Rational Acoustics

Arthur Skudra
September 6th, 2012, 07:16 PM
....when you can run TWO of them together with Smaart 7.4? :D

So here starts my journey in using 15 microphones simultaneously in Smaart, some of you may have seen my postings in Facebook.

So a little background:

Venue: 5000 seat sports arena in Northern British Columbia, Canada (I've never been this far North)! Total of 22 JBL AM5215 loudspeaker arrays, plus concourse speakers, plus private box suite speakers. 24 channels of processing in a BSS London BLU 100 and two BOB's. Amplification using Lab Gruppen CX 4 channel amps. Giving credit where credit is due: Contractor was Sapphire Sound, consultant was Engineering Harmonics. I was hired to commission the system and program the DSP.

Equipment: 2011 MacBook Pro 17" 2.5 GHz Quad Core i7 with 16 GB Ram, 750 GB Hard Drive, OSX 10.8.1 and Windows 7-64 bootcamp. Two Roland Octacaptures, 4 Earthworks M23's, 3 Earthworks M30's, 6 Rational Acoustics RTA-420's, 2 Apex 220's, B&K Type 4231 Calibrator, NTI Minirator MR-2, iAudioInterface2, and a boatload of mic cable and stands!

Here's a few pics of the room:

So on to the next post where I describe setup of the two interfaces....to be continued....

Arthur Skudra
September 6th, 2012, 07:34 PM
A bit about how I approached optimizing this system. First I did the usual verification that all loudspeakers are equal more or less in level, everything in correct polarity, tested the impedance of each speaker line using Dayton Audio's DATS impedance/speaker tester, and of course verify that the configuration file I wrote for the London BLU was working correctly (had that running in the background via Parallels Desktop). I applied the macro supplied by the manufacturer for that particular speaker in my design.

Most of the loudspeaker arrays were two JBL AM 5215-64 stacked horizontally on top of each other (with the exception of 3 arrays where only to top seating section needed coverage, so one speaker only at those locations). Since the room and the system was symmetrical down the long dimension of the room, I decided to leverage that and split the 2 speaker arrays with separate processing for the top and bottom boxes. This allows me to adjust the volume of each speaker to ensure even SPL coverage front to back. It ended up needing a small 2-3 db boost for most arrays for the more distant box, depending on their position in relation to the seating areas.

Then I divided the arena into four, concentrating on the quarter of the room closest to me to determine the best voicing of the system. It wasn't hard to get a representative measurement of this quadrant of the system with 15 mic positions! Then after that, I spread the mics across the entire half of the room, essentially putting one mic per speaker (near throw, far throw) to balance the system in level. Then confirm that the opposite side of the arena measured the same. It took longer to lay the cables out than it did to make the measurements and the appropriate changes in the DSP! Then we put a half dozen mics in the concourse to optimize the speakers there, and time them to the main array. The same with the private box suite speakers. Edit: Oh, and I optimized the center speaker arrays separately as well, using a couple mics on the floor in a ground plane configuration.

Gotta go...train is arriving at the station...

Arthur Skudra
September 6th, 2012, 11:37 PM
So now onto the setup of the Octacaptures....

So a few steps (you can also follow the instructions in the manual):
1. First things first, you need to make sure that the sampling frequencies are set correctly in each unit so that they are both the same. You do this with the front panel controls of the interface.

2. You have two units, one will be considered the timing master, the other sources the word clock from the master. On the timing master Octacapture, set the "VS Expand" to "ON", the other Octacapture this setting remains off.

3. Plug a RCA cable from the output of the timing master octacapture (the one with VS Expand set to ON) SPDIF Out port (output 9/10) to the SPDIF In port (Input 9/10).

4. Now do something that the instructions don't tell you to do: Shut the two interfaces off (push and hold the power button so that it save the configuration changes and shuts off). Then power up the two Octacaptures.

5. Plug the usb cables from the interfaces into your computer, using adjacent ports so they appear on the same controller inside the computer. If you're a mac user, skip the next step and continue on with step 7.

6. For Windows 7 users, you simply need to launch the control panel for the Octacapture, select "Driver Settings" from the menu, and check the box next to "VS EXPAND" and make sure an "OK" appears next to it. Now skip to step 11 below....


