Active Silencer/Muffler Design 4

[RodH]

Hi everyone. I'm in the process of building up an active silencer for a typical mid-size car. Has anyone had any experience with doing this before and can reveal tips/gotchas to help get this project off the ground?

The first prototype will be added to a typical "high flow" (read noisy) passive silencer with the aim of combining both passive and active silencing. I've yet to select processor/circuitry/sensors, etc, and recommendations are welcome.

I intend to use the filtered-x algorithm, with a microphone at the exhaust tip, synchronisation from the crank firing rate, and a loudspeaker trying to cancel in the midpipe (attached through a stub) ... Let me know if you know a better way!

TIA,
Rod

 

[maxh]

Hi RodH
Sorry don't know much about Automotive applications, but I have seen this done on diesel engine gen sets. However, here the speaker was put at the end of the exhaust pipe to get around temperature limitations of the speaker design.

The system I saw was like yours and used a single mic although their software characterised the exhaust due to the single speed nature of the machine it was designed for.

If I can find the references tomorrow I will post it for you - it was 5 years ago however !!!

 

[GregLocock]

Can you briefly describe the filtered x method? or point to an on-line resource?



Cheers

Greg Locock

 

[RodH]

There's many out there, this one's fairly concise:

http://www.its.bth.se/courses/etc004/fxlms.pdf

One of the key things is, this algorithm can take the crank angle sensor as input, avoiding a microphone in the worst (hot) part of the exhaust system. The crank angle sensor should be noise free too.

Rod

 

[Highspeed]

Hi RodH,
My experience on Active Silencer wasn't on exhaust pipes but here, few tips that could save time
Before going into the LMS or equivalent stuff, I did put a lot of efforts on making sure that it could work, meaning that with a well known simple(sine) noise signal at the inlet, the setup can effectively cancel it.Using a simple microncontroller with AD/DA allowing a phase shift can do the job as long as you use it with your synchronization
on top of it it will allow you to have a model of your setup.
On my setup I did have a tacho signal and that made things a lot easier than having a mike at the noise source.
Also, when adding extra passive silencers to the system, I did find that it was easier to put them after the Active noise cancellation module.
In my system, I finally didn't implement LMS / DSP to get to the level required but kept things simple using the original µcontroler having Lookup tables recorded during the device calibration based on the assumption that my setup would not change over time.
A good article, you've probably read :

http://www.analog.com/library/analogDialogue/archives/34-02/noise/

Good luck

 

[GregLocock]

The broadband results in that article are rather good, in fact even the narrow band results are impressive. I suspect that is because the system is running at a stable speed, whereas on a car you have to cope with slew rates. I don't think we ever reliably got more than 10 dB (which is still pretty impressive subjectively).



Cheers

Greg Locock

 

[bluenoted]

Hi! RodH

Sorry, don't know much about ANC muffler, but I know the man working in the ANC Lab got the experience on a ANC muffler. Here's his e-mail.

choong@ihanyang.ac.kr

Good luck!

 

[davefitz]

Instead of active silencer using a loudspeaker, why not utilize an adjustable helmholtz resonater. The resonator adjoint volume or orifice area can be adjusted in real time baased on monitored peak frequency from the tail pipe micorphone ( actually , place the microphone closer to the source, or simply assume the peak frequency is proportional to engine speed).

 

[RodH]

Thanks for bringing up the idea of a resonator - from what I've read these are not as good as ANC because they only deal with one frequency per resonator (making it very complex for signals other than tones). I'll be trying to make the ANC work if I can.

Cheers,
Rod

 

[Highspeed]

Another idea that can also help a little in ANC design:
If you feed your system with a tacho signal and the output mike only you'll have a delay in noise cancellation when the motor "load" will change (ie : at same RPM, full power vs Low power gives a different noise level), If from your injection you can have an image of the load and feed your LMS with it, then you'll kill 1 or 2 dB's out specially on transcient modes.

Helmholtz resonator:
If, on your setup, the connection from your exhaust pipe to where you have your loud speaker can be seen as an Helmholtz resonator in the usable bandwidth, that will not help the LMS, specially if the resonator has a high Q:
in that configuration the resonator will act as a band stop in a narrow band with funny things on phase shifts that LMS dislike.
Cheers.

 

[RodH]

Are you saying I should also take MAP or TPS as well as RPM? Teaching the system this transient behaviour may be difficult, no?

I suspect using a fuel injector's signal pulse won't work under all conditions. Although it has variations in both pulse width (load) and frequency (like the tacho) this will be bad in overrun where there may be total fuel cut.

Also, I'd need to use multiple injector signals which gets confusing as they switch to & from sequential/batch control strategy.

Can you expand more on your idea? I'm interested.
Thanks,
Rod

 

[Highspeed]

Rod,
Here is the idea:
The noise source is dependant among other parameters of Power and RPM's
let's assume that you can have a model of the source noise vs RPM & torque or output power.
Whatever technique you are going to use to actually cancel the noise will work better in transcient if it has this knowledge embedded as opposed to acquiring this knowledge.

As opposed to the classical 2 mikes design, the tacho only design has some lag for transcient because the system needs to see at the outlet that the noise has increased to feedforward it and is having some (kind of) integral parameter taking care of that for the upcoming sound pressure waves.

I am not in the automotive industry so I don't know how to capture from the motor control an image of the output torque or power and I don't even know if it is feasable but if it is, then I think it will improve the system.

