Tuning your car using a pc based measurement setup

[npdang]

This article is a very basic, step by step look at how to tune your car using a pc based measurement setup.

First, position the mic in the listening area. Here, I placed the mic just in the middle of where my head would normally be.



Now we will measure the time arrival of all the speakers.

Start by disabling/muting all the speakers. Starting with the left side tweeter, enable or un-mute it. Begin your mls measurement, looking at the time response (impulse response). I use WinMLS software for this application, a calibrated pre-amp/mic combo from Germany, and the M-Audio transit usb card. Be sure to read the other article on assembling a pc based measurement system for other options. Repeat again, testing each speaker by itself with the others muted.

It is critical when doing this that you make certain your soundcard is "time synched". Most sound cards have a delay between the time that you begin playing a signal, and you begin recording. To make it even more frustrating this delay can vary between measurements. You would just want to make sure that your software+hardware is setup to compensate for this delay, so you can get accurate time measurements.

You should see something like this, which shows you the time arrival of the left tweeter in RED, and the right tweeter in BLUE.



Now, notice the first big spike with the RED line tweeter is going negative. That mean's your tweeter is wired in reverse polarity. A quick trip to your tweeter or amp to rewire and the problem is fixed. Also look at the BLUE line tweeter. The first initial spike, which corresponds to when sound from that tweeter first reaches the microphone, is exactly .6 milliseconds behind the RED tweeter. Now you know that you need to delay your RED tweeter by .6 seconds.



Now look at both tweeters. The initial spike is going positive, which means the tweeters are in the correct polarity. Also, the spikes are "almost" overlapping. The time difference between the 2 tweeters is now about .1 milliseconds which is good enough.

Basically, you would want to continue doing this until all the speakers are in the correct polarity, and the initial spikes are all overlapping.

Next up, we want to set the equalizer. To do this we would measure the frequency response of each driver separately and then take a look. You want to make sure that you include strong early reflections in your measurement, but not later ones. Generally, early reflections are responsible for tonal aspects, while later low amplitude reflections account for a sense of space.

RED = left tweet
BLUE = right tweet
BLACK = left mid
GREEN = right mid

If you're not using an accurate mic+preamp with a known, generally flat response make sure to take your frequency response measurements with a grain of salt.

(NOTE: in the bottom time window, the green/black lines are the mids time response. See how the initial spikes lineup? That shows you that the drivers are correctly time aligned. Look at the second large spike. That is possibly the cone breakup or a strong reflection from the Seas Excel w18 mid/bass. Also, note that although this driver shows a +12db spike when measured in an anechoic chamber (in the spec sheet), in MY car door there is no audible cone breakup at the listening position, and thus no need for a notch filter.)



Let's do an example starting with the left side. I like to begin with the tweeter. Looking at the RED line, we can see that below 3khz there's a bit of a dip, and above 10khz there's a -5db rolloff. So you would probably want to cut out a bit between 3 and 10khz, and boost a little bit around 2khz in order to flatten out the response. I level matched the mid and tweet pretty good by ear so I'll just leave that alone (BLACK and RED lines) Here are the settings I used, inputted into my Behringer dcx2496 processor.



Now let's do the left side mid (BLACK line). It's pretty obvious there's a big hole centered around 750hz, and some bumps around 100hz, 200hz, 400hz, and 1800hz...probably from door cover diffraction and resonance.

Here are the settings I used for the mid. Notice how the boost and cuts corellate to the peaks and dips in the (BLACK) line frequency response:



Now to setup the crossover points. Look at the black line and the red line around 1-2khz. We know that for the Seas w18 Excel used in this example, distortion rises above 2khz. So keeping this driver below 2khz would be a good start. Also, the ribbon tweeter used in this example begins to show higher distortion around 2khz as well. So it looks like 2khz is a good starting point. I also like to use steep crossover slopes in order to minimize distortion, and make sure that the driver really isn't stressed too far beyond the crossover frequency.

Now that I've decided on a 4th order Linkwitz Riley crossover at 2khz, let's look again at the frequency response of the RED and BLACK lines at 2khz. The BLACK line or mid/bass already naturally rolls off at 24/db octave roughly at 2khz. So we don't need to set anything here. The RED line or tweeter "almost" seems to rolloff at 2khz 24db, but actually starts a little later at 1.5khz. So, I decided to try a 1st order (-6db) rolloff at 2khz to see what would happen.

Here is the final response of the left side mid and tweet. Notice the frequency response is much smoother overall, the mid/bass plays down flat to almost 40hz (no highpass set), and the crossover point actually ended up being around 2khz, 4th order (-24db). You also want to make sure that the tweet and mid are playing at the same level.



