer as with all the other mics. But it's way lower around 300Hz. It does not fit.
 
Member
Joined 2005
Paid Member
Thanks for continuing with this topic and digging deeper.


About the topic - I'm still wondering about some of the M23 THD, e.g. H2 at 500Hz is higer as with all the other mics. But it's way lower around 300Hz. It does not fit.

You make a good point about the distortion measurement at 500Hz.

Astute readers like yourself are aware of the limitations of taking frequency response (and distortion) measurements in the farfield in a non-anechoic environment.

In fact, the distortion measurements below 800Hz are highly suspect. Furthermore, frequency response below 2KHz, as measured by the (Farina) log sine sweep needs carefully interpretation.

Give me some time to amend the original post to explain why this is the case, in plain language and minimal jargon.
 
I think most of us measure distortion at about 30cm from the driver. Measuring close range does dismiss cone breakup and surround radiation contribution, present in the far field. You could argue about 'nonlinear vs. linear distortion' of course, but imho that is nitpicking. I'd rather know the whole story.

This being not completely relevant to the topic subject of course, but we don't need mikes that do distortion well at 1cm from the cone, unless one wants to examine specific parts of distortion. These are quite irrelevant to those who don't build their own drivers though.
 
  • Like
Reactions: 1 user
Just quickly about noise. A measurement in my living room during the day but with kids in school. :geek:
I'm living in the countryside but 400m to the motorway. Normal windows, nothing special soundproove. Selfnoise of mic + instrument is 19dBSpl(A) - so there is significant influence!

Living room.JPG


As you see with the C rating there is low frequency rumble from the road - but you can't hear that! We are not sensitive for lf at low levels. Therefore the A weighting makes sense in this case.

So noise in a quiet room is quickly as low as most measurement microphone noises.
 
Near field measurements have the advantage of better SNR; ambient noise as well as room reflexions are drowned out by the sound pressure of the DUT. However, as even low far field SPL values are high SPL values in the VNF (very near field), microphone distortion will increasingly become a factor.

In general, realize that not only microphones produce their own distortion and noise, but also your mic preamp, your power amplifier, etc., have their own distortion and noisefloor.

Especially microphone pre-amp noise will become a big issue. You will always want to set the gain of the pre-amp so that it stays just below the clipping threshhold of the pre-amp, as to maximize signal to noise. Afair this can be even done with the Umik, as the Umik has switchable gain settings.
 
  • Like
Reactions: 1 user
The A weighting is relevant wrt speaker measurements. 125Hz is more than 15 dB down, so actual SPL isn't 20 but 35dB. Thus you need well over 95dB level to measure -60dB (0,1%) HD at low frequencies. If 0,1% at low frequencies would be relevant that is, I think that anything below 0,5% is quite OK for sub 100Hz. Furthermore the A-curve descends at high frequencies too, so all of those wanting to measure 3d to 7th order HD have to check their background levels too. I think A-weighting is pretty useless for us.
 

Attachments

  • dba.dbc_.jpg
    dba.dbc_.jpg
    164 KB · Views: 25
The A weighting is relevant wrt speaker measurements. 125Hz is more than 15 dB down, so actual SPL isn't 20 but 35dB.
You can't calculate from the 20-20kHz integrated value to the 125Hz narrow band value! The spectrum values are way lower as the 20-20kHz values.

My quick measurement was just referenced to the noise level table and that real numbers can be very different.
Analysing surrounding noise level for S/N calculations is a total different thing - every full band noise level (A, C, Z, Leq, ...) is pretty useless for that.
 
No I know, did my acoustics and noise courses, be it long ago. Afaik one 3d octave band can be louder than the overall mean level if the rest is hardly contributing. We don’t have to argue about such things, I only wanted to point out the relative uselessness of broadband noise measurements when evaluating conditions for acoustic measurements. Certainly if A weighting is involved.
 
Member
Joined 2005
Paid Member
Well it appears to me that whenever a driver/loudspeaker is being measured, what we are measuring is in fact our complete measurement chain.

From the audio output to the amplifier to the speaker to the microphone back to the input.

So whilst a USB mic is very convenient, they were probably never intended for distortion measurements because they are limited to the built-in interface. And the electret condenser mics are limited by their own self noise eg. 32dB for Panasonic WM61A or distortion limits.

Of course we would all love to have an industrial grade measurement setup like an Audio Precsion or Klippel Near Field Scanner. Maybe there are necessary to measure drivers like Purifi or Bliesma… but that opens a new can of worms (if the drivers are lower distortion than the microphones…)

On the other hand a lot of microphones for the “prosumer” market seem to be for music/content creation/recording where anything goes.

A microphone for DIY distortion measurement would need to be

1) Omindirectional
4) Flat from 20 to 20 KHz( (+/- ? dB)
3) low self distortion of 0.01% when observing-
x dB @1m
Hence max SPL of
(x+34) dB with 0.5% THD
(x+40) dB at 1% THD
(x+50) dB at 3% THD
4) self noise of 20dB? 15dB? Other?
5) Phantom powered
6) closer to $200 instead of $2000

Is this even possible?
Then we’ll have a winner


Of course there are some tricks that may or may not help. Eg. use of repetitions to lower the noise, or stepped sign measurements.

