Poor man's frequency response analysis

[crazydoc]

I'm trying to find a cheap (read "free") way to test the frequency response of mics. The purpose would be to compare a mic's response before and after modification, either against an unmodified version or against a reference such as the ECM8000. As such, it doesn't have to be correct in absolute values, but only reproducible for comparison purposes.

I've used the analysis tool in Cool Edit 96 demo version (this is the only recording/wav software I have) to compare several mics, using the noise of my computer fan as a source. Unfortunately it has quite a few peaks and valleys in it.

What I think I need is a white noise source. I've thought of using the air escaping from a tire valve, but can't figure out how to get the car into the house.

Any thoughts on a white noise source? I have a couple of circuits I could build if worse comes to worst.

I'll attach the graph I've done. Sorry about the resolution. The graphs would be a lot easier to compare with even a smooth noise source, not necessarily even white. Also, I tried to put too much in one picture - it all got bunched up at the high frequency end.

Again, any thoughts on a cheap source of noise? Thanks

 

[esactun]

Here's some free stuff, including an oscilloscope, a frequency analyzer, a white/pink noise generator....
and it's all FREE!!! Bwahahahahahaha..................

http://www.maxim.abel.co.uk/utilities.htm

Look for "WaveTools".








BUT.... the white noise will be shaped by the response of your monitors... but for approximation it should be OK. Otherwise you need big $$ signal generator units and B&K test mics, etc. (The ECM8000 has a *slight* boost in the treble per published specs, so they aren't perfectly flat.)

 

[crazydoc]

esactun

That's exactlly what I need, and much more too.

I feel like I've died and gone to cheapass heaven. Thanks so much.

 

[esactun]

esactun

That's exactlly what I need, and much more too.

I feel like I've died and gone to cheapass heaven. Thanks so much.

You're welcome!!! I felt the same way when I dug those up-- cuz after all, I'm a cheap-ass!

 

[knowdoubt]

Also, any cheap Soundblaster card you might have lying around would make a good noise source . Actually I think you need to use the pink noise generator to equate a flat response to a flat reading on the spectrum analyzer. Never mind any nonlinearity in freq response of your monitors. As you said, you can use the ECM8000 as reference & allow for it's slight treble boost. I would highly recommend blocking the sound path to the back & sides of the ECM8000 with as much pillows & sound absorption as possible to eliminate as much as possible any comparison error due to it's omni-directional pattern. Having all mics as close as possible to the monitors & keeping the volume low will also help eliminate room reflection pickup differences being an error factor. Of course you have to beware with a multi-way monitor that your mic placement isn't only or excessively picking up one driver (1 band of freq), so you wouldn't want to get to close & each mic must be tested positioned at exactly the same point in a 3 dimensional space with identical on axis alignment.

Looking forward to seeing your results, especially interested in seeing what you plot for the SP B1. By the way, what color was the color code dot on your B1 box? I'm thinking of ordering a couple & from what little I could tell from what I heard in your other post, I think I feel safe enough to order the same as you recieved.

 

[trdn1]

if you're already using cool edit, you can generate white noise in that program. Go to generate > noise > then white noise. I guess you'd then have to mic your output speakers, which could introduce more issues, but anyway just an FYI in case you didn't already know...

 

[crazydoc]

Looking forward to seeing your results, especially interested in seeing what you plot for the SP B1. By the way, what color was the color code dot on your B1 box? I'm thinking of ordering a couple & from what little I could tell from what I heard in your other post, I think I feel safe enough to order the same as you recieved.

green

I never did understand what these codes were about, as we ran Alan off before he could explain it. Can anyone tell me? Thanks

Thanks for the ideas. I still haven't been able to get a nice smooth reading from a noise source - I think it's the crappy speakers. I'll try changing them first.

if you're already using cool edit, you can generate white noise in that program. Go to generate > noise > then white noise. I guess you'd then have to mic your output speakers, which could introduce more issues, but anyway just an FYI in case you didn't already know...

