DiyMobileAudio.com Car Stereo Forum banner

1 - 13 of 13 Posts

·
Registered
Joined
·
464 Posts
Discussion Starter #1
Most of you known I run a car PC for my audio...
http://www.diymobileaudio.com/forum...cs-water-cooled-car-computer-95-complete.html

The last few days I've been taking measurements of the car speakers using my Berringer measurement mic...
Here it is 'in action'...







This how-to will focus strictly on using Acourate...with that said, I'm assuming you already have an idea on creating logsweeps and crossovers...and for that matter, using Audacity to play/record your sweeps.
Here is Audacity recording a driver response...



I'm going to be focusing on the high-pass crossover with this...low/sub use the same process....

Once you have your crossovers designed, go ahead and load up the highpass. You'll notice it's loaded in 'Active Curve 1'...more about curves later.
This show a 3rd order Neville-Theile crossover at 2500Hz...



From here, we want to convolute the logsweep, so we choose, aptly enough, LogSweep Convolution...





Load your highpass logsweep...
Note the file structure I'm using...everything has it's own folder! There are a lot of proccess and files generated. Until you become fluent with this software, I recommend keeping a comprehensive folder structure as well...



You must also select your 'inverse' sweep...
I made my sweeps ahead of time and seperated them in respective folders, i.e. HI and LOW folders. I used a sweep of 45 to 3500Hz for my drivers, and 2000 to 24000Hz for the tweeters.
Make sure you don't mix them up! Your hipass sweep MUST use the hipass inverse file. Acourate saves them all with default names, in a default directory, so if you create a new sweep, your old ones will get overwritten.



Once the Convolution has finished, you'll see your sweep in the bottom graph in the 'time domain' window...

 

·
Registered
Joined
·
464 Posts
Discussion Starter #2
From here, you need to make the peak area...left-click to the left of the spike and right-click to the right of the spike...(see above pic).

Now we have to extract just the data we need, so we call up F4, Cut'n'Window...
Make sure Before peak in area...is selected, and enter 6000 in the 'Position/Count' window...this will get all relevant data...



Once that finishes, it'll load in the Amplitude Domain window...go ahead and select 'Amplt' in the top, and display this...



At this point, you can compare your tweeters output to the crossover...you can see where mine are a little weak from 7000 to 9000Hz...but from 2000 to 6000, it's a respectably flat response...
That's correct that!!

Now, we call up 'Phase Extraction', F2...
At this point, we just want to use 0Hz in both fields, and make sure minimum phase is selected...

Save the result to the next available series, in this case, series 4...



Now, so far, you may have noticed, we've used up 4 'Active Curves'...and we'll need a lot more! So now, we're going to save our work so far...
Save it in a seperate directory, i.e. 'Working'...



After you save, we want to 'Clear All Curves' with the double-trashcan icon...



You'll get confirmation dialogs for each curve you want to delete...
Confirm 'Yes' to all of them except Curve 1, our crossover we loaded back at the beginning...we want to keep this loaded for the work we do next.



Now select Curve 2 (because we don't want to replace Curve 1!), and reload the work we just saved...



Now we're back to where we were, but now with 4 available curves left...



From here, we call up 'Frequency Dependant Windowing', F3. This is just a fancy term for frequency smoothing. We want to smooth out the excess data and use just the 'profile' response of the logsweep. This makes the Impluse Response a LOT easier on CPU cycles...



Select 'Both Sides' and enter in this data in the Peaks fields. You can experiment here with different figures, depending on how accurate you want your results to be...

When that completes, you'll see the smoothed response (brown line)...

 

·
Registered
Joined
·
464 Posts
Discussion Starter #3
With a smooth response to work with, we need to apply a 'difference' to the response...a 'mirror' response curve, which will be used to flatten the original tweeter response...so we call up Amplitute Difference...



