Live RTA of a car's transfer function

greenwire · 01-31-2005, 01:42 AM

Hi. I'm wondering if anyone has ever tried doing any live real time analasys in their car before.

i'm thinking about mounting a measurement microphone permanantly in the driver's headrest and getting a full duplex sound card. I want to be able to get a live transfer function of my system and my interior acoustics and apply it to the EQ, so my truck always sounds perfect. windows up, windows down, loaded or empty, turned on or turned off

Does anyone know if signal analasys and DSP software exists for such a thing? i haven't seen any but i'm sure it's possible. and hey, with a little bit more work it should be able to do some active noise cancellation for the engine noise too..

am i nuts?

techy101 · 01-31-2005, 02:17 AM

Are you talking about the types of software we use for live sound applications? Things like SmaartLive-$695 plus a decent measurement mic-Earthworks M-30 $475. (http://www.siasoft.com) This software has a 30 day free trial available.

If you didn’t want to spend that much, (keep in mind that price pretty closely = quality with these software’s) you could go with something like True RTA, but you'd need to get the Level 4 version for $99.95 (plus a descent microphone again. looking at a few hundred for a properly calibrated mic) to have any decent measurement ability. Anything lower only gives you up too a 1/6 octave RTA.
(http://www.trueaudio.com)

As far as active noise cancellation, what exactly do you mean? Are you talking about having the software decide how to eq your system to compensate for whatever noise there is? If so, I have never seen such software, one may exist, but a computer can't make things sound good. It can tell you what your spectrum looks like, and allow you to decide how to compensate, because in the end it is you who decides if it sounds good or not.

OOORRRRR you could just go all out and put a SIM3 machine in your car

(http://www.meyersound.com/products/i...sim3/index.htm)

Also keep in mind that for an accurate reading, you will need to periodically have your microphone re-calibrated.

PS. If you did try something like this, I wouldn't try to run it on your carputer. These software’s take up system resources like none other. I'd recommend putting it (maybe the SmaartLive demo) on a laptop, making your measurements in the car, and creating a few presets for your eq based on these measurements. That would save you a lot of money. You don't have to use a calibrated microphone, but if you choose not too, you should try to find the most flat freq. response mic you can. Take a look at the freq. response charts, and compensate your measurements accordingly.

If you are feeling ambicious enough, please keep me updated on how it goes.

Vacheron · 01-31-2005, 02:18 AM

Well I'm sure the software will be out there somewhere because I had a Pioneer DEQ-9200 way back in 1997 which did the same thing (Balanced the parametric equaliser, adjusted the time delays from the speakers and set the balance and fader positions) though albeit with a custom DSP chip instead of a CPU.
I have no idea if they still make it, but the amount i got for it when i sold it on ebay a couple of years ago suggests they may not!

It used a small microphone you tied around the drivers headrest and left the car, then it would fire a series of pink noise out of the speakers and analyse the FFT responses of the results.

You will probably initially need to borrow an audio CNE (constant noise emitter) to provide fixed values of the initial stages the sound will go through

* Mic Frequency Response
* "Mic In" Frequency Response
* Sound card software input fixed EQs

Anything beyond this point can be controlled and calibrated by your software.

The next tricky part will probably be interfacing with the sound card drivers/software to adjust the time delays to the speakers if that it the way you are thinking of going? However I will leave that to others because I'm not a software writer if I can at all avoid it!

Sounds like a very interesting project though.

techy101 · 01-31-2005, 02:21 AM

Quote: Originally Posted by Vacheron

You will probably initially need to borrow an audio CNE (constant noise emitter) to provide fixed values of the initial stages the sound will go through

The softwares i spoke of earlier all include a tone generator, pink/white noise generators, multiple sine wave generator, and custom sine wave sweeps.

Vacheron · 01-31-2005, 02:34 AM

Quote: Originally Posted by techy101

The softwares i spoke of earlier all include a tone generator, pink/white noise generators, multiple sine wave generator, and custom sine wave sweeps.

Sounds good (sorry for any repetition in my post but you beat me to the submit button by 1 minute)

.

The reason I mentioned using a separate audio CNE is that if the transfer function of the signal into the processing software is non linear, the software cannot know or compensate for this, and so the final frequency profile will be the inverse of the response of the signal input components.

So if you are looking, for example, for a totally flat frequency response, then the software will need to know this data which will then have to be periodically calibrated by an independent system every now and again for best results.

