That's a (bad) dynamic pressure-gradient microphone! These are WAY more sensible to vibrations as an omni (=pressure) condenser. (A pressure mic is also less sensible to wind noise btw)
Behringers H3 looks EXACTLY as it should - more steep as H2 and significant lower. Just the other curves are not well to read caus most of it is noise or amplifier distortion.
It's only noise there - does not make sens in this fixture.
It's a pressure chamber - there is no distance to the source!
Actually that's a good test if you are in pressure chamber mode at frequency X - change position and check if the level stays the same.
There is no stereo here? I'm afraid you are thinking about a completely different measurement.
This table is correct. I would use an average for M215 cause the difference is with the sensitivity of your capsule and you don't know what you get - probably an average value.
Btw when you look at the 0,1% value and the 1% value - it's 20dBSpl difference when electronics stay linear. As it should be.
But this one is totally wrong:
You used the values of the graph - but for e.g. H3 this is not THD, it's noise of the fixture! ONLY noise, it has nothing to do with H3. You need to calculate these values!
e.g. M215 at 122dB -> about 0,1%. -> at 102dB 0,01%. 94dB about 0,004% That's for H2. (remember -20dB -> Factor 10. )
H3 of the Behringer mic falls down in a rate of -20dB -> factor 100! What seems to be right, higher harmonics go down way faster.
At 115dBSpl it has 0,05% - that would be 0,0005% at 95dBSpl. You wil never measure this - it's burried in the noise of the microphone! And any electronics adds probably more H3 to the result as that.
M215 - pretty hard cause there is only a small area where measurements seem serious. I would estimate 0,05% at 135dBSpl. 0,0005% at 115dBSpl. 0,000005% at 95dB Spl. Remember - THD produced by the microphone capsule.
H4 - the Behringer already has about 0,001% at 115dBSPL, M215 about 128dBSpl ... it's lower as what a speaker produces. Always.
Conclusion:
There is no influence at these measurements levels from a proper measurement mic at H3 and higher.
(and that's one of the reasons why condenser microphones are pretty good in recording stuff - they can be made really good!)
Be careful - implications based on wrong interpreted acoustical measurements. Way to much of these around the internet ...
IMPLICATIONS
1. Behringer ECM8000- past (based on Panasonic WM61A capsules) and present models are unable to characterize the distortion of ultra low distortion drivers with H2 <0.03% (-70dB) and H3 < 0.01% (-80dB). This microphone has higher distortion than eg. Purifi PTT8.0X - @ 94dB at 1m; H2 < 0.03%, H3 <0.01%. Too much self noise, low signal to noise ratio.
2. MicW M215, depending on the sample, may, be able to characterize these low levels of disortion. The better of the 2 samples certainly can, the slightly poorer performing sample needs to measure a driver at 94dB/1m driver at 31.6cm, in order to observe 104dB, improve the signal to noise ratio.
3. The Earthworks M50 can characterize drivers with distortion H2 <0.03% (-70dB) and H3 < 0.01% (-80dB).
4. The G.R.A.S. 40BD is designed for high frequency, high SPL. Suitable to measuring up to 70KHz, and at higher SPLs.
Thus may be suitable for measuring woofers in the nearfield.
edit: If I've interpreted it incorrectly, would you please show H2/H3 based on SPL levels for each microphone.
I was reading % distortion based on Generator Level...
1. Behringer ECM8000- past (based on Panasonic WM61A capsules) and present models are unable to characterize the distortion of ultra low distortion drivers with H2 <0.03% (-70dB) and H3 < 0.01% (-80dB). This microphone has higher distortion than eg. Purifi PTT8.0X - @ 94dB at 1m; H2 < 0.03%, H3 <0.01%. Too much self noise, low signal to noise ratio.
2. MicW M215, depending on the sample, may, be able to characterize these low levels of disortion. The better of the 2 samples certainly can, the slightly poorer performing sample needs to measure a driver at 94dB/1m driver at 31.6cm, in order to observe 104dB, improve the signal to noise ratio.
3. The Earthworks M50 can characterize drivers with distortion H2 <0.03% (-70dB) and H3 < 0.01% (-80dB).
4. The G.R.A.S. 40BD is designed for high frequency, high SPL. Suitable to measuring up to 70KHz, and at higher SPLs.
Thus may be suitable for measuring woofers in the nearfield.
edit: If I've interpreted it incorrectly, would you please show H2/H3 based on SPL levels for each microphone.