Duh!
Now that you mention it, I do remember that there are more options in the "generate" menu. I'm at work now so can't check it out, but I'll look at it tomorrow. Thanks for the reminder.

 

[knowdoubt]

crazydoc ;
When I e-mailed Studio Projects asking about the color dots, Alan e-mailed me back that they represented different levels of sensitivity (no sonic/tonal diference) & that all mics of the same color code would match within 0.05db (which is really close tolorance for ordering matched pairs). Odd thing is he stated only 3 color's - black, blue & red (no green)....?? I sent an e-mail in to him today asking what the order of sensitivity rating highest to lowest is pertaining to the color dots. Don't guess it really matters though. How are you liking yours? I'm thinking/hoping these will be great LD drum overheads & all around instrument mics, especially for mid to far field. Also should be good for jazz vocals or any music where the vocals need to sound relatively natural (as is often with acoustic istrumentation style music) & not to hyped.

 

[crazydoc]

Well, I figured out a way to test the freq response of mics. My methods certainly aren't rigorous, and if you see any obvious fallacies or weaknesses, let me know.
Here's a shot of my organized, up-to-date mic testing lab.

I guess I'm going to have to do this in 3 posts as I can only seem to attach one file to each.

 

[crazydoc]

I set up a Behringer ECM8000 mic 15cm in front of a "quality" RS speaker putting out white noise and recorded the signal to a wav file in Cool Edit 96. I did a frequency analysis, and then made a FFT (fast fourier transform) filter to add EQ to the file in an opposite manner to the speaker deficiencies, so that the ECM's freq spectrum was as flat as I could get it (without extensive tweaking.)

This shows both linear and logarithmic views of the same file.
I think that if I add this corrective filter to any white noise signal mic'd through this speaker with the same setup, it should correct for the speaker's abberations and give a comparitive freq curve to the 8000, which I realize is not perfectly flat in reality but has a smooth 2-3 dB rise and fall from 3k-20k, at least per their published specs. (I did test both my 8000's and they were almost indistinguishable.)

 

[crazydoc]

Anyhow, here's some of my poor man's curves. I also tested the CAD GXL2200 with and without bass rolloff, the Behringer B! with and without rolloff and pad, and MXL603s, which I could post if there's any interest.
I was surprised at the 2 high freq dips in the SP B1 - maybe that's why it sounds flatter in the highs to me.
Be all this as it may, it's how it sounds that counts.

 

[knowdoubt]

crazydoc;
Great! You're on the right track but I have a few suggestions for changing your test setup to improve the accuracy & validity of your results. As shown, your speaker & mic is to close to the ground & side wall/book case. As your setup is, off axis reflections from floor & side wall can skew the results especially when off axis freq response of different mics varies. Most likely the mid band (upper bass to lower HF) would be suspect with your current setup. I suggest placing the speaker against a wall (facing out of course) smack dab in the middle of the wall (equal distance from floor & ceiling & equal distance between side walls). This maximizes the distance & reflective angle from all reflective barriers (including back wall at rear of mics) & decreases the relative pickup amplitude of their reflective contribution. If possible some additional assurance would be added by surrounding the immediate test area with a makeshift absorption barrier (curtain of blankets, quilt or foam for instance). Clear all objects away from the test area (nothing that can be within an angle of reflection path that would reach the mics). Also you should repeat the intire test sequence from 1st mic to last mic to compare & varify the results (basically to confirm the validity & consistantcy of the test setup & procedure itself). This all may seem excessive but I bet you would yield noticeably different (if only slight, mostly mid band) & more valid results than those yielded by your current setup. It's important to not combine off axis & on axis signals & response in a test like this.

Yeah, I'm also a little concerned about the 2 HF dips (around 16k & 18k) that your plots show for the B1, but they're narrow band dips & that's pretty high up. It looks pretty good up to 14k (similar to the V67 up to that point). The thing I see in that plot that equates most to the difference I heard between the B1 & V67 in the mp3 comparison you posted is the slight relative dip you see in the B1 around 5k which would equate to a little less sibilance. I'd like to see the Log plots posted as well for all these mics as I find them easier to visualize the overall freq response curve & easier to see whets going on in the bass & midrange.

