mic freq response

[mulletchuck]

Howdy.  my CV4 showed up today, and i'm pretty interested in the frequency plot for this guy.    I haven't powered it up yet or recorded anything with it yet, but i thought i'd get some of your thoughts on this plot, cuz i've never seen one like it.

this is what's listed in the manual:

and this is what the plot that mine came with looks like:

 

[PRR]

So what is your question?

The two plots look different because the individual plot has octave-smoothing and is run on +/-70db (!) scale. This makes it look REALLY FLAT.

The flat-to-20Hz is not too unlikely on omni, dubious on directional patterns.

 

[mulletchuck]

ah, you are correct.  the scale is misleading.  I should ask them to email a pic of it using a smaller db scale.      Could you explain how they actually test the mics for freq response?    Do they put it in front of a speaker and fire a swept sinewave at it?  cuz then you'd be at the mercy of the speaker being used, right?   

 

[zebra50]

Do they put it in front of a speaker and fire a swept sinewave at it?  cuz then you'd be at the mercy of the speaker being used, right? 

That is one way to do it. Normally a calibration mic is used to correct for speaker, amp and chamber response.

From the sampling data at the top, this one looks to be a FT (Fourier Transform) method, where the mic is subjected to a burst of noise, or a sweep, and then the frequency information is calculated.

I would treat a response plot like this as a guideline only - it is smoothed, but it does at least show the mic is working! It doesn't tell you about noise level, and of course you will do your own tests.

 

[mulletchuck]

Did some testing last night!  https://www.youtube.com/watch?v=Vklr9CbeQzQ

 

[MicDaddy]

If you look close, the response is up some ~9dB @ 10KHz, far from flat.

Also as others have pointed out, 1/3 octave smoothing has been applied which makes it pretty to look at. 

Enough smoothing applied and almost any microphone will look flat.  That along with the scale they use, cause the plot to appear as if the mic is fairly uniform and flat.

The software that generated the plot is SoundLab from Gold-Line using Time-Delay-Spectrometry or TDS.

TEF is the same format I use as well as other microphone manufacturers.

Take a look at the top you'll see they've also only taken 1024 samples in .9 seconds.  I've seen Telefunken capture as many as 8192 data points.  The more data points the less flat it will appear IME.

It is also nearly impossible to tell at what frequencies they start and stop their sweeps.

Their plot shows down to 20Hz, but I doubt very much that they are able to test accurately to 20Hz (or much below 100Hz), let alone the mic performing as it appears on the plot.

On another note regarding your video, did you simply switch the input on your mac from 'internal mic' to 'interface' during the recording of your short movie?

 

[mulletchuck]

For the video, I'm using the soundflower audio driver which lets you send audio from one app to another.  so, in logic, I have 2 channels set up, one using the built in Mic inputs and the other using my interface.    Theyre set to output to the soundflower outputs.  In quicktime, the audio inputs are the soundflower driver inputs.  So, whatever input in logic is monitored or record enabled is what shows up in the quicktime audio input!  Makes it really easy to sync audio to live video recorded in quicktime.  Soundflower is mac only.  I used the same technique for the other video I did, In a sentimental Mood recently.

 

[ricardo]

This TDS measurement has a freq/time resolution of 2.6kHz/0.38ms.  This is a very large amount of smoothing.

It is meaningless below 2.6kHz.  On top of that 1/3 8ve smoothing has been applied.

About all you can say of the measured mike is that it has a broad peak AT LEAST 6dB high around 9kHz and it is AT LEAST 6dB down at 20kHz.

Today, most mikes are measured by measuring a speaker system with a calibrated measurement mike and then doing another of the DUT at the EXACT same position.

This will be via an impulse response method using a pseudo noise signal like MLS or preferably an exponential sweep.  The frequency responses are obtained by FFT.  The measurement mike response is then "subtracted" from the DUT measurement using complex arithmetic.

The measurements need to be quasi-anechoic, in a room large enough so that the first reflection is as far from the direct sound as possible.

Read chapter 10.4.3 Clio manual for the best explanation.

http://www.audiomatica.com/download/cliomanisa.pdf

I highly recommend Clio for an inexpensive and accurate R&D acoustic measuring system and even more so for production testing.  Their measurement mikes are one of the very few whose calibration I trust.

There are NO cardioids or fig-8s flat below 50Hz.  The most extended cardioids are the Shure SM81 and Hebden Sound.  These have a true 50Hz -3dB response.  If measured at 1m, they will show response flat to 20Hz cos proximity.  All other cardioid curves showing 20Hz flat response are done by the Marketing VP.