7. For Mac OSX users, you need to set up an Aggregate Device to get the two to synchronize together. So go into your Applications folder, open up the "Utilities" subfolder and launch the "Audio MIDI Setup" application.

8. Click on the "+" at the bottom left and add an aggregate device to your list that includes "Octa-Capture" and "Octa-Capture EXP", you can rename the "Aggregate Device" to another name if you wish.


9. Checkmark the Drift Correction box next to the Octa-Capture EXP device in the list for the aggregate device.

10. You're done the setup. Now I strongly recommend that you close all apps, shut down the computer and power off the Octacaptures. Then start up the computer and the Octacaptures. For some reason my computer didn't recognize the aggregate device properly until after a power cycle.

11. Now one last step, launch Smaart 7.4, and in the Options/Transfer Function window, check the box "Allow Multi-Device Transfer Function".


12. Now you can configure your aggregate device as you would any audio interface.

More later......

Arthur Skudra
September 10th, 2012, 11:03 AM
Ok, the continuation....

So now you have a new device in your "Audio Device Options" list... whatever you named your aggregate device!


Note that the two Octacaptures also show up individually as well in that list, make sure you select your aggregate device! Ok, moving right along, you can go into that device and assign friendly names, and apply mic correction curves to the inputs. Note that mic preamps on the first device 1-8 appear as inputs 1-8, SPDIF are inputs 9&10, then inputs 13 -20 are the mic preamps on the second device. Inputs 11 & 12 and 23 & 24 are not used for reasons that are beyond me.


When setting up this many microphones, you'll stumble across a few limitations:
1. There's only so many tabs that can be squeezed in the Measurement Config window! I believe it's limited to a total of 18 tabs including the "group" one.


So what to do? You have two ways to click on those "missing tabs" the first is to click on the "+" you see on the upper right, which will pop up a list of additional tab options:



Arthur Skudra
September 10th, 2012, 11:26 AM
...you can click on the individual engines in the expanded tree on the left:

That wasn't hard, was it?? ;)

Now the second limitation you might stumble across...your computer! I was running Smaart on a new Quad Core i7 2.5 GHz with lots of ram and a hybrid SSD drive. Yep, this is a fast machine!


I ran the Activity Monitor while doing this crazy exercise of measuring a Whirlwind QBox with all 15 mics plugged in and lying on my desk in my office, and an external NTI Minirator MR2 for pink noise, so far so good with just 16 Spectrum engines running simultaneously! The computer is barely flexing it's muscle...

Now add on 15 Transfer function engines....

and 6 averaging engines....

and now the fan in my laptop turns on in a few moments! Still though, I'm only using 540 out of 800 % CPU! Plenty of room to spare to run Parallels desktop in the background and the DSP app!

Now before someone pipes up that my phase curves are all screwy, keep in mind I was doing this test on my desk in my office. I plugged my RTA 420's and Apex 220's directly into the Octacapture with out a mass of mic cables. I was a bit lazy. I did plug my Earthworks mics in via mic cables, only because the barrel of those mics is too thick to plug directly into the interface. So what you're seeing is eight mics imperfectly aimed at a tiny 2 inch speaker sitting on my desk, hence the squirrely HF phase response. Ignore that and concentrate on 250-2000 Hz, close enough for this experiment. No delay drift!

Now on to my next experiment....

Arthur Skudra
September 10th, 2012, 12:15 PM
I copied my new aggregate transfer group, and created a new one, this time using the internal generator loopback. So now I'm using the internal Generator in Smaart!

So I turned on the generator, then all the spectrum engines, the generator stuttered a bit initially, but then settled down just fine (don't know why it stuttered). Then I turned on all the Transfer measurement engines and six average engines, so now we see the laptop running a little harder with the internal generator making an internal loopback, so yes, I can still fit one more mic in there! I think we're running around 580 % out of 800! :)


Note that there's an additional 48-49 ms latency doing the internal loopback, so now you're a bit outside the acceptance window of the delay tracker. So initially it's highly recommended that you manually find the delay times for each measurement engine before doing anything else!

Finally, I wanted to point out the nifty control panel with the Octacapture:

Click on those buttons on the bottom left to choose between the two interfaces, and simultaneously the LCD screen on the Octacapture you're clicking on lights up in reverse contrast. Handy to keep track of which interface is which! Otherwise everything else functions the same as with a single Octacapture, except that you need to choose between the two interfaces.

