Determining Sample Frequency

[wantolearn]

I need to determine the sample frequency for measuring pressure pulsations. I have a V6 running at 3300 RPM (maximum). If this produces 165 exhaust events every revolution, how do you calculate the resolution necessary to capture all the significant data?

 

[CESSNA1]

WANTTOLEARN: The resolution of the sample is 1/N where N is the total number of samples. If you take data with 10000 samples then the resolution is 1/10000 or .0001 Hz. The other part of your question is more difficult. You need to filter at the highest frequency of interest. That is for you to decide. Say you are interested up to 2000 Hz. Your antialiasing filter should be 2000 hertz, BUT filters are rated at the 2 Db down point. 3 Db corresponds to a 50% decrease in power. That means you have to filter at a higher frquency, up at 3000 or 4000 Hz to capture the data you want. The Nyquist Criteria says that to minimize aliasing you should sample at twice the frequency of interest or sample at 6000 - 8000 Hz. Shannon's theorem states that in order to reproduce signal exactly there should be at least 2 samples at the highest frequency of interest. From a practical standpoint however to be within 5% or the actual signal you should sample at 10 times the highest frequency of interest, or in this case 20000 Hz.

I suggest ou look at

http://www.tti4edu.com

and sign up for some of their courses.

Regards
Dave

 

[wantolearn]

Thanks Dave. I think I posed my question incorrectly. I guess I'm looking for verification of the highest frequency of interest for a V6 running at 3300RPM. Since the V6 will incur 3 exhaust events every revolution, I've calculated 165Hz as the maximum frequency of interest. Does this sound correct? If this is the case, I should use a tranducer that samples at a minimum 1.7KHz?

p.s. thanks for the info on the TTI......

Steve

 

[CESSNA1]

WANTTOLEARN: It depends on what you are looking to find, and how you define an event. How did you determine 165 Hz. 3300 rpm is 55 rps (Hz). Assuming one intake valve opens and closes (2 events) and one exhaust valve opens and closes (2 events) then I see 4*55=220 Hz. Also for a four cycle engine there are 2 crank revolutions needed to complete the valve operation.

Regards
Dave

 

[IRstuff]

That assumes that the pressure pulses are purely sinusoidal, which I assume are not. Since you know nothing about the actual timing, I would suggest that you use something like 100 to 500 times the frequency.

You can always throw out extra data. Trying to infer missing data is always significantly harder.

TTFN

 

[wantolearn]

CESSNA1: I'm focusing my evaluation on back-pressure in the exhaust for a turbocharged diesel system. I determined the frequency based on the pulse of the exhaust valve opening (assuming the peak pressure pulse occurs during opening events and not closing). Considering 1 exhaust event per cylinder and that there are 3 firings per revolution for a 4-cycle engine, I arrived at 165 Hz (3300rpm/60*3firings).

Should I re-think this?

Regards,
Steve

 

[Rob45]

You calculated correctly: 165 Hz is in fact your firing frequency.
You will very likely however have things you're interested in occurring at several multiples of the firing frequency.
I would suggest looking up to the 9th order, at least: i.e., 3 x 165 = 495 Hz.

This is not of course the resolution.
But if you are as you say interested in the backpressure, not acoustics, you could get the information you need with a far lower sampling rate; just a few times per second would suffice.

BTW, it's not the transducer that determines the sampling rate, in general, but the data acquisition system itself.

When I do this type testing, I use a pressure transducer from PCB Piezotronics 1501A01FJ15 which nominally is a 0-15 psi static pressure sensor but appears to have a -15 - +15 psi range, and has good dynamic response.

 

[wantolearn]

We'll do. Thanks again Rob45, IRstuff and CESSNA1.