3
$\begingroup$

I am working on ''Active noise cancellation for air vent'' project. I am using NI myRIO1900, Audio Technica ATR 3350 omni directional microphones and Tevion Loudspeakers. I have done the programming using LabVIEW and I have used FPGA module.

I am using sinusoidal noise source of (200-500 Hz) as test noise signal.

I have implemented LMS on LabVIEW, as a preliminary results I am getting around 8dB noise reduction but these results are pretty random in a sense.

Only at certain volume levels( if I adjust very carefully), the noise cancellation is happening and the variation in step size is not giving any decent change ( again I have to make volume level changes to see some results).

Over all, I am getting some results but I don't have control over the results.

This is my set up enter image description here
[the pipe length is 1 meter]

I know I have to consider secondary path modelling and feedback modelling. I have actually completed the coding of FBFxLMS also but I don't have any reasonable results with this also.

But my question is how to improve the acoustical design of my setup.

What are the problems using just cardboard material to simulate the environment of air vent?

I have read in 2-3 places (here in 2nd answer) that without proper acoustical design it is not possible to reduce the noise effectively.

Quoting from a paper

If the acoustical design of the system is not optimized,the digital controller may not be able to attenuate the undesired noise adequately.

The only point I am aware of is using low frequency noises (<= 500Hz) to consider the sound wave propagation as plane acoustic wave. I am completely unaware of any other details on how to better my acoustical design.

I am wondering how to find the limitations of how much a DSP will affect and how much the actual acoustic design. If anyone can provide some reasonable resources to go through, it will be highly appreciated.

Improve this question
$\endgroup$
2
  • $\begingroup$ I am not sure whether your 'Active noise cancellation for air vent' project got completed. I would appreciate if you share your knowledge on how you overcame the above issues. I am doing an implementation of feed-forward Filtered X-LMS for active noise cancellation in lab set up with an engine accelerometer data of low frequency (<500Hz) and Fs = 2000Hz. I created a setup almost similar to yours (figure above) with speakers and mics. I do not have a directional speakers and high efficient mic. I have a Creative SBS 2.1 Speaker and a Table Desk Microphone to capture the data in real time. The a $\endgroup$
    – user20076
    Mar 18 '16 at 8:48
  • $\begingroup$ Hey, I could not exactly solve the above problem. It is too complex for me to handle the physical (acoustical) aspects of the project, may be I need to use an-echoic chamber or completely isolated room!! As of now, I quit!! rather I did simulate the results. They are decent. Even though I would love to share my work, I can't now! As it is my thesis work,it should get evaluated before I can share with anyone. I can help you with any specific questions though. $\endgroup$
    – charansai
    Mar 20 '16 at 7:16
1
$\begingroup$

I first wasn't sure that this was in the scope of a digital signal processing site, but then decided that it is about limits and environments for doing DSP, so it's quite on-topic.

Your paper (what paper? Please, please, get used to always correctly cite!) hit the spot:

If the acoustical design of the system is not optimized,the digital controller may not be able to attenuate the undesired noise adequately.

I think you understand quite well what they're saying: You've got to deal with signals in the real world, i.e. with actual, physical, pressure waves. You're doing that by modeling the real world mathematically.

This can of course only be a success if your reality is "easy" enough to fit your model. For example, your ANC probably assumes that there's a position for the noise emitter, and position(s) for the cancellation signal emitter(s). What it probably doesn't involve is that with a bad acoustic design, noise might be leaking out of the back of your emitter, bounce off a distant wall, and come back with a huge, random phase shift.

This is science. In science, you try to reduce the amount of uncontrollable effects as far as possible to make as many things certain as possible.

Your design is a good one, in principle: Cardboard is pretty effective at absorbing sound. However, you weren't really too careful when constructing your system:

  • gaps. Close them.
  • Walls. Make them thicker.
  • backs of speakers, open for absorption and emission of ambient noise. Pad them.
  • Not too sure about the bottom of your "tube": Is it dampening cardboard or actually the table, which probably is much better at transporting power of a speaker lying flat atop than air is?
  • Overall system exposed to ambient noise. I'd get a big cardboard, put down a layer of insulation (e.g. styrofoam), put everything inside, fill things up with insulating material.
  • Is omni-directional really the pattern you're looking for in microphones? This all sounds like you want your noise source mic to be highly directional to only capture the noise, and not the effect your own emissions have at the noise source.

Then, make sure all the components, especially your sensors, work very well:

  1. are your microphones (especially that right in front of the noise emitter) perhaps clipping?
  2. if not 1., then, how clean is the signal your speaker produces? Maybe you want to try and produce a few tones at different volumes, and look at the spectra (coming from an RF background, I'd propose full correlation-based channel sounding, but I don't know how well that works).
  3. Verify your model: unplug your cancellation speakers. Run your device, and record what they would have produced as sound to a file, as well as what your microphones heard. Simulate what the mics at the places where you want your noise to be cancelled should have heard based on these files. Do exactly the same with the speakers plugged in. Compare simulation with actual microphone recording.
Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks for your detailed explanation. I want to cite the paper but with amount of credits or privileges that I have, I can not add two links in the question. Yeah, as I was quoting, I would have preferred to add this link. I will consider all of your suggestions. I want a bit more clarifications. When you said ''backs of speakers, open for absorption and emission of ambient noise. Pad them.'', do you mean to use just the cone(core) part of speakers ?? and Isolating the ambient noise is big headache for me, I am out of ideas ...(Contd.) $\endgroup$
    – charansai
    Jan 14 '16 at 20:30
  • $\begingroup$ I am actually working late at night when no one will be there :D. If I close the entire setup that I had with another closed box with proper insulation, taking care of wall reflections is a big problem, I think. I had a bad experience with directional microphone, their frequency response and build quality is so bad that I had to go for omni-directional and better quality microphones. Thanks for the last 3 suggestions, I will try to follow those. Finally, can I get some references discussing practical acoustic design problems (especially for duct type systems).. Tq :) $\endgroup$
    – charansai
    Jan 14 '16 at 20:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.