 

[PRR]

> I should ask them to email a pic of it using a smaller db scale.

They probably won't. Back in the old days, the plot in the box was the only copy. This is obviously on a computer, and could be saved.... but affordable mikes are made by the thousands. Why would they keep the plots? Why would they pay someone to dig in the logs to find "your" plot?

And as others say: it appears to have been run at very low Measurement resolution. 1024 samples?? My spectrum tool hardly goes that crude.

Computers make some thing easier. The old way was to permanently mount a flat mike near the mike under test and use the flat-mike's output to AGC the sweep to the speaker. This could give a constant 0.1mV 100Hz-5KHz out of the flat-mike by forcing 1V to 10V to the un-flat speaker. The mike under test gets the same flat sound level. Yeah, two adjacent mikes will influence each other... this is production check not precision work. Today we can mount the flat-mike, blip it, save the response, use that to correct the output of the test-mike, AND free-up the precious flat-mike for better things.

> it does at least show the mic is working!

I think that is exactly what it shows.

In the old days we'd run a sweep to an oscilloscope with two lines crayoned onto the screen. Low-pay workers would put each mike in a small sorta-anechoic box and see if the trace fell between the lines. Pass/Reject. Rejects went to QC who would look for the problem to see if it was a fluke or if something was going wrong (sloppy work, out-of-spec parts, missing pieces....)

You usually had two types of rejects. Slight reject (all the tolerances fell the wrong way and it wasn't up to the published spec for this model... if production can't be tightened maybe you announce a new "value"-series with a different model number to get rid of these); and GROSS rejects which are dead or way-way off.

The low-pay worker would see a screen like this (below). Chuck's mike is OK. Someone mis-placed the damping on the next mike, that's a gross reject.

 

[mulletchuck]

i'm very glad i posted this thread, because i have learned a whole lot about what goes into a frequency plot for a microphone!  this forum is AWESOME.    ;D ;D

 

[CJ]

smart low pay workers just draw new crayon lines so they don't have to put a tag on anything, 

happy holidays to all the audio freaks,

 

[MicDaddy]

smart low pay workers just draw new crayon lines so they don't have to put a tag on anything, 

happy holidays to all the audio freaks,

Now there's the elephant in the room    if it doesn't pass the test, change the test. 

Merry Christmas to all!!

 

[Speedskater]

ah, you are correct.  the scale is misleading.  I should ask them to email a pic of it using a smaller db scale.      Could you explain how they actually test the mics for freq response?    Do they put it in front of a speaker and fire a swept sinewave at it?  cuz then you'd be at the mercy of the speaker being used, right? 

This is a paper about how Earthworks measures microphones:

http://www.earthworksaudio.com/wp-content/uploads/how-earthworks-measures-mics.pdf

 

[ricardo]

http://www.earthworksaudio.com/wp-content/uploads/how-earthworks-measures-mics.pdf

This is a good paper but over-the-top unless you need accurate response above 20kHz.  Even then, the spark method they use is notoriously difficult to calibrate properly.

Their concerns about the source used are valid but over-the-top (as are the AES comments)  You can check if your source is OK by simply moving your reference mike 100mm in each direction and comparing curves.

These days, there are loads of good small "mini-monitors" which are suitable for measuring mikes up to 20kHz.  The Infinity M20 is good as is the cheaper L150.  In each case, a little experiment is necessary to find the axis that gives the most consistent measurements (see the "100mm in each direction" above).

The best is a dual-concentric like an old Tannoy 12" DC in a large box.  Modern equivalents would be the modern Tannoy DCs & KEF Uni-Qs but greater care needs to be taken to interpret the LF with these smaller units.

The Clio manual is probably a more relevant tutorial in the 21st century.

In case you're wondering, I've use all the Earthworks methods to measure both speakers & mikes in Jurassic times.

You need a reference mike to do any serious mike work.  The B&K & GRAS 1/2" mikes probably set the standard.  I've been unhappy with ACOS.  At the inexpensive end, I recommend the Clio mikes used with their hardware & software.

If you use a Behringer ECM8000, get it calibrated.  They are pretty far out.

Using Clio or Professor Farina's Aurora extensions to Audition, you can get pretty accurate results with omnis cos most (all?) calibrated measurement mikes are omni.  Accurate results below 200Hz for cardioids & fig-8s is another matter.