From my experience that is based on a Blower at variable speeds, I did not used LMS as every thing was teached during a calibration phase and embedded in 2 entries lookup tables, entries beeing RPM & current consumption of the blower, ouput beeing a timed cancellation waveform with a gain adjusted by the integral of the rms value of the output mike "band pass" filtered at F0.
result : from 6 to 10 dB's killed but fundammental frequency killed by more than 12 dB's
Hope this helps ...
Cheers

 

[GregLocock]

Just use TPS (throttle position sensor) and then fudge in a torque vs TPS curve.

However, in practice, cancellation even in first gear full throttle is quite acceptable using the error mics as the amplitude control. - yes, you are probably only getting 3 instead of 10 dB cancellation, but it is only for a couple of seconds.

Note that the slew rate will be a function of which gear you are in so in this complex version you'll need a gear indicator switch.

If I sound a bit negative, well, this seems to me to be a nice refinement rather than a crucial detail.

Cheers

Greg Locock

 

[snowrat]

You will want to check out Leuven University.

http://www.mech.kuleuven.ac.be/pma/research/mod/topics/topic03/default_en.phtml

Papers from Sas R. Boonen are quite good

 

[RodH]

Thanks Snowrat.
IIRC, the Lueven work uses a variable restriction rather than noise cancellation. To my mind this reduces the benefits of active noise control on a silencer (muffler) because backpressure is probably the last thing you want on an IC engine. I don't doubt it could be made to work, but I suspect the cancellation method is superior so will pursue that.
Cheers,
Rod

 

[Rob45]

Hey Rod, I don't mean to rain on your parade, but there has been quite a lot of ultimately unsuccessful work done in attempting to do exactly what you're working on.

I reviewed in person the efforts of the two attempts to do this commercially in the U.S., by NCT (Noise Cancellation Technologies, Lanthicum, Maryland, no longer in the vehicle noise business) and ANVT (Active Noise something-or-other in Phoenix, no longer in business at all).
Only the former's system worked even passably well, since it used a throttle position sensor along with a tach pickup to "feed-forward" - really, to use a look-up table, to guess at what anti-noise signal would be required. Still, it was a very "throaty" sounding system, and not in a good way, since it cancelled the fundamental but didn't do much for odd-order harmonics, and the lag in response was quite audibly noticeable, and also rather unpleasant.
As for the ANVT system: without a tps,looking only at rpm and the tailpipe noise, it was fine at constant speed (they only wanted to demonstrate it on a chassis dyno, not on the road) but absolutely awful under any kind of transient operation.

My recommendation was that my company not invest anything in either technology.
Chevrolet also closely examined NCT's noise cancellation technology in planning the (current) C5 Corvette, since Walker Muffler had licensed it, but they ultimately went to a conventional system - which, BTW, I designed.

 

[RodH]

Thanks for sharing your insight, Rob45. It's good to know the TPS input proved useful (I was going to use MAP, but may look at both now).

Do you happen to know if (a) these systems used some passive noise control as well, or just active, and (b) if the failure to cancel harmonics was due to limitations in the system response time, or some other factor?

Cheers,
Rod

 

[GregLocock]

I only used a very old system (1988) and we were heavily limited by computing power. Off the top of my head we struggled with cancelling random signals greater than about 100 Hz, even with only two transmission paths, or with engine nosie above 200 Hz. We could do multiple harmonics within that limitation. Slew rate was a big deal.





Cheers

Greg Locock

 

[Rob45]

"Do you happen to know if (a) these systems used some passive noise control as well, or just active, and (b) if the failure to cancel harmonics was due to limitations in the system response time, or some other factor?"[p]
Rod:
Both the systems I described used rather large "glass-pack" - type passive silencers to reduce flow noise and miscellaneous other higher frequencies, since as I recall, and also as Greg notes, neither the NCT nor the ANVT systems were much good at attenuating frequencies above 100 Hz or so.
The failure to cancel harmonics had to do with this latter effect, phase error: much of the harmonics were at frequencies that the active noise system could not address, and these systems were only demo'd on 6-cyl and 4-cyl engines, respectively. A V-8 would have been worse, and the guys working on this knew that.

For a V-8, operating at (for convenience in calculation) 1500 RPM, 1st-order firing is 100 Hz, 2-times is 200, -times is 300; and it's not uncommon for there to be 2nd and 3rd-order (8th and 12th rotational order) components in the tailpipe noise spectrum. Not to mention some "odd" order parts such as 1.5 and 4.5-times firing, though these can often be tuned out with good piping layout.
-----------------------------------
I will tell you where I saw an active noise system perform very well, very effectively:
NCT, again, took several of us to the Baltimore railyards, where a high-horsepower gasoline-driven air compressor (150hp or so) was being used to vacuum grain (wheat, etc.) from railcars and transfer it to trucks.
These air compressors were incredibly noisy in spite of having mufflers on the air inlet to the compressor, a big (if I recall) Roots-type blower - something on the order of 125 dB(A) at 3 feet or so.
One of these compressors had been fitted with an active noise cancellation system consisting of a speaker (maybe two) at the compressor inlet. Out of curiosity, I walked back and forth between the active-noise compressor and a conventional one, stepping off the distance, and found that the place where they both seemed to be equally loud was about 8 feet from the active noise one, and over 100' from the other!
That was an effective noise cancellation system!