'Difference' meaning subtraction, so we're going to select our crossover (XO3R48) and subtract that from the last procedure we did, which was the Windowing (FDW2)...save it to the next available Curve Series...



And the output of that procedure...(blue line)



So now we have an 'anti-response' of our tweeter to flatten out the frequencies, but it's still not utilizing the crossover we designed for it!

At this point, we want to elliminate any data outside of the crossover range, so again, we use the right and left mouse buttons to select an area inside the crossover domain...



Anything at -20 or -30dB isn't likely to be audible, so that's a good area to select. On the upper range, your likely not going to hear anything at 24000Hz, so 20 to 23KHz is workable...

Now we're going to run Phase Extraction again, F2, but this time, in the Hz dialog boxes, we want to enter the targeted range we selected...if you look at the bottom left side of the program window, you'll see the Hz data displayed for you...



Now it's convolute time! We're going to convolute, F5, our response to the crossover (in Curve 1)...So again, we select the crossover, and the last process we performed, which was the Phase Extraction (MinPhas4)...save it to Series 6...



You'll see the black line as the result...note how it's conforming the the crossover now...



From here, since we're out of Curves again, we're going to unload all (double-trash cans) Curves, EXCEPT the last one, Curve 6, and of course, our crossover, Curve 1



We're left with our crossover profile, and the convoluted response for our tweeter...



Because of all the data visible outside of the auditory range, we're going to run Cut'n'Window again, F4...
Note the settings...we want a count of 32768 before the peak. Save to Series 2.

 

·
Registered
Joined
·
464 Posts
Discussion Starter #4
Because of how some of the response data is above 0dB, we want to normalize the response to prevent clipping. Normaize function is F8.
Make sure you select the previous saved Curve, and use 'Individual Gain' option...



The resulting window is now corrected to 0dB max...



Now the highpass cross over for that tweeter is complete! All that's left to do now is save it, and view it with your lowpass...



Remember to save it to a folder you know will be safe!

Once saved, go ahead and unload all the curves...



...and return to 'All' view...



Load up your completed low-pass response in Curve 1, and your high-pass response in Curve 2...



Adjust the scale on the left size of the program window to get a better view of your work...



For using Pristine Space, you need to save your work as a wav file...since Acourate can only save stereo wav files, you need data in both L and R channels, or in this case, Curve 1 (left) and Curve 2 (right)...



Since I've been using LEFT channel output for my lowpass and RIGHT channel data for my highpass since day 1, this output configuration works perfectly for my setup....

You'll note in that last pic, the output of the tweeter, generally, is quite a bit below the driver for most of it's range...
There is a technique with Acourate which allows you to do output level matching, basically by performing the same high/low sweep as you used before, but running it with these IR files loaded up.
Once that's done, you load the high/low sweep together, and normalize them together...
 

·
Registered
Joined
·
464 Posts
Discussion Starter #5
You'll also notice the lowpass corrected response of the Premiere PRS720's...pretty darn flat!!
 

·
Registered
Joined
·
4,702 Posts
'Difference' meaning subtraction, so we're going to select our crossover (XO3R48) and subtract that from the last procedure we did, which was the Windowing (FDW2)...save it to the next available Curve Series...



And the output of that procedure...(blue line)

Wow, I've never seen anything quite like this. My tunning on the WinMLS is manual labor compared to this. So this 'Difference' function creates a perfect EQ. contour to flatten the original response. Very impressive.

I was getting some information the other day from MP3 Car and I've noticed you post there too, then I saw this. Major points scored for the carputer setup for sure, the Audison Bit one /Alpine combo looks like a toy by comparison.

So let me try to understand a bit what you are doing here, correct me if I'm wrong. You took a frequency response from one of your tweeters. You uploaded the recorded signal, set an xover curve superimposed, created the mirror image to flatten the response and then superimposed it with the midrange.

I have a 'few' questions base on my experience tunning so far:

Why do you test the FR right in front of the speaker and not at the driver's spot?