Also, here's a line diagram on how things are hooked up!
The top octacapture is inputs 1-12, the bottom inputs 13-24.

Arthur Skudra
September 10th, 2012, 12:37 PM
Some closing thoughts...

I tried out the same configuration, this time running Windows 7-64 in Bootcamp on my MacBook Pro, didn't really notice any difference in speed, works just as well!

Lugging all the equipment was a chore! I filled up my roller suitcase with most of my test equipment, one folding mic stand, clothes, weighed in at 49 pounds! Filled a medium sized duffel bag with 1500 feet of mic cable, 49.5 lbs! So those two bags were my checked bag limit of 50 lbs each! I carried on all my mics, laptop, calibrator, ipad, valuable gack, and I ended up with two carry-on bags at 22 lbs each, stretching things to the limit! I couldn't bring anything else with me! The contractor provided at least another 1500-2000 feet of mic cable, AC cable, network cable, and 14 mic stands on site!

Setup was time consuming, but with three of us we got all the equipment, mics placed and wired up in 2-3 hours. We spent a few hours measuring, DSP programming, moving a few mics around, finally got done around 2:00 am in the morning (18 hour workday for me)! Without the multiple mic channels, I'm sure it would have taken us a lot longer to commission something of this complexity. If I were to do it again, I definitely would go wireless for at least a half dozen "roamer" mics, if not more! It was fantastic to be able to establish an "overall" eq of the system within minutes using all 15 mics! Level balance of each array was just as quick!

If I had the time, I would have set up multiple transfer function measurement configuration groups in addition to the one master group of all the mics/engines. Maybe one for each major part of the system. That way my focus is instantly on only the mics in one section of the arena without having to guess which mic is where. There's a lot of potential for what the groups can do. If only these groups could also switch which output the generator appears on my Octa-capture as well (Adam please add this to your list)!

Well the second Octa-Capture has been returned to Long & McQuade....ooopsie, forgot to turn off the VS Expand option!!! :o If you are going to rent an Octa-capture from the Burlington store in the near future, you might want to check that setting!!! :D I gotta say it was an excellent investment of $40 for the week!

Michael Häck
September 10th, 2012, 04:26 PM
Hey Arthur,
great job - sounds very interesting using 15 mics @ the same time! Did you measure the Dayton impedance/speaker tester using v.5? What was your setup for that measurements?

Arthur Skudra
September 10th, 2012, 08:05 PM
Hi Michael,

I used the DATS software that came with my WT3 USB device. It was really easy and straightforward:

1. Calibrate the leads by shorting them together
2. Calibrate using a precision 1K resistor
3. Then measure many speaker lines (this project I had around 45 speaker lines to test).

It measures high and low impedances with equal ease.

For $129, you really can't go wrong with this device! A must have for the serious system commissioner. Very lightweight and compact.

Here's screen shots, one of 2 traces of high impedance, the other of 20 traces overlaid on top of each other (low impedance). It's so sensitive that the impedance was changing ever so slightly because the manager of the rink was cranking his tunes through the other parts of the system while these speakers were under test (which explains the ripple you see on the curves). Lesson learned: do your impedance tests in a quiet room!



Rasmus Rosenberg
September 11th, 2012, 05:10 AM
Awesome work Athur! and thanks for the detailed rapport, very inspiring. Thats DATS software and unit looks like the next thing going in my tool box :) BUT:
I would for one vote that Rational should get a "partnership" up and going with a product like that, small lightweight, usb, and should just use an new "Impedance mode" in Smaart.

Arthur Skudra
September 11th, 2012, 08:05 AM
I would for one vote that Rational should get a "partnership" up and going with a product like that, small lightweight, usb, and should just use an new "Impedance mode" in Smaart.
RasmusYeah, it would be nice not having to boot up in Windows to run special software just to use a fraction of it to make impedance measurements! Obviously the DATS software is focused on the loudspeaker designer, and has a whole bunch of tools in it that I would very rarely use for system commissioning. But it does impedance measurements rather well. I used the impedance measurement capability in Smaart v5 previously, but found it to be too constraining for measuring the very high impedance that you might find in a 70 volt system. It worked fine for lower impedance speakers. Part of the problem was finding a USB interface with a high enough input impedance to make these measurements.

