Help me understand: Measuring speakers

ani_101

 I would like to know how to go about measuring speakers. This is from a pint of view of having the box ready and the divers in the box. 

Usually I only measure Garfield (1m) on axis. I tried nearfield once with the mic 0.25 inch from the cone. This was a ND105, which doesn't have a dustcap and I ended up denting the cone as it but the microphone... Didn't measure nearfield since and this was the first speaker I did from scratch that wasn't a kit. I also never understood how to properly measure port response and integrate with the driver response...

So safe to say I am way behind in the measurements department. 

So teach me the basics and discuss the advanced topics, good enough as well as the full nine yards...

I have two microphones a cross spectrum umik and a Dayton umm. No Omni mic, but would like to know what can and can't be done with freeware and omnimic...

Lots of good info in Erin's thread about the ultimate measurement setup, but not all are practically possible...

This thread is from the point of view of measuring a speaker for crossover design and system response evaluation. I'll start a separate thread for measuring, evaluating and selecting drivers for a build and simulating it to determine suitability for a project.

Ron_E

John Kreskovsky's SoundEasy Design Guide has step by step instructions on how to do the measurements you want to do. I'm pretty sure other software can do the kinds of merges you need to do since you don't have Soundeasy. The design guide is a free download.

http://musicanddesign.speakerdesign.net/Guide.html

Ron

dcibel

There's nothing that the Omnimic does that you can't do with other "free" alternatives, the Omnimic is just a plug and play package with relatively easy to use software. The one big feature that Omnimic has that is difficult to replicate outside of tat system is absolute SPL measurement, however for most DIY use this is not a necessary feature as relative SPL is all that is needed for design purpose.

Here's some reference material I've saved in my google drive:

https://drive.google.com/file/d/0B-GpgK57lLMNOTljaGdrNl83WW8/view?usp=sharing

https://drive.google.com/file/d/0B-GpgK57lLMNMWYwajNHWjduaE0/view?usp=sharing

https://drive.google.com/file/d/1cy7F_w76v_WlTi-GE7H2t8rVFergdl5k/view?usp=sharing

https://drive.google.com/file/d/1cy7F_w76v_WlTi-GE7H2t8rVFergdl5k/view?usp=sharing

https://drive.google.com/file/d/1ImTtBEPO5y0KYgb060lBa0tCZwQH_94M/view?usp=sharing

https://drive.google.com/file/d/0B_65c8z6pbn2SGZWbVN5cnVpamM/view?usp=sharing

Also, some good reference material on nearfield measurements here:

https://audioxpress.com/article/measuring-loudspeaker-low-frequency-response

One you've gone through the material and played around with your gear and software and have more specific questions then it's easier to provide assistance.

jhollander

So what measuring system are you going to use? That's probably the first thing to get confidence in. Then we can walk you through the measurements and manipulations.

When I first started I kept a notebook of my steps and a file naming convention so I knew what I did to get the measurements and where I screwed up along the way.

ani_101

I am most comfortable with HolmImpulse. So I'll be found a set of measurements with it. Also want to try REW, as most room corrections can directly integrate with REW. And want to try ARTA for distortion measurements.

Is there any way to set the amp output to a set voltage or measure? So I can do absokute SPK for say 1v or 2.83V?

dcibel

Set amp output for a specific voltage is easily done, you may still have trouble translating the mic input into a specific absolute SPL.

REW, ARTA, Holm are all capable of SPL and distortion measurements, so just pick the one you like the most and stick with it. REW is probably the most fully featured, however the user interface is also probably the most cumbersome of the three.

ani_101

dcibel said:

Set amp output for a specific voltage is easily done, you may still have trouble translating the mic input into a specific absolute SPL.

How do we set amp output for a specific voltage?
Can a spl meter help to determine absolute SPL. In this case is pink noise the best to measure SPL?

dcibel

Play a 50-60Hz tone and measure the amp output with your multimeter. 50-60Hz because this is the frequency where multimeters have greatest accuracy, and make sure the tone is at the same amplitude as any subsequent measurements. You will only be limited by the accuracy of your meter for this purpose.