Do you use a mic preamp? How does it network to your audio card?

Acourate Audacity free? Is this Asio, VST? What other software do you need?

What does Convolution do, is it just importing your recorded audio file?

Did you try to record back what the FR looks like after the inverse curve is EQ.'d in? I'm asking this because a lot of the times what we see as imperfections are not fixable with an eq. Room nodes and antinodes cannot be EQ'd out and it's better to back off the EQ. manipulation.

And with that being said can you manually EQ. out some frequencies if you want to?

How do you set in delay and phase? Can you do the same inverse kind of treatment on phase?

Carputer looks soo good from what I've seen so far. I might join the ranks with a full setup this summer. One thing I haven't found an answer to was how people deal with the low Pre-amp voltages out of these cards. Most I've seen are 1 to 2V. Line drivers?

Congrats on the setup and keep up the good work! The water cooling in particular is to die for. My setup is A/C cooled, so I have a few tricks too :laugh:
 

·
Registered
Joined
·
464 Posts
Discussion Starter #7
1) I measure at the speaker location, because firstly, I want to correct the speakers response alone, not the speakers response as heard from the drivers position. Basically, before I do any 'room' corrections, I want to tune the speakers so they are outputting a flat response BEFORE I go any further. I measure all 4 corners, as subtle differences in the enclosures will like produce different results...

2) I use a Marshal XML MicMate...it's a XLR to USB preamp with phantom power...it also has 3 gain settings. Plug it in and go!

3) Acourate is (gulp) $500 a license. ASIO4All drivers are free, VST (the plugins) I use Pristine Space (there are others, like x-volver) and Audiomulch to process it all.

4) Good article on Convolution
Convolution Processing with Impulse Responses
There is a LOT more going on with Acourate than just an EQ 'difference'...checkout the crossover white paper...
http://www.acourate.com/XOWhitePaper.doc

6) I ran a second sweep with my corrected filters in use, and you can see a definite difference. Check out the tutorials on the Acourate site, as you can also 'witness' the response manipulation with the sweeps super imposed over each other...

7) There are a lot of different EQ VST plugins you can use...but this whole exercise is so you don't need to EQ! :)

8) Phase is already corrected in the process. Time alignment I haven't tried yet, but consists of a full sweep and marking the 'peaks' from the sub/tweet/mids on both L/R sides, and the software automatically gives you the difference in values.
The outputed data is designed to be used in BruteFIR, but they say it can be used in all kinds of plugins...
I just haven't gotten this far yet! lol

Thanks! I don't think I'll ever go back to a conventional HU setup again!
My Xonar cards, and a lot of commercial grade hardware use pretty decent output stages...
There is plenty enough line level going to the amps, that I have all the gains at full minimum, and in Windows, I cut the wave level volume (kinda like a preamp stage) to 75% max, and there is plenty of music volume there...
And what's really cool is, the sound cards don't clip, or at least I'm not aware of it!
 

·
Registered
Joined
·
464 Posts
Discussion Starter #10
Been awhile since I've logged in...
The wife got the tripod from a camera shop...pretty cheap for camera use, but perfect for a mic stand!
 

·
Registered
Joined
·
58 Posts
How did you fit it?
Ive just tested 2 camera tripod's and the hole on the cameras are much smaller than the mic adapter hole..
 

·
Registered
Joined
·
464 Posts
Discussion Starter #12
How did you fit it?
Ive just tested 2 camera tripod's and the hole on the cameras are much smaller than the mic adapter hole..
I don't remember...lol
I think I drilled out the tripod hole to allow the mic mount to fit?
Sorry, been awhile since I've used it...
 

·
Registered
Joined
·
112 Posts
bringing up this old post again. i hope you're still out there WuNgUn.

$500 is a huge 'gulp' for ((acuorate)). are there any other (possibly freeware) software options of accomplishing this level of filter creation?
 
1 - 13 of 13 Posts
Top