Moving the mic closer will minimise room reflections for clearer measurements of bass frequencies that increases the observed SPL (and distortion) by the mic by 10-25dB.

Is there a way we can “subtract” the microphone distortion, the way we can subtract the microphones non-flat frequency response?
 
Last edited:
Not that I know of. But do we have to? Loads of research papers show that our ‘hearing chain’ is quite indifferent to nonlinear distortions and measured by the popularity of tube amps, fullrange speakers and studio gear adding harmonic products to the original signal I’d say a lot of us even like them. So is that what we have access to not already more than good enough?
 
Member
Joined 2005
Paid Member
Yes you make a good point.

If our measurement imprecision means that all our drivers are much better (by 6-10dB; an order of magnitude) than we previously thought they are, then do we rejoice?

That there's no need for brand B or P drivers, because model K or J has already been available for a long time (decades)

No need for Scan-Speak Ellipticor, Illuminator or even Revelator. In fact the good ol’ classic with the copper sleeve, nylon or polypropylene cones and good old classic like JBL, Altec Lansing, Beyma, RCF and Fane of the last century were already GREAT!

The only thing this century has done has made things smaller. And more affordable. by making them more affordable (great sound for all) it has become also more popular and more disposable (Fast electronics)

OR does it mean that harmonic distortion, intermodulation distortion (two tone or multi-tone), higher order or non-harmonic distortion measurements, amplitude distortion taken by DIYers to be discarded interpreted with a high degree of caution due to poor less than ideal methodology?

eg. microphone already contributing significant distortion, microphone not sensitive enough, measurement affected by reflections, affected by poor signal to noise ratio


1714461834708.png


Reference:
https://www.klippel.de/fileadmin/kl...aningful Tolerances for Signal Distortion.pdf
 
Last edited:
  • Like
Reactions: 1 user

TNT

Member
Joined 2003
Paid Member
Well it appears to me that whenever a driver/loudspeaker is being measured, what we are measuring is in fact our complete measurement chain.

From the audio output to the amplifier to the speaker to the microphone back to the input.

So whilst a USB mic is very convenient, they were probably never intended for distortion measurements because they are limited to the built-in interface. And the electret condenser mics are limited by their own self noise eg. 32dB for Panasonic WM61A or distortion limits.

Of course we would all love to have an industrial grade measurement setup like an Audio Precsion or Klippel Near Field Scanner. Maybe there are necessary to measure drivers like Purifi or Bliesma… but that opens a new can of worms (if the drivers are lower distortion than the microphones…)

On the other hand a lot of microphones for the “prosumer” market seem to be for music/content creation/recording where anything goes.

A microphone for DIY distortion measurement would need to be

1) Omindirectional
4) Flat from 20 to 20 KHz( (+/- ? dB)
3) low self distortion of 0.01% when observing-
x dB @1m
Hence max SPL of
(x+34) dB with 0.5% THD
(x+40) dB at 1% THD
(x+50) dB at 3% THD
4) self noise of 20dB? 15dB? Other?
5) Phantom powered
6) closer to $200 instead of $2000

Is this even possible?
Then we’ll have a winner


Of course there are some tricks that may or may not help. Eg. use of repetitions to lower the noise, or stepped sign measurements.

Moving the mic closer will minimise room reflections for clearer measurements of bass frequencies that increases the observed SPL (and distortion) by the mic by 10-25dB.

Is there a way we can “subtract” the microphone distortion, the way we can subtract the microphones non-flat frequency response?
Could this be the Line Audio Omni-1?

//
 
  • Like
Reactions: 1 user
Member
Joined 2005
Paid Member
Omni1 is a leading contender based on my brief survey...

All figures are from manufacturer spec
italic denotes independent measurement
? denotes derived calculation

bold denotes leading contenders based on above requirements
* denotes MEMS based microphone
BrandModelNoise FloorMaximum SPL based on 0.01% THD
dB10%3%2%1%0.5%0.3%0.1%0.03%
SonarworksSoundID Reference(2024)*29?
Merry ElectronicsMMA209-003*?8413313012812011094
SonarworksXref20/SoundID24?
miniDSPUmik-1??93133
miniDSPUmik-220?125
Line AudioOmni118?99133
iSEMconEMX-715030?96144136130
CPX-1212??155
micWM21518?95135
EarthworksM2320?94
EarthworksM23R20?94140
EarthworksM3020?94140
EarthworksM5020?94140
AudixTM1/Plus28?90130
ACO Pacific7502PH18?90140
PCB Piezotronics376A3140?115165156
376A3322?100150
376A3215?87137
Dayton AudioEMM-624?87127
BehringerECM8000?32?
NTI AudioM201024?92142
M221525?103153
 
Last edited: