Need help from Electrical Engineers Please (CFC60 and CFC36 reqs)

[WCIGAFETW]

Hello everyone,

I am needed to apply a filter to some acceleration data. That I think I can handle (I have chosen a Butterworth). What I do not understand are the CFC (Channel Filter Class) specifications. In particular, I do not know what a "Stop Damping" number specified in -dB is. Below is an example of a CFC60.

Filter type
CFC 60

Filter parameters
3 dB limit frequency: 100 Hz
Stop damping: –30 dB
Sampling frequency: At least 600 Hz

My sensor samples at 400 Hz which puts my CFC at about 36. Please correct me if I am wrong. At CFC 36, what would my parameters be? Do they define cut-off frequencies?

Thank you for any help. Google-skill"z" are failing me.

David

 

[NascentOxygen]

Hi WCIGAFETW, welcome to Physics Forums.

I can't help much with this, but maybe that is the gain in the stop-band? There is a pdf that seems to cover what you are doing, it might be of some help. http://203.158.253.140/media/e-Book/Engineer/Automotive/Vehicle%20Crash%20Dynamics/0687_PDF_C01.pdf

 

[Enthalpy]

Why has my answer disappeared?

A butterworth is the worst possible choice because it rings much worse than the others, as compared at the same selectivity, and this may be important for an accelerometer filter. A Chebychev is better, an elliptic even better; the good ones among known names are minimum-Q elliptic, hourglass, and the best is inverse (or type II) Chebychev.

Your "stop damping" must be the attenuation in the stop band. To define a simple low-pass, you must still tell the corner frequency of the stop band.

The main difficulty here is that your software uses terms uncommon to the filter people.

 

[WCIGAFETW]

Thanks for your help Enthalpy and NascentOxygen. I have been doing some reading into this stuff and think the only real parameters I need to be concerned with are the cutoff frequency and the order of the filter. I am still running a Butterworth (sorry Enthalpy) as it is the filter used by SAE for all of their analysis.

I found a 2nd order filter online with the coefficients explained and applied it to my data.

http://biomech.byu.edu/Portals/83/docs/exsc663/part01/filtering_considerations.pdf

The equation and coefs are simple enough. However, I am making assumptions as to what a CFC36 would be. I believe the parameters for that filter class would be as follows:

-3 dB limit frequency: 60 Hz
Stop damping: –30 dB
Sampling frequency: At least 360 Hz

Are cuttoff frequencies and this "-3 dB limit frequency" the same? Are these two things synonymous with rolloff frequencies? If these numbers and my assumptions are correct, my cutoff frequency to be applied is 60 Hz.

Thanks again for the assistance.

David

 

[NascentOxygen]

Are cuttoff frequencies and this "-3 dB limit frequency" the same? Are these two things synonymous with rolloff frequencies?

-3dB and rolloff frequency are I'd say synonymous with "corner frequency". But "cutoff", I'm hesitant to say one way or the other.

If these numbers and my assumptions are correct, my cutoff frequency to be applied is 60 Hz.

If this is performed in software, then you could at a later stage easily double or halve the -3dB frequency and see whether there is any significant change in your results. If there is little change, then it's probably safe to conclude that you are capturing all significant components.

 

[NascentOxygen]

A butterworth is the worst possible choice because it rings much worse than the others, as compared at the same selectivity, and this may be important for an accelerometer filter. A Chebychev is better, an elliptic even better; the good ones among known names are minimum-Q elliptic, hourglass, and the best is inverse (or type II) Chebychev.

The Tchebychev is very underdamped, so shows a lot of ringing compared to a Butterworth of the same order.

 

[Antiphon]

Please tell us that you're not programming the control system of a real vehicle.

 

[Enthalpy]

The Tchebychev is very underdamped, so shows a lot of ringing compared to a Butterworth of the same order.

The aim of a filter is not to have a certain number of poles, but a certain selectivity, for which even the type I Chebychev has fewer poles and lower Q factors. This must be the reason why it rings less than a Butterworth at identical selectivity, and an elliptic even less.

 

[Enthalpy]

Please tell us that you're not programming the control system of a real vehicle.

Or crash tests! I believe to remember such names like CFC36.

Fun: I studied and defined filters for CrashSoft, with much better ringing than the common ones, with a trainee called David like the OP here.