Check the specs of your SPL meter, most cheap ones are +/- 1.5dB at 1kHz, is that accurate enough for you? You can play whatever noise you want, but I'd calibrate against 1kHz level if I was using a SPL meter for that purpose.

You could also use a speaker that you are 100% confident of the sensitivity as a calibration device. Something like the XT25 with a very smooth response could work well.

Anyway, as I stated in my first post, none of this is necessary for the purpose of speaker design. Only relative SPL between drivers is important, which is easily done by measuring the drivers at the same distance without touching any volume knobs between measurements.

dcibel

Some more reference material, this one is fairly math heavy:

http://www.artalabs.hr/AppNotes/AN4-FreeField-Rev03eng.pdf

ani_101

I am going to continue the voicing thread here...

https://diy.midwestaudio.club/discussion/1417/voicing-problem#latest

I took a few measurements, which i am posting here. The Simulation had 1 ohm in the tweeter circuit, but i found 0.66 ohm to sound the best when the speakers were on the desk close to the wall and 0.82 to be the better when they are out in the room. These affect the response >4k only, so not affecting the BSC - looking at the graph, maybe BSC can be increased a bit, but sound full both on the desk and out in the room, so not much inclined to change it.

what was surprising to me was the values (i went through 1ohm, 0.82ohm, 0.75ohm and 0.66ohm) was that the voice tonality changed between the resistors and was audible. the on axis graphs were pretty similar between the resistors, but off axis, the 1 ohm fell off faster. the 0.82, 0.75 and 0.66 were almost overlapping, but much closer to the on axis graph.

I am inclined to go with the 0.75 as it seemed to sound good both on the desk and out in the room, whereas there is a distinct preference between 0.66 and 0.82 (blind test too) on desk and middle of room. the sample size for blind test was very small, just my wife and me, but the difference was audible...

BUT none sounds good with Uptown funk... 

ani_101

here is the 0 degree comparison between 1 ohm, 0.82 and 0.66

very similar and can't really tell a difference. But 1 ohm sounds flat, 0.82 sounds good in the middle of the room, 0.66 is best on the desk close to walls (12 inches from wall behind speakers)



here's the plots on axis for 0.82, 0.75 and 0.66 ohm - nearly indistinguishable, 0.75 ohm is favored...

ani_101

These are the offaxis. The 15deg and 30 deg are guesstimates, only 0 and 45 and 90 are true, but all measurements were taken together for the different resistor values at each position, so they can be compared.

First up is the 1ohm response at 0, 15, and 30. note the fall off in the higher frequencies.



Here is the 1ohm at 0, 15 and 45...


Compare these to the 0.75 ohm at 0, 15 and 45 degrees


Putting the 1ohm and the 0.75 and comparing at 15 degrees to the 1ohm on axis:
Blue = 1 ohm on axis
Red = 1 ohm 15 deg offaxis
Green = 0.75 ohm 15 deg offaxis
0.75 tracks a bit closer to the on axis till about 12k


This is at 45 degrees offaxis (i have overwritten the measurement for 30 degrees for 0.75)
Blue = 1 ohm on axis
Red = 1 ohm 45 deg offaxis
Green = 0.75 ohm 45 deg offaxis
0.75 again tracks closer to the onaxis than the 1 ohm

ani_101

Since there is a tonality difference between 0.82 and 0.66, here are the graphs for these two values tracked from on axis:

This is 0.82 ohm, 0 degrees, 15 degrees and 30 degrees


This is at 0.66 ohm,  0 degrees, 15 degrees, 30 degrees


Here's 0.82 ohm (blue), 0.75 ohm(red) and 0.66 ohm (green) at 15 degrees


Here's 0.82 ohm (blue), 0.75 ohm(red) and 0.66 ohm (green) at 45 degrees

ani_101

Next step would be to do the vertical offaxis and  then I would like to try to capture the low end frequency response - this is ported, so would like to try both microphone in the box and nearfield response.