And here is the final summed frequency response for the left side. Tweet and mid/bass measured together. Notice around 2khz, the transition is smooth and seamless.



Now you basically want to repeat all that for the right side. Make sure that the right side is as good a copy of the left side as possible. Here is the frequency response of both the right and left side independently:



And here is the summed frequency response for the entire system, compare it to the unequalized, un-tuned frequency response at the very start (below). The optimized response is about +/- 3db ... so not bad. The system with only "textbook crossover" of 2khz 18db, and level matching by ear and no equalization has a response of about +/- 10db.



Finally, a good place to highpass the mids would be around 80hz -12db. It keeps the mid from distorting at high output levels, and it's generally a good place to bring the subwoofer in.

Here is a frequency response of the sub (located in the trunk rear firing) with no eq or crossover overlayed with the rest of the system:



Notice the dip around 60hz, and the natural 24db rolloff at 80hz. Doesn't it seem redundant to use a 24db slope on your sub if it's already got a natural rolloff at that point? I'll leave the issue of sub integration as an excercise, but my first thought would be to lower the output a little, boost the response at 60hz, and highpass the mid/bass at 80hz -24db.

Now that was pretty easy wasn't it?

Some points and conclusions I'd like to sum up that I think are fairly obvious.

First off, "one size fits all" passive crossovers and/or texbook crossovers just don't work. That's because we can see that the in car frequency response is completely different from the frequency response on the spec sheets, which were taken in an anechoic chamber. Also, we have to take into consideration the actual natural rolloff of the driver before implementing a crossover.

Also, equalization is nearly a MUST have for typical car setup. Here I have the mids in the doors, and the tweeters in the kickpanels. Everything is heavily dampened. But still notice the somewhat extreme, and strange eq settings I had to use just to get the frequency response flat. It could take weeks or even months of discriminating listening to find all those flaws.

Also, notice the shape of my equalization curves above. Those would be almost impossible to achieve unless you are using a parametric equalizer. Notice that I am equalizing the tweeter and mids separately from each other, as well as from each side. In order to get the best possible/flattest response and good imaging, an equalizer with this kind of flexibility and power is absolutely needed... and chances are you won't find it in an analog unit.

And make sure that even though you have a flat frequency response, you want to cross the speakers over where they will not distort or strain. Always bear that in mind when choosing crossover points, or boosting on the eq.

Lastly, notice how the summed system response looks a bit different from the left/right side response. If that happens, you can go back in and eq both sides to adjust the response flatter. However, although your tonality may improve from having a flatter summed response, the further apart your left and right side frequency response is from each other the worse your imaging will be. Try and fool around with small adjustments and balancing the need for tonal improvement with imaging together, and of course ALWAYS let your ears be the final judge!

In this case, I was very pleased with the sound right out of the box. Definitely one of the best cars I've heard so far. Imaging and staging is nearly perfect (for a car), and tonally it's very good. With the ribbon tweeter response easily goes out flat to 20khz unlike most tweeters. However, I still feel that it will take weeks of hard listening and fine tuning by ear to get that last 5%.... although you could certainly stop at this point since it sounds quite good.

[npdang]

Equipment used:

WinMLS 2000 software
M-Audio transit usb soundcard
microphone and mic preamp from http://www.ibf-akustik.de/2005/index05.html

And let's not forget the Behringer dcx2496 for eq/time alignment/crossover

[Hobbes26]

Awesome writeup.

Were you sitting in the drivers seat at the time of measurement? Legs get in the way of speakers in the kicks or doors quite a bit when you're doing the measurement. Especially when you've got the seat moved forward. ...Though this IS just the first/basic step... :P

[npdang]

Actually no I don't. My head would block the mic

The only difference I can really notice when sitting in the seat though is the imaging smears a small bit to the right side. Moving my leg a little more out of the way or pulling the seat back so that I'm not blocking the mid clears up the problem. Tonally, I don't know what the effect would be although it sounds balanced to my ears.

[aaron]

Definitely required reading! Thank you for all the info. Now I know why such few cars ever seemed to truly impress me. Not many people were doing these kind of measurements ( that I knew about ) when I was involved with car audio about 10 years ago. It was all RTA.

[FaintReality]

When checking the frequency response of the drivers, the tweeters in particular, is it safe to not have the crossovers set? I am assuming you set the crossovers on the tweeters to protect them, just lower than ideal so you can find out where there natural roll-off is? Then again, the software does not use test tones so it makes me think they might be ok with no crossover during testing?

Also, when performing the test, is there an ideal volume setting on the HU like 1/2 way up, or can you just turn up the HU a few clicks since the frequency response should be the same regardless of volume?

