TOPIC:

I finally have a Dirac tune that I'm happy with, and thought I'd share my process, especially the stuff not covered in the instructions.
My system is 8 channels, run off a JL Audio XD800/8v2 8 channel amp. Front stage all Audiofrog: gb25 tweets, gb40 mids, gb60 midbass. Rear doors have JL Audio ZR800-CW 8" subs.
I decided to do my calibration with just 2 dirac channels (stereo), including the door subs in the mix. One reason for not doing a 3ch (separate mono sub) was that I want my system to be set it and forget it, not wanting to accidentally change the sub level and then not be able to get it back to zero since there is no sub level display other than through a computer. Another reason is simplicity; this way I just deal with one curve screen instead of having a separate sub target and keeping that synced up with my L/R. And finally this way Dirac compensates for the crossover region. Those of you with mono subs could probably do something similar by just mixing both your right and left into the sub in the c-dsp mixer.

As far as computers, the MacOS Dirac program was not working when I wrote this (there are many posts with possible solutions but none worked except perhaps Windows emulators; it seems that Apple changed the security around the audio system breaking the Dirac program and there is no fix yet as of Feb 2020). I originally tried a very old Dell m1330 Windows 10 laptop. I got all sorts of strange developer errors which went away when I strung ethernet to the garage and used a more recent Windows 10 desktop. Moral of the story is that having decent hardware makes everything more reliable (and faster).

Both the C-DSP and especially Dirac programs are somewhat buggy; I sometimes had configurations not load or sync properly, so it’s a good idea to double check that your configs look how you think they should look. Dirac can give weird developer errors when your levels are out of bounds or internet is not constantly connected.

Before getting into the dirac calibration, I set up my system mostly following the audiofrog tuning guide:
testgear.audiofrog.com/wp-content/upload...out-Why-it-Works.pdf
and started with the audiofrog target curve on the bottom of this page:
testgear.audiofrog.com/
I set the time alignment distances using a tape measure (my understanding is this is required even if using dirac), double-checked polarity with an NTI minirator and NTI acoustilyzer. My crossovers were 3.2k/400/70Hz, and I did the trick mentioned in the audiofrog guide of overlapping the sub/midbass crossovers to help get a bass boost. I didn't use the "Bass Management" tab but rather did the sub crossover in the Outputs tab.

I originally tried setting the relative levels with the analog amp pots, but later found it was easier to just leave all the amp pots at the same midpoint, and adjust levels in the c-dsp so I could go back to previous saved configuration and calibration files. Since the DSP has 24 bits inside this should not cause extra noise. To get the levels set, I used the REW program with my u-mik1 (which incidentally I got calibrated from Cross Spectrum cross-spectrum.com/measurement/calibrated_umik.html).
I started with all the levels in c-dsp plugin zeroed, except +12db in the mixer tab from dirac to sub. I ended up also having +5db in mixer from dirac to midbass, with 0 for tweets and mids.

Tip: if you don’t hear any sound when listening, press the volume button on the controller to unmute; it’s easy to accidentally mute it while the controller is loose before installation. Also unplug the usb from the computer if you are not using the c-dsp or dirac programs.

I mounted the controller on a Proclip mount (designed to hold a cell phone) which was customized for my vehicle. It puts it in a very comfortable position, and doesn’t require drilling holes in the curvy car panels or dealing with unreliable adhesives. The plate on the Proclip is just tall enough to mount the controller vertically and still be able to access the screws on the back of the controller. I screwed the controller into the Proclip plate from inside the controller.

I did some parametric EQ in C-DSP while playing pink noise (left then later right) and adjusting to the audiofrog curve (later my own curve) in REW RTA mode for each channel. I avoided boosting dips much (sometimes max 5db if they were horrible), because when I had tried boosting dips I ended up with uneven rattly bass. I mostly was focusing on lowering broad peaks in the frequency response to get it near the target curve. I had a nasty dip around the sub crossover frequency which went away when I inverted the subs; this should not be necessary according to audiofrog since my polarity had been verified, but I did it anyway because it smoothed the response. Just one of those voodoo things that have no good explanation I can see.

The reason for doing this PEQ before Dirac is that my understanding is that Dirac has limits on how much it can boost or cut; the difference between the signal and noise floor if you adjust dirac mic levels according to the on-screen instructions is only about 24db, and my sense was that it was only boosting or cutting on the order of 12db. My system needed more than this to get near flat, hence the PEQ before Dirac. I was somewhat impressed at how the PEQ made the system fairly listenable even without Dirac; this was my first successful use of PEQ after many unsuccessful attempts in the past without proper instrumentation. My reading of the minidsp directions is that they actually encourage you to do a FIR autocalibration with REW before the Dirac calibration, but that just seemed like too much automation for me.

