How Can Active Spatial Noise Control Reduce HVAC Noise in Meeting Rooms?

[fooser-q]

Hello everyone,

I'm new to this forum, and though this question may be addressed already in another thread somewhere, I'm hoping to get some focused discussion about solutions going here.

Problem:
HVAC noise is propagating down the HVAC duct, entering a meeting room / boardroom. The noise source is two fold in that it is a result of both the motor itself and the vibration in the duct. These are periodic noise signals with frequencies that can reach up to about 6KHz (top end) from generally around the 60Hz or harmonics thereof. The focal point is to remove as much of the low frequency with Active Noise Control, and the high frequencies will have to be dealt with using conventional damping / insulative techniques.

Solution:
For the low frequency portion of noise, produce an "anti-noise" signal by measuring the source, using an inverting amplifier, and outputting the phase-shifted through a speaker mounted in the duct to cancel the incoming noise waveform. I implemented this in a small scale using a loudspeaker to output a single/mixed tone signal to simulate the noise and used an electret condenser mic to measure the noise, and another loudspeaker to output the anti-noise.

New Problem:
The anti-noise now has a little bit of the desired effect of canceling out the original noise, but the microphone now is picking up the "anti-noise" and feeding that back into the system. This is causing oscillation and eventually system instability (ie: loud audio feedback squeal).

How do I solve this problem of removing the oscillation? Will I have to implement a whole new system using DSP with FIR filters? Should I use a noise gate on the measuring mic? Will I need to physically move the "anti-noise" speaker further away from the mic? Even if it cancels out the noise at the point where the "anti-noise" speaker resides, will that noise be destructively interfered with further down past the speaker in the duct?

Thanks for taking the time to read this lengthy post.

Regards,
Fred

 

[marcusl]

Welcome to the Forums, Fred!

I am not at all an expert in HVAC, but I have implemented many electronic adaptive noise cancellers (ANC's) in radar, communications and instrumentation systems. The ANC is more complicated than your reversed-phase system, but may be something to consider if higher performance is needed. Here are some observations (opinions, actually):

1. Passive noise reduction should be implemented first. Mount the motor or entire HVAC on vibration isolators, use a rubber collar to connect the duct, use duct dampening, etc. Even if this is a large job involving equipment removal, do it! You'll achieve better results, over a larger frequency range, than any band-aid applied downstream. Done well, you won't need anything else. In this case, boring old-school beats exciting new.

Example: In a lab-space shuffle, my computer gear was once moved into a previously unused small room that had a large fume-hood duct running through it. The hood ran 24/7, and the noise could be heard up and down the hall. In fact, you could literally see the sheet metal sides of the duct vibrate and jump, so intense was the noise. The solution was to build a stout box of 1" plywood around the duct, pack it full of sand, and seal it. The room was quiet and nearby occupants up and down the hall were delighted that their rooms were quieter, too.

2. Your system must put out as much power as the noise. If the duct is vibrating, for instance, then your speaker and amp must be sized to produce equal duct vibration in order for cancellation to occur. Beware that even big speakers usually don't produce well below about 10 - 20 Hz. You need to broadcast full power at all noise frequencies.

3. Try low-pass filtering your sensor signal. You only get feedback when there's a 180 degree phase shift between sensed and injected signals, which won't happen at low frequencies where the wavelength is long.

If fc = filter cutoff frequency, v = speed of sound, and d = separation between mike and spkr, you want

d << v/(2*fc) = 1/2 wavelength.

Roll off at 55 Hz if you have 1 foot separation. You'll then need to use other techniques to cancel higher frequencies.

4. Separating the speaker and microphone will eliminate feedback but will also reduce cancellation since antinoise is injected at a different spot than measured noise. An ANC is the answer--it uses sophisticated digital signal processing to determine the antinoise waveform that results in optimal sound cancellation at the microphone. Here's an article on the concept:
http://en.wikipedia.org/wiki/Adaptive_filter"
This is a last resort, since such systems are expensive and require an expert.

 

[fooser-q]

Thanks for your reply Marcus!

I have tried low-pass filtering input and output, reconfiguring the circuit and various polarity changes in an attempt to stabilize the audio output, but with no success. Near as I can figure, I need some sort of Echo cancellation filtering. We are likely going in a new direction along the lines of DSP with digital filtering to make this work.

 

[marcusl]

If you aren't getting understandable results the easy way, then DSP (which is even less intuitive) may work even worse. I'd try to understand where your feedback is coming from. What kinds of filtration are you using? What settings? Are there other, hidden, feedback paths (ground loops, etc.)?