Keep it coming. Great post.

-Stephen-

 

[darrin_h2000]

Most surround sound recievers have Pink noise output for testing the levels of the speakers. If you can put that on a cd.

 

[crazydoc]

knowdoubt

Thanks for the great review and suggestions! This is the kind of feedback I need.

I can't really put the speaker in the middle of the wall, because the walls all have shit all over them. However, I will try to get a more open area of the room, and put some isolation coverings around the setup as you suggest, then try the tests over again. It will be interesting to see what differences there are.

You're right about the 16k and 18k dips in the SPB1 probably being insignificant - I'm sure my hearing itself is down at least 30 or 40 dB at these frequencies.

I'll post the log plots in a little bit.

I've got to actually try recording some music today. What a strange thing to do.

Thanks again for the feedback. BTW, I love the double entendre of your handle - are you an agnostic?

 

[h kuhn]

I don't know if this will serve you somehow, but here is a wave file of pink noise:
http://www.ethanwiner.com/pinknoiz.wav

it is wave file that you can loop in cooledit (or any other audioeditor)
served me for checking the frequ. response of my speakers

greetings
Harald

 

[crazydoc]

Here's the log plots of the same 4 mics. The freq values are impossible to read. The dog-ear artifacts on all the mics between 2k and 4k also present on the linear plots will give a rough reference.

A couple of you have posted references to pink noise, versus the white noise I have been using. Should I be using this instead? If so, why?

 

[knowdoubt]

knowdoubt
BTW, I love the double entendre of your handle - are you an agnostic?

Pretty insightful of you.

On the pink/white noise thing. In this case I don't think it matters to much, since pink noise is just white noise with a continuous 3db per octave high freq roll off & you have to EQ the speakers anyway. All that's important is to get as close as possible to a flat reading from the ECM8000 (minus it's little high freq boost) so whichever is easiest to EQ to reach that goal. Pink noise would be important if you were trying to calibrate the speakers to put out a spectrally flat equal power response from a spectrally flat equal power signal source, which is not important in this case. All that matters in this case is that the final noise coming out is acoustically flat regardless of how flat or unflat the source noise is. Uhh... does that make any sense or have I confused you as much as I've confused myself?

Anyway, When looking through all the graphs I think your test was at least accurate enough to draw some rough conclusions regarding the freq response of the mics. Testing several mics helped by showing common patterns that show up in all the response plots such as the dog ear pattern (between 2k & 4k) which since it's common & nearly equal on all graphs can be discounted as part of the nonlinearity in the test signal. Same goes for the hump in the bass that's common to all the graphs, though it would also be in the range that would be suspicious of reflective standing waves. Since high freq are beamy I suspect the plot's above 3k are fairly accurate enough for Rock'n Roll, at least referenced to the ECM8000 plot.

Wish I could tell what the freq readings are on the log plot's. I've not used Cool Edi. Are you sure there isn't a way to set it to abbreviate so it's more legible? Oh well one can tell the general range well enough.

-Stephen-

 

[crazydoc]

On the pink/white noise thing......

Yeah, that's what I figured too. In fact, pink noise might be even worse with its high freq rolloff, since with the single cone speaker I'm using I already have to compensate so much at high freqs.

Again, thanks for all your feedback - it was very helpful. If I get this process refined to where it gives more useful results I'll post some more plots.

 

[Mr. Moon]

Bumpage!!

I wanted to bump this post because:
A. there's some great info in it, and
B. because I wanted to see if there are any newer "cheap" (free) tools/techniques people have been utilizing to plot the frequency reponse of their mics.

TIA!

-mr moon

 

[kgirl72]

Very interesting guys...also way over my abilities of total understanding, but I'm going to read this again.

-Kirstin