Does anyone have experiance with generating wind sounds for a car that the user would hear from the drivers seat. We already have the front and side air preasure on the car, we are calculating this already for our dynamics. We need something fairly realalistic does any one have a suggestions on how to do this?
This does have to be done real time, but since we have a dedicated machine (2 CPU) we can dedicate a good deal of CPU time to this.
- dheitbri asked 12 years ago
Do you know if any perticular noise function would work better, like Perlin Noise, or Gauss. We were thing about trying to use a noise function to generate a wind sound, then using band-stop filter to simulate the mufflering effect the cars body has.
Right now we are under kind of a time constraint, we need to crank this out pretty quick, our old system we are replacing sucked so we are not aiming for the stars here we need something passable, but we also plane on going back in improving everything once we have something that works.
If you have recording gear, record while driving in a quiet car with quiet tires. You can study the sampled audio and recreate, or just filter the sample as needed (remove any tire or engine noise) and loop.
In a race car, the dominant sound will be the engine and exhaust.
You can find examples for both on YouTube, Google videos, etc.
To create colored white noise, randomly perturb a sine wave at the desired frequency (this is how I created a low frequency jet engine boost sound). Combine multiple perturbed frequencies to add depth. When playing back in the car, the noise will not be continuous, but will waver a bit.
Tire and wind noise are going to be the dominant sound in this case. We are modeling fairly "normal" driving conditions, you mostly only here the engine in most cars when you apply the gas. We have a good engine model, or good enough, there is some good lit on producing engine sounds.
We are using a late 90’s malibu, and and a late 90’s Jeep SUV.
Anyways we will probully try to get a quite car, maybe a hybrid and drive it on new black top at 25,45,55,65 then blend the sounds based on speed and impact pressure.
The questions are:
Whats the best way to when recording to isolate the "wind sound", this may not realy be possible, since the road sound is pretty close from what I can tell.
Also how much to we need to vary amp and freq of our noise vs impact pressure.
I think we have enough to make ok sounding stuff. I guess a good deal will be just playing around with stuff until we get something that sounds right………
A directional shotgun mic pointed at the window(s) could help reduce road noise (can rent if you don’t have one).
You could try to isolate the road noise using advanced spectral (FFT based) filtering in SoundForge and Audition. However, if you are going for realism, road noise should not be removed. Just record at various speeds, and blend betwen the samples at runtime based on simulation speed. The best effect would be a stereo mic placed near the driver’s head, playing back the stereo recording at runtime (providing depth). I think you’ll find a good solution without too much work.
We are not worried about the tire/road sounds as we are modeling these sounds seperatly, we already have a model based on these built, and they seem to work ok for now.
What we have decided for now is to record the wind sounds using a shotgun mike pointed at the windows, recorded at regular intervals, then interpolating between sounds in our simulation based on forward windpressure. If we have time I think I will experiment with overlaying different noise functions as well.
I would use FMOD Designer to prototype.
We actually have a wind example, but its not as realistic as you would like. (it is like a whistling wind).
First, add a white noise oscillator, then start adding lowpass and bandpass filters and tweak the cutoff / resonance values. Thats what i did.
Please login first to submit.