It all depends how accurate you want the final result to be I suppose.

techy101 · 01-31-2005, 02:51 AM

haha, *races to the post button*

Ya, i say just go with the SIM machine. It'll only cost your house, car, computers, soul, and two animal sacrifices, but it's a great machine

Or better yet, just have the guys who did this setup come in and do your car

greenwire · 01-31-2005, 09:04 PM

HAH.. I used to do a bit of production rigging, that looks like a nightmare to me.

I don't think i want the guys that broke the roof of that stadium to come anywhere near my truck

About the white/pink noise thing, i was going to try to get away from that.

What i want to do is much cooler, i want to replace the pink noise signal usually used for calibrating audio with real music, real sound.

Do an FFT of the source live out of the soundcard, and take the FFT of the measurement microphone. Apply the appropriate delay to the signals to compensate for latency, take the difference between them and apply it to the EQ, so while you're driving and rockin' out, the software is compensating by the DIFFERENCE between the source and the measured return.

And about noise cancellation, you all know about the noise cancelling headphones that measure the ambient noise, invert the phase and apply that signal to the headphones, using the property of destructive interference to "cancel" the noise.

Why not do the same thing in a car? measure the source waveform, measure what the microphone hears, weed out what isn't good signal from the measurement with a subtract operation after the appropriate delays, invert the extra noise and apply it to the signal.

The art is in calibrating and setting up the feedback loops, has this been done before? and if no, why the hell not?

starfox · 01-31-2005, 10:00 PM

You might need to buy a high end soundcard that can sample fast enough to do this, eg. if you're playing CD quality stuff you'll need a soundcard that can sample at 88.2khz else your result from the FFT won't be accurate enough to seperate noise from the signal and noise from sampling (nyquist).

I can imagine that you're going to need a lot of processing power to do this in realtime as well

greenwire · 02-01-2005, 01:46 AM

Sure, but a lot of soundcards can do 24 bit at 96khz on 2 channels easially, and a measurement microphone is mono, so there's half the bandwidth right there.

The thing wouldn't need to write to HD, and i doubt it would take more than a P4 at 1ghz to do it if it was dedicated. This is the kind of thing i would need to build a little bit of specialized hardware for anyways, so why not just run it on a good fast PC running a stripped down nothing-but-audio linux build?

I can't see any issues with doing it in realtime on a modern PC if it's specialized and not running XP or anything bloaty like that.

just buffer the input. You can always compensate for speed in your control loop. 1gb of ram is a ****load of audio data, so if you don't plan on listening to any 2-hour long songs, it would be able to buffer the whole load and not flinch.

Vacheron · 02-01-2005, 02:13 AM

Quote: Originally Posted by greenwire

Why not do the same thing in a car? measure the source waveform, measure what the microphone hears, weed out what isn't good signal from the measurement with a subtract operation after the appropriate delays, invert the extra noise and apply it to the signal.
The art is in calibrating and setting up the feedback loops, has this been done before? and if no, why the hell not?

Well the franticly sounding dual airhorns of a rapidly approaching semi trailer would be a pretty good reason!

On the subject of sampling, 44KHz would still be enough fopr Nyquist, and the original CD spec is sampled at 44.1K for exactly the same reason, however the output in the analog domain will still be 20K tops.

starfox · 02-01-2005, 05:50 AM

Quote: Originally Posted by Vacheron

On the subject of sampling, 44KHz would still be enough fopr Nyquist, and the original CD spec is sampled at 44.1K for exactly the same reason, however the output in the analog domain will still be 20K tops.

Hmm, yup that's absolutely true. I must've been half asleep this morning.

greenwire: alright good luck

If you manage to write the linux apps to do this, please show us the source code.. it'd be interesting to see!

greenwire · 02-01-2005, 08:05 PM

That's the problem! i have no idea where to start with something like this, i've never coded anything in my life

I'm a sound tech, not a computer geek

Anyone want to help?

BassBinDevil · 02-01-2005, 08:29 PM

I don't know if it's helpful, but there's the kxproject driver which lets you exploit the DSP abilities of Live and Audigy cards in various ways; could be a way to implement the EQ you'd need: http://kxproject.lugosoft.com/dsp.php?language=en

Liberty Instrument's Praxis software can do some useful measurements: http://www.libinst.com/PRAXIS.htm
"a high resolution dual channel RTA spectrum analyzer (up to 1/48th octave) with a pink noise generator, averaging functions, and table readout of data. "

Speaker Workshop could be useful:
http://www.speakerworkshop.com/SW/Feature%20Summary.htm

greenwire · 02-02-2005, 05:29 PM

Wicked

Hey! you're from the island! rock on... thanks for the tips man

User Name		Remember Me?
Password