by Tony » Sat Aug 09, 2008 6:20 pm
Rick,
I doubt it is electronic induced noise, because there is no direct electrical connection between flow bench and pressure transducer, just a plastic air line. The only really sensitive electrical nodes on the board are extremely short, and relatively low impedance, so are unlikely to be susceptible to radiated emissions from the motors.
Far more likely, is that it is direct acoustic coupling into the pressure transducers down the air line. That is not going to be a problem, unless there are some fairly large cyclic air pressure fluctuations due to vortex shedding, blower surge, motor rumble/vibration, or something similar. There is a lot of violence being done to the air in there.
Just realize that pressure transducers very similar to these can accurately track the exact pressure variations within the combustion chamber of a gasoline engine running at 10,000 rpm+ They are extremely fast, and very sensitive. They not only respond to static air pressure change, but acoustic sounds and the fastest small pressure fluctuation.
Anyhow, it will be impossible to get a completely noise free output, and funnily enough a bit of random noise will actually be an advantage for us. Your continuous averaging algorithm will interpolate nicely between the theoretical nine bit maximum resolution, and you will get a beautiful smooth output to much higher resolution than nine bits. In some analog to digital applications they actually deliberately add random noise or "dither" to achieve this same effect. But we get it for free!
That is exactly why I deliberately did not add any low pass filtering to the amplifier circuit board. Much better to keep the dithering least significant bits intact, and average in software. Any pre filtering slows down the overall response so you cannot then be quite so aggressive with the averaging program.
Don't worry about the noise as far as the project goes. All noise averages out to true static pressure, so your averaging program will get rid of whatever is there very effectively. All of us will have plenty of noise, to a greater or lesser degree. Just use the longest averaging time you feel comfortable with in your algorithm, as far as speed of response goes, and it will work fine.
Also known as the infamous "Warpspeed" on some other Forums.