Skip to main content

Hello everybody,

My application aims for Active Noise Cancellation, I have to use Microphones and a DSP to digitize the received signal then act on speakers to cancel the noise.

Could you please help me on which criteria that i have to put into account in order to choose the right DSP, and if you have any suggestions concerning DSP evaluation boards ?

Thanks in advance

Comments

Brother Junk Tue, 01/03/2017 - 07:21

We need more than that. Can you tell me what you will actually be doing? Are you test driving cars and making videos? Are you making videos in the same car?

If you are testing cars, you can't deaden the vehicle (which would get rid of a lot of ambient noise). If it's your own vehicle, that would be part of the process.

So, we need details, and how far are you willing to go? E.g. Welding, full vehicle deadening, cutting holes in the car etc...if you are in other people's cars, none of those are an option.

See what I mean? We need to know the whole situation, or we are all just wasting each other's time. Everybody is willing to help, but at this point, I have no idea what you are trying to do. It's hard to help you mitigate any problems without that knowledge.

But I'm interested...

Marouane Tue, 01/03/2017 - 08:05

Actually I'm hardware development engineer working on sound systems, so I think that I gave you the essential requirement for my application. I'm seeking for an active solution not the traditional ones (Welding, full vehicle deadening, cutting holes in the car ...). All I need is : how to chose an appropriate DSP that can support interfacing with Matlab.Image removed. Please take a look to the picture in order to know what am talking about.

Attached files

Boswell Tue, 01/03/2017 - 09:15

Marouane, post: 446235, member: 50284 wrote: This application will be used in automotive domain (moving vehicle), in order to cancel the air-born noise for each passenger (Two Mics + DSP+ Two speakers).

Marouane, post: 446238, member: 50284 wrote: Actually I'm hardware development engineer working on sound systems, so I think that I gave you the essential requirement for my application. I'm seeking for an active solution not the traditional ones (Welding, full vehicle deadening, cutting holes in the car ...). All I need is : how to chose an appropriate DSP that can support interfacing with Matlab.Image removed. Please take a look to the picture in order to know what am talking about.

Your diagram shows the standard simplification for analysis of nulling a single sound source propagated down an acoustic waveguide. Unfortunately, this is a long way from representing the inside of a vehicle, particularly when there is more than one occupant expecting the noise to be reduced.

I do know that there are many car manufacturers that have had teams of skilled acoustic and DSP engineers working on optimal solutions for their own particular vehicles. All the solutions I have been involved with have used multiple microphones and multiple speakers (i.e. more than two of each) in order to achieve results that have been deemed worthwhile. Some other manufacturers have abandoned their attempts at this time because of the difficulty of the problem and the cost of a workable solution.

I did once try a prototype single-seat version of a two-microphone system that played through the car stereo, and it was eerily effective within a narrow band of vehicle speeds if you held your head still. However, outside the very small cancellation zone, the phasing noise associated with imperfect nulling was quite unpleasant.

Good luck!

Brother Junk Tue, 01/03/2017 - 16:43

Marouane, post: 446238, member: 50284 wrote: Actually I'm hardware development engineer working on sound systems, so I think that I gave you the essential requirement for my application.

I'm not trying to give you a hard time. I sincerely want to help. It's nice to see another car guy around. But I didn't know you were a hw developer working on sound systems, till just now (vs a diy person). I didn't know that passive was out of the question, till just now. You said what you wanted, you didn't specify the things you can't use to get there. Or what is a non-factor.

I still don't know if you are doing this for a specific car. Is this for a new Cadillac SUV? For Cadillac cars in general? For every car in America?

If it's for a specific car, your job will be much easier, as you can take a lot of measurements. You will also have known cancellation speaker installation points (or better yet, you may be able to dictate them). What you are attempting to do, is quite unnatural. I get that it's your job now...but is the design intended to cover a specific vehicle (it can't work any other way imo, at this point in tech) or every car in the future?

What I can tell you is that every car is very different. Cars have different resonant frequencies, they have different cabin gain, different frequency trapping areas, different tires create different frequencies, the speed of those tires alters that. The motor noise...is it a 4 cyllinder, or an 8 cylinder? Where are you going to mount your cancellation speakers? Depending on the amount of noise, you may need several speakers, that are perfectly placed, in order to work. bouldersound absolutely nailed it. The speaker placement in the car will be the main point on which success hinges. Frequency cancellation (afaik) relies on a driver to invert the pressure of the sound wave. If those two waves are not, and don't remain aligned, = fail. This is why it's much easier to do with headphones, than with a car. Especially as Bos pointed out, if there is more than one occupant. It's just not possible to invert all those waves for every occupant at this point in time.

Let's assume you already know all about it all (I'm being sincere)...the dsp board... I have no idea what type of devices you are looking at. I feel like this area is more of your forté rather than mine, but I would want the most powerful board I could get. At least a 24 bit/96khz....depending on price (car mfr's are notoriously cheap) a 24/192. Anyone feel free to correct me, but the more samples you have, the more accurate the inversion will be. And since we know that the inversion must be very precise, logically, you would want the most samples possible. Or that "Cadillac" will pay for.

Is that your only concern? A board that will integrate with Matlab? Could you just call Matlab for suggestions? And is mic and speaker placement someone else's job? Bc I can't see how they could be separated and be effective, unless you are part of a team. But I would bet that as far as hw is concerned, the mic and def speaker placement would be far more important than the board selection. One is dictating what should be cancelled, one is doing the cancellation.

For the board, I would want the most samples possible. But that's just my logic working, I've never actually tried this endeavor. Maybe a 16 bit board would be fine...ideally, I would still want the most samples possible.

If the design is not protected, I wouldn't mind seeing what you come up with in the end. And I'm curious where they plan to place the primary mic?

pcrecord Fri, 01/06/2017 - 06:14

This is funny to read about car makers trying to hide noises instead of reducing them at the source...
I know how to do in in a DAW (recording software) but in a car.. it's a long shot to ask a recording forum. ;)
Anyway, I so curious, I'll check how this goes..

BTW My Chevy Equinox 2012 has a noise reduction system with a phase cancellation speaker in the trunk. You may want to check how they did it...

Brother Junk Fri, 01/06/2017 - 13:01

pcrecord, post: 446359, member: 46460 wrote: This is funny to read about car makers trying to hide noises instead of reducing them at the source...

I agree. Maybe the OP can enlighten us, but I bet it's due to the added weight and reduced mileage of deadening a car. I had to install different shocks in my van. Heavier door hinges. I mean, I went overkill, but deadening a vehicle adds substantial weight and reduces performance and gas mileage.

Or maybe it's just to see how much they can improve cancellation tech in cars. Like I said, Headphones, the variables are far fewer than a car. If they get it working super well in a car for all passengers, I'd love to see how they have done it.

I can't see how you get past the sound pressure wave needing to be precisely inverted the whole wavelength. Maybe lots of mics and lots of speakers?