For the Dirac calibration, I removed the driver headrest, and used quickgrip clamps to attach the bendy mic mount that I got from mindsp to the headrest mounts on the top of the seat. I found that the range of that bendy mic mount was just about right for what I thought was good spatial sample spacing.
In Dirac, I loaded the 90 degree calibration and mounted the mic vertically. I chose the single seat visualization and did the cube shaped 9 measurements: first with mic between my ears, then a cube (well actually rectangle) with:
Top of the mic just above the seat top for bottom measurements, and bottom of mic near seat top for top measurements. Avoid placing the mic too low so the seat doesn’t directly reflect.
Mic close to seat for back measurements, and approx. 6”+ forward for forward measurements.
Mic near left edge of seat for left measurements, and extending past the edge of the driver seat toward the passenger seat, near the range of the bendy, near but not quite at the middle of the car, for right measurements. The thinking is that this avoids getting the left too close to the windows, while getting range towards to passenger seat where there is open space. Also I personally hear better out of my right ear 

I started with the audiofrog target curve in Dirac, but given that cutting is always better than boosting, in Dirac I lowered the main part of the curve 5db+ below the 0 normalized level on the Dirac target. After many iterations of trying different tweaks to the target curve in different config buttons, the curve I settled on looks like this (approximate points):
0db at 10Hz
+4db at 40Hz
+4db at 60Hz
0db at 100Hz
-5db at 200Hz
-9db at 8kHz
Natural roll off above 8k which was down to about -22db at the slider at 20k
The curve will vary depending on your vehicle and preferences. I’m still not completely happy with how the bass rattles door panels on some songs at normal listening levels. Holly Cole’s song “Temptation” from the album by the same name always rattles the bass, but most music without too much bass sounds pleasing. In the end I will probably have a separate config with much less bass, for bass heavy music. Those of you with real subwoofers might not have this concern. I do wish that there were some way of doing Fletcher-Munson loudness compensation or just straight compression to reduce the bass at higher listening levels and avoid overexcursion but you can’t have everything.

After experimenting with boosting the left channel by 1-2db post-Dirac in c-dsp to move the center image in front of the driver, I decided I preferred to leave it at 0 (no correction). This pulls the image to the right of the driver toward the middle of the car, but gives a wider imaging stage that I decided I prefer.

For completeness I’ll mention how I got a full digital signal path from my iphone: The magic sauce is the Apple Lightning to USB 3 Camera Adapter (www.apple.com/shop/product/MK0W2AM/A/lig...usb-3-camera-adapter ). This solves two issues: 1. The older unpowered version (titled Lightning to USB Camera Adapter) wouldn’t give enough power from the phone. And 2. It allows you to charge your iphone while you listen to tunes. Most people would be happy powering it from some sort of hardwired or cigarette lighter USB power adapter, but I wanted optimal charging so I installed a mini DC/DC converter with a 120V plug where I plugged in the apple wall wart charger. Then you run a USB cable to an audio adapter; I use the “Signstek HiFi USB to Coaxial S/PDIF Converter“ (www.amazon.com/gp/product/B00FEDHHKE/ref..._title?ie=UTF8&psc=1 ) and then run a coax digital cable to the c-dsp.

The end result is that I have amazing imaging, being able to hear instruments in rock solid spacial locations much more than I ever have in a car. The bass is smooth and clean (within the limits of my door-mounted subs). It was definitely not just a matter of pushing a button; I went through eight iterations, each with 4 different configs of different calibrations, before I resolved all the issues.

Nice write up. Thanks for sharing.

A few items:

Dirac will and can in fact delay individual drivers even when they are all on a single Dirac channel but I do believe entering delays in the plug-in's output tabs prior to running Dirac is advisable as resources can be used elsewhere to result in a more precise tune.

Many of us actually use a multi-channel Dirac optimization to get even more precise delay measurements then can be determined just by measured distance which does not account for xover phase rotation & group delay. Once loaded, the delays on the plug-ins Dirac tab are transferred to the output tabs. Then the Dirac file for that preset deleted before running a 2ch Dirac Optimization.

I also use ZR800-CW but as mid-basses. They are not really meant to be used as subs. Be careful using them as subs long term without a sub-sonic (HP) filter around 35hz as I have been advised of failures occurring otherwise.

I found the solution for the MacOS (Catalina for me) mic problem. You need to delete minDSP program and the DL program. Re download both. Now open Dirac Live from within minidsp (link at the top of the software). This will prompt you to allow mic access if you were not promoted at the time of the redownload of Dirac 3.

Also, EXCELLENT write up man.

Thanks for the detailed write up!

When you say vertical mic position, do you mean that you have the mic pointing upwards to the roof?

Yes, for multi channel dirac, mic is pointing to the roof with the 90° calibration file loaded.