September 23rd, 2012, 05:00 PM
Arthur, can you explain in more detail what the DATS plots are showing and how you use them. Are you just looking for consistency between the traces or something greater? What does the peak on the plot represent? Thanks.

September 23rd, 2012, 09:42 PM
Arthur, can you explain in more detail what the DATS plots are showing and how you use them. Are you just looking for consistency between the traces or something greater? What does the peak on the plot represent? Thanks.

I was wondering the same thing..
if you don't have a reference trace and you aren't testing multiple speakers is there any way to measure just one and know anything useful about it?


Arthur Skudra
September 23rd, 2012, 11:36 PM
Jivey & Jason,

The DATS plots simply show impedance (left vertical scale), phase (upper right vertical scale) over frequency (horizontal scale). The lower plots are impedance, and the upper plots are phase. Please note that the first plot the vertical scale is 0 - 200 ohms, the second plot is 0-100 ohms.

On the first plot, careful study shows that through most of the audible frequency range of the speaker (a 70 volt control series speaker from JBL), the speaker line has a nominal impedance of around 20-40 ohms, indicative of a high impedance transformer distributed system with a lot of speakers in parallel. I compared two similar lines here of almost the same number of speakers, just different sides of the rink. Already having the knowledge that this is a transformer based system, I knew that the impedance would be high (could be even higher depending on the taps used and the number of speakers), and having two to compare gives me increased confidence that things were installed correctly. If I had time, I would have measured one speaker by itself, it should follow the same shape of curve. The "peaks" in the curve show parts of the audible spectrum where greater impedance occurs, typically due to crossovers, tuned ports, and the bandpasses of the box itself. If a 70 volt (or 25, 100, or 140 volt) transformer distributed system was wired incorrectly (eg speaker set to 8 ohms impedance instead of a 1, 2, 5, 10 watt tap) it would have shown quite clearly with a low impedance plot that screams "low impedance" at you. Remember those electrons follow the path of least resistance/impedance, so you're seeing the lowest common denominator here if there is an error in the wiring!

On the second plot, here we have the impedance of two loudspeakers wired in parallel each, a total of about 20 measured lines. You're seeing 6 ohms on the nominal impedance in this plot, consider that you need to factor in the resistance of the wire, which can account for 2 ohms added to what should be 4 ohms (two 8 ohm speakers in parallel) without the cable distance (this is an arena, where cables are hundreds of feet long). Again peaks are due to things like tuned ports, crossovers, and the natural bandpass of the box itself. What this second plot shows is that we have a bunch (20) of loudspeaker lines that show agreement with each other, no short circuits, no loose connections, all following more or less the same pattern. Again, if I had time I could have climbed on a lift and measured the individual speaker directly, it should show more or less the same shape of curve. The "squigglyness" of some of these curves is due to the manager of the arena having a good time playing his tunes through part of the system while I was taking measurements of the other part. The woofers essentially were behaving like "microphones" and altering what would have been a smoother curve. Oh well, I had a tight schedule to follow!

You could ask the manufacturer for the impedance plot of the loudspeaker that you can compare your measurements to if you only have one or two speakers to compare. Otherwise, as I did here, have a whole bunch of measurements (40+ here) to compare to each other for increased confidence that things were installed right. The reason why this is a valuable tool is that when you document these impedances, should something go wrong with a speaker or the wiring, the impedance should change, and now you have a baseline for the way things should be under normal circumstances that you can compare field measurements to. Great for troubleshooting systems. Particularly valuable in checking 70 volt systems, because you will quickly be able to identify a line in which the janitor hooked up a home stereo speaker to because it will show up as a low impedance measurement on your screen. Then you get out your sine wave generator, set it to 60 Hz, connect it to the input of the amplifier, defeat the high pass filter, crank it up a little, the offending speaker should "announce" itself quite readily with the woofer loudly emitting a 60 Hz tone, while the transformer tapped speakers typically don't reproduce that frequency as loudly unless some really big transformers (with a lot of iron inside) were used on each.

Hope this helps!

Ben Clarke
October 3rd, 2012, 08:33 PM
great case study Arthur.

Diligence through procedure is the path to repeatable accuracy!