Anything standing off in these graphs? The measurement distance was 1M (onaxis, from the tweeter), but the distance to the tweeter would have changed when i rotated the speakers - so the mic was stationary and the speaker was rotated around the box's vertical axis.

rjj45

I would be satisfied with almost all of those graphs - my approach is to get response as flat as possible with as few parts as possible, and try for a nice reverse null at Fc. I find when I dial it in, the best recorded tracks sound great, and the rest sound as good as they might. We can't fix a bad mix!

dcibel

If it were me with my voicing preferences, based on what you've shown I'd be bringing down most everything above 6-700Hz a couple dB, especially for a speaker with such wide dispersion. Your last plot at 45 degrees is closer to what I'd be looking for on axis to 10kHz, minus the bump from 2.5-7kHz.

Something you might try with the Omnimic software is loading in all of you on and off-axis plots, and you can create a weighted average within the software. Apply a higher weight to 0-30degrees than to the wide axis, and see what you get for an average. I found it to be a useful tool vs trying to interpret a bunch of overlaid off axis responses. File menu -> Make weighted average.

rjj45

Also, 5 dB for presentation is good, but for us "insiders", consider working with and posting 2 or 3 dB per division. Shows all the problems better.

ani_101

What's the best way to add in a slight 2db tilt in the high frequencies? The BSC is already at 6db and the woofer is contributing till 6-7khz...

jhollander

Regarding Uptown Funk there's nothing in the graphs to account for your dislike. You would need to listen to each driver separately to see if one is more offending that the other. Or you could target the woofer breakup and push that down with a tank.  I'll try to find the other thread.

jhollander

I didn't find the simulation with the driver responses or the drivers.  Can you post those here?

ani_101

This is with the 2db per division for the 0.75 ohm in 0 deg (Green), 15 deg (red) and 45 deg (blue)

ani_101

jhollander said:

Regarding Uptown Funk there's nothing in the graphs to account for your dislike. You would need to listen to each driver separately to see if one is more offending that the other. Or you could target the woofer breakup and push that down with a tank.  I'll try to find the other thread.

jhollander said:

I didn't find the simulation with the driver responses or the drivers.  Can you post those here?

This is the other link: https://diy.midwestaudio.club/discussion/1417/voicing-problem#latest

jhollander

Regarding voicing preferences I'd keep the dip below 850 and set my max SPL to the 850 SPL value and go flat as best I could.  This is where I find XSim to be very good.

ani_101

Drivers are the SB Acoustics SB12CACS25-04 in MTM with SB21SDC-000-4 dome tweeter

The Simulation:

ani_101

XO schematic:
Tangerine XO V6.pdf

ani_101

PCD file is uploaded in google drive: 

https://drive.google.com/file/d/1hTYEM9MTwx7X_or3PJ0UgcHdRH7NG2il/view?usp=sharing

ani_101

jhollander said:

Regarding voicing preferences I'd keep the dip below 850 and set my max SPL to the 850 SPL value and go flat as best I could.  This is where I find XSim to be very good.

I am not sure I understand. What should i do?

dcibel

You google drive file share isn't made public, I had to send a request for access.

ani_101

dcibel said:

You google drive file share isn't made public, I had to send a request for access.

Updated the permission to anyone with the link. Should work now...

dcibel

Try this for adjustment:

https://drive.google.com/file/d/1F4pipF4bFOa6Xevk13arfFd_Rj7kytS8/view?usp=sharing

You may want to attempt re-measuring your drivers for simulation as I don't see the best agreement between the simulation and your final measurements. It looks like you are working from just a far field measurement as well for the woofer, merging near and far data can make a world of difference for determining that BSC range to see what's going on from 200-800Hz. Just a couple tips as well, remember to enter in the effective driver diameter to PCD so the off-axis sim is accurate, as well try to load in measurements in the 80-90dB range so they appear in the middle of the plot and the zoom buttons don't shift them off the chart. You also haven't entered any X or Y offsets which is important for accurate simulation. 

rjj45

If it was my build, I'd drop 800 and up about 1.5 to 2 dBs, but that's because I am now a stickler for flat as possible.
Agree with dcibel on that.
Really, the CSS guys kicked my ass on the Criton 1TD kit. Plus - Minus 1 dB over a wide range. That's some finesse!