Thanks, Dave

[sqkev]

Please bear with me, a few noobish question

What should I gate all the measurements at?

In your summed results, are they actual measurements or simulated?

I see that you measured tweet + tweet and then woofer + woofer
THEN adjusted TA on each and xo and eq afterwards.
For someone who doesn't have a dcx and considering passive, is it better to:

do gated FR of tweet and mid on one side, adjust level, xo, do measurement again

then
repeat the same for the other side and EQ and TA both sides?



one more question that is slightly off topic,
when going active and setting all different levels for tweets, mids and possibly midbasses and subs, would the system be all linear at all different levels?
Let's say that we match the voltage on all channels, yet not all amps produce the same amount of power at that certain voltage. In theory, this is the disadvantage to a all active system, correct?

[npdang]

Originally Posted by sqkev

Please bear with me, a few noobish question

What should I gate all the measurements at?

In your summed results, are they actual measurements or simulated?

I see that you measured tweet + tweet and then woofer + woofer
THEN adjusted TA on each and xo and eq afterwards.
For someone who doesn't have a dcx and considering passive, is it better to:

do gated FR of tweet and mid on one side, adjust level, xo, do measurement again

then
repeat the same for the other side and EQ and TA both sides?



one more question that is slightly off topic,
when going active and setting all different levels for tweets, mids and possibly midbasses and subs, would the system be all linear at all different levels?
Let's say that we match the voltage on all channels, yet not all amps produce the same amount of power at that certain voltage. In theory, this is the disadvantage to a all active system, correct?

I didn't gate any measurements, however I did use 1/3 octave smoothing. These are actual measurements, not simulations.

The steps I used were:

1. time alignment for all drivers
2. level matching
3. handling driver to driver transition which does include use of eq as well as xover
4. equalization

It wouldn't matter if you went active or passive, the steps should be the same.

Your last question I think is a bit confused (or perhaps I am). Level matching is matching the sensitivity of all the drivers... not the output voltage of your processor or amplifier.

You can measure a change in system frequency response with volume, however it isn't *that* big of a deal Imo... and it would be due to things like driver non-linearities, reflections, power compression, etc.

And for things like the sensitivity of our ears to different frequencies changing with spl, that's fine as well. As long as you playback the recording at the original volume, it should sound accurate. You shouldn't expect it to sound the same at a significantly lower, or higher volume Imho.

[corrado]

hi npdang,

I have already bought all the hardware require for the measurement, now just deciding on which software to get.

one question, must I point the mic to the rear when measuring the sub, assuming my sub is in the trunk?

please advice, thanks.

[300Z]

Cool, another Corrado owner here... how did you found out about this site?

As for your question about the mic placement when measuring the sub, it shouldnt matter since low freq arent controlled dispersion...

Regards
Leo

[corrado]

hi there, thanks for the response....

but I do not own a corrado...how I wish I do....

got to know this site via a friend....

btw, my SS exact should be in next week....

bought it after reading review from sounddomain....I believe you are there

cheers.....

[300Z]

Yes i'm on sounddomain, ECA and some other forums as well...

What sized of the Exact you got?

Leo

[Grand Masta]

Originally Posted by FaintReality

When checking the frequency response of the drivers, the tweeters in particular, is it safe to not have the crossovers set? I am assuming you set the crossovers on the tweeters to protect them, just lower than ideal so you can find out where there natural roll-off is? Then again, the software does not use test tones so it makes me think they might be ok with no crossover during testing?

Also, when performing the test, is there an ideal volume setting on the HU like 1/2 way up, or can you just turn up the HU a few clicks since the frequency response should be the same regardless of volume?

Thanks, Dave

Nguyen, I am curious about this as well.

[daitrong]

Originally Posted by Grand Masta

Nguyen, I am curious about this as well.

i would guess you have your xovers set before measuring... You're not going to need the Frequency response at 350hz for your tweeters since you're never be playing your tweeters down that low. Plus if you were to measure it that low, the FR would like all jacked up.. lol Maybe nguyen will chime in and let us all know for sure.

[npdang]

I do set the highpass crossover when measuring tweeters.

[Grand Masta]

Do you set the highpass on the midrange or midbass drivers to prevent them from overexcursion in the lowest octaves?

Also, why is the tweeter present in this graph? Looks like the "midrange" of the tweeter was properly attenuated by the filter and then the tweeter began playing again in the lowest octaves. I am confused.

[npdang]

It's background noise... you see it's present for the mid's as well.

For the mid/bass I don't highpass, but you can. It's really not even necessary for the tweet as long as you're not driving it too hard.