I was reading % distortion based on Generator Level...
Last edited:
Thank you for your correction.
Now @IamJF
0.004% is -88dB down.
When observing 94dB, how is the M215 able to measure 6dB (94-88dB), when it's own self noise is 18dB(A)?
I understand, but it's good to know what we are actually talking about.
Often such numbers are translated back to 1 meter reference.
For 1 meter distance, just add 6dB to all numbers basically to know how loud the source can be.
It doesn't really matter, but again it's important to know what context we are talking about
Okay so basically the graph just flat lines because of the noise?
Do you mean the noise is coming from the test setup or from the microphones/electronics (preamp etc) themselves?
That's the point - you can't. Same with higher harmonics. This means the noise of the mic is higher as the harmonic it produces -> you will not get harmonics of the mic in your measurements.
Did I already say "get a proper mic for THD measurements" ?
(and not a cardioid dynamic ...)(Actually ... when you look at the spectrum ... noise level there is lower as the summed noise from 20Hz to 20kHz so you will very likely see H2 there. So it doesn't get too simple
)
For proper absolute THD measurements in those cases when no 3rd party data is available
https://www.diyaudio.com/community/...hones-actual-measurements.411218/post-7649541
Best we can get with this fixture.
A pressure chamber is pretty different to a normal speaker. You are linear down to 0,1Hz here, resonance frequency is at >300Hz, not 30-40Hz. You loose pressure chamber behaviour at higher frequencies -> frequency response goes down.
It's just a way to produce a LOT of clean sound pressure.
Oh sorry, I missed that one
I know what they are
(was planning to build one myself many years ago)
I am talking about practical numbers here.
Sometimes those numbers are translated to performance at 1 meter.
It doesn't matter which number is being used, as long as it's clear (which it is now)
However, this still doesn't show how they perform above 200Hz
Could be better, could be worse.
edit: A bit like what @TNT is saying (more or less)
It seems there is some confusion about THD over level measurements - so I made some nice graphs.
Let's start with electronics cause it's easier to measure and maybe to understand.
At low levels - most circuits produce VERY little THD. They are linear. Small signal behaviour. But there is noise and when you try to measure harmonics you will find a signal - but it's not the actual harmonic, it's just noise at the same frequency.
When you do a % graph and have a fixed noise level - it get's less % of the signal level when you rise the signal. ....
In other words, when your noise level is fixed and signal level rises, you get higher S/N. It's shown in % instead of dB in this graph.
Here measurements of a Hypex power amp I modified - most of the graph is noise dominated.
At the right part you overdrive your circuit - it can't produce this high level, your rail voltage is not high enough. You run into clipping, THD jumps at that point. Don't do that with a speaker
And then there is a pretty small part with actual THD! This is a highly optimised circuit with a lot of feedback - there is little THD.
In the transition zone we have noise and THD mixed. I estimated noise with the red line and THD with the green line (roughly).
At 7Vrms we measure some noise and some THD. When you want to estimate THD at 5Vrms ... it's ABOUT 0,0001% - and it's masked from noise.
Here a graph from a FET input buffer with no voltage feedback. Less feedback - more THD. Noise on the left, THD in between, overdrive right.
Back to our microphones - M215-2:
We only have valid THD data from 110dBSpl to 135dBSpl. The rest is noise or clipping from the preamp. But this data is reliable!
(as we had proven with our 1/4" mic measurement. This is important to do, otherwise you have no idea what you are measuring!)
Always check your data. Twice at least. Where is my noise floor. Is my result possible in theory? Is it similar to others measurements? What is the behaviour of my measurement devices? What influences the result - how can I minimise this influence? (e.g. the frequency response of the device when you measure THD/IMD over frequency! It needs to be linearised to be comparable.)
For most of you this is already clear but I hope it helped someone to understand these funny colourful lines in graphs like that better.
Let's start with electronics cause it's easier to measure and maybe to understand.
At low levels - most circuits produce VERY little THD. They are linear. Small signal behaviour. But there is noise and when you try to measure harmonics you will find a signal - but it's not the actual harmonic, it's just noise at the same frequency.
When you do a % graph and have a fixed noise level - it get's less % of the signal level when you rise the signal. ....