[Grand Masta]

Hmmm... What is causing this background noise in the lowest octaves? Does it affect the subwoofer FR measurement as well?

[npdang]

It's not a big deal since it's mostly under 20hz. If you really need accuracy around that area you can always use a different stimulus along with pre/post emphasis to remove noise.

Imo when tuning bass in the car, it's not always about having a flat response but having a smooth response and seamless integration between mid and sub that makes it sound good.

[Greg200SE-R]

Originally Posted by npdang

Start by disabling/muting all the speakers. Starting with the left side tweeter, enable or un-mute it. Begin your mls measurement, looking at the time response (impulse response). I use WinMLS software for this application, a calibrated pre-amp/mic combo from Germany, and the M-Audio transit usb card. Be sure to read the other article on assembling a pc based measurement system for other options. Repeat again, testing each speaker by itself with the others muted.

It is critical when doing this that you make certain your soundcard is "time synched". Most sound cards have a delay between the time that you begin playing a signal, and you begin recording. To make it even more frustrating this delay can vary between measurements. You would just want to make sure that your software+hardware is setup to compensate for this delay, so you can get accurate time measurements.

You should see something like this, which shows you the time arrival of the left tweeter in RED, and the right tweeter in BLUE.



Now, notice the first big spike with the RED line tweeter is going negative. That mean's your tweeter is wired in reverse polarity. A quick trip to your tweeter or amp to rewire and the problem is fixed. Also look at the BLUE line tweeter. The first initial spike, which corresponds to when sound from that tweeter first reaches the microphone, is exactly .6 milliseconds behind the RED tweeter. Now you know that you need to delay your RED tweeter by .6 seconds.

npdang, I am trying to measure the time arrival for each individual driver in my system, just like the measurement in your first graph above. I am having trouble getting such a nice, "clean" and distict spike to appear in my measurements.

My setup is: M-audio Mobilepre, Behringer ECM8000, demo versions of lsplab, WinMLS and Freeprobe running on a current model Vaio notebook. In my car I have a Clarion DRZ9255 with active 2-way front comps and a single sub. All I want to do right now is accurately measure and set my time-alignment.

Can you go over the initial settings for WinMLS (Seems like you use this app often)? AFter calibrating the output/input levels, are there any special settings that need to be changed?

the results I am getting are jagged lines, and then a not-so-distinct spike... not a flat line leading up to a clearly defined spike like in your graphs. The readings can change somewhat between measurements as well. I know it's not easy to elaborate on such a complex subject but any info you (or anyone else) can provide would be appreciated!

[JohnSmallberries]

I am also trying to set up a system like this. I have the ECM8000 as well and what I want to know is, how do I calibrate this microphone, or do I really need to in order to get tests that would work for setting eq etc for my pxa-h701?

Not being much of a german reader I don't have any idea what the site that npdang linked has to do with the mic...

Thanks a bunch!

[dennisp]

Anyone who has the MobilePreUSB and WinMLS, could you tell me what settings you have here? Thanks.

I'm just confused as to how to calibrate the sound card.

[JohnSmallberries]

i have a similar issue.. im not sure how to use the mobile pre in winmls...

I have a ecm8000 and tried using truerta as well but I think it was not setup right as the adjustments I was making in the car made the audio sound change but the response in these programs didn't seem to change.

[03blueSI]

That is wierd, I have the ECM8000 and a behringer mic preamp and I have used TrueRTA in the past and I could easily see the changes. I am now using FuzzMeasure (MLS for the mac) and can see changes fairly easily as well. One of the big things to do though is to use multiple mics if you can afford to and sum the response

[audionutz]

Originally Posted by Greg200SE-R

npdang, I am trying to measure the time arrival for each individual driver in my system, just like the measurement in your first graph above. I am having trouble getting such a nice, "clean" and distict spike to appear in my measurements.

My setup is: M-audio Mobilepre, Behringer ECM8000, demo versions of lsplab, WinMLS and Freeprobe running on a current model Vaio notebook. In my car I have a Clarion DRZ9255 with active 2-way front comps and a single sub. All I want to do right now is accurately measure and set my time-alignment.

Can you go over the initial settings for WinMLS (Seems like you use this app often)? AFter calibrating the output/input levels, are there any special settings that need to be changed?

the results I am getting are jagged lines, and then a not-so-distinct spike... not a flat line leading up to a clearly defined spike like in your graphs. The readings can change somewhat between measurements as well. I know it's not easy to elaborate on such a complex subject but any info you (or anyone else) can provide would be appreciated!

Ditto!!!! Anxiously awaiting the answers