In other words, when your noise level is fixed and signal level rises, you get higher S/N. It's shown in % instead of dB in this graph.
Here measurements of a Hypex power amp I modified - most of the graph is noise dominated.
At the right part you overdrive your circuit - it can't produce this high level, your rail voltage is not high enough. You run into clipping, THD jumps at that point. Don't do that with a speaker
And then there is a pretty small part with actual THD! This is a highly optimised circuit with a lot of feedback - there is little THD.
In the transition zone we have noise and THD mixed. I estimated noise with the red line and THD with the green line (roughly).
At 7Vrms we measure some noise and some THD. When you want to estimate THD at 5Vrms ... it's ABOUT 0,0001% - and it's masked from noise.
Here a graph from a FET input buffer with no voltage feedback. Less feedback - more THD. Noise on the left, THD in between, overdrive right.
Back to our microphones - M215-2:
We only have valid THD data from 110dBSpl to 135dBSpl. The rest is noise or clipping from the preamp. But this data is reliable!
(as we had proven with our 1/4" mic measurement. This is important to do, otherwise you have no idea what you are measuring!)
Always check your data. Twice at least. Where is my noise floor. Is my result possible in theory? Is it similar to others measurements? What is the behaviour of my measurement devices? What influences the result - how can I minimise this influence? (e.g. the frequency response of the device when you measure THD/IMD over frequency! It needs to be linearised to be comparable.)
For most of you this is already clear but I hope it helped someone to understand these funny colourful lines in graphs like that better.
No, the software didn't allow that ... you have to compare the graphs. It's a little pain but the best I have. (5mV - 108dBSpl. 10mV 115dBSpl. 50mV 128dBSpl. 100mV 135dBSpl)
The graph is from 20Hz to 700Hz?
250Hz. These measurements are at 250Hz.
I showed all the calculations you need and even showed the higher order harmonics with the conclusion that they are extremely low and will never show at measurements.
What else do you want me to do? Maybe read again, it's all there.
Can you not measure distortion products below the noise floor via coherent averaging (or equivalents like long freq sweep times, narrow filtering on harmonic frequencies, etc)? Or is the noise you refer to non-random such as line spurs and the like?
Not that it matters particularly, as you say it (at least the H2 part) should keep dropping at the same slope rate as seen up higher).
Not that it matters particularly, as you say it (at least the H2 part) should keep dropping at the same slope rate as seen up higher).
I think the suggestion about measuring 2-tone IMD (the "stereo" suggestion) was about a way to isolate the fixture distortion from microphone distortion. Two speakers, each producing a different frequency, can cause an IM product in the microphone that doesn't appear in either speaker's harmonic distortion spectrum, depending on how much you can keep the speakers from influencing each other. Don't know how you could do that in a pressure chamber like you are using though.
This is what I used to create the tables:
Eyeballing
1mV: 94dB
3mV: 104dB
10mV: 114dB
30mV: 124dB
100mV: 134dB
So are we not that different? (my reference 100mV -> 134dB)
Then I looked at % H2 and H3:
eg. H2:
H3:
eg. 10mV ->114dB. Behringer H3 = 0.05%
If my numbers in the tables are incorrect by a couple of digits eg. 0.008% instead of 0.006% H3 @104 dB for ECM8000, I probably need to visit my optometrist.
But that was my process. Please advise.
Last edited:
It's the Audio Precision stepped level sweep - there is no averaging or other stuff going on. The good thing is there are no calculation influences on the results. Downside - with some optimisations you could probably get a better S/N range. I wanted to be sure that at least the measurements we get are reliable.
You could try to do something similar with analysing spectra ... but that would be way more work as I spent here already. (cause you would need a lot of points to get a useful graph)
That's the point - 2 speakers in a pressure chamber will influence each other A LOT. So you would first build that chamber with the speakers and try to understand and define how much they influence. Then try if you can go in a useful SPL area with clean enough signal to measure microphone IMD.
I understood the idea but that's a bigger project ...
You could just check with the 1/4" mic if the speaker has low enough IMD for your measurements which is likely cause the membrane doesn't move a lot.
But IMD will behave closely to what THD shows as unlinear behaviour (more THD -> more IMD) ... there is not a lot of information you gain.
There is no doppler effect on a microphone membrane ... or iron/magnet unlinearities. These behave.
Seems correct!
Also correct.
Just don't use measurement points where we measure noise instead of THD.