Discrepancies between MathScript and Matlab decimate function

[PatCarbo]

I would like to obtain in Labview the same output as Matlab's decimate function.

I was expecting that MathScript's "decimate" function would do that but this doesn't seem to be the case.

To make comparison easier I will be using the decimate function with three parameters.

decimate(A,B,C)

Where

                A = [1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 16, 16]

                B = 2

                C = 8 (Chebyshev type I, 8^th order filter)

Matlab returns:

1.2708

2.1873

3.2438

4.1731

5.2076

6.1679

7.1663

8.1662

9.1254

10.1559

11.1004

12.1261

13.0992

14.0648

15.1451

15.8024

Labview returns:

1.0

1.72746

2.75001

3.75001

4.75001

6.75001

7.75001

8.75001

9.75001

10.75

11.75

12.75

13.7495

14.7619

15.6993

Could anyone explain how to get from one to the other?

Regards,

[IanS]

Hi,

I'm an applications engineer at NIUK and I've been looking into this.

We don't have matlab, but I have reproduced the problem in mathscript. It seems that whats going on under the hood is slightly different between the two, as decimate is a high level function. Perhaps mathscript doesn't generate quite the same filter. 

I took a look at the matlab help pages to see what the component functions are. By replicating these lower level functions in mathscript we can get the same output!

Assuming you have used A, B and C as before, you can use the following mathscript code:

[g,f]=iir_cheby1(C,0.05,0.8/B)
y=filtfilt(g,f,A)
z=downsample(y,b,b-1)

 where z will give the desired output.

Line 1 generates a chebyshev type 1 filter of order C, with ripple 0.05 and cutoff 0.8/B.

Line 2 applies zero phase filtering.

Line 3 takes out every 'b'th element, starting at index b-1. (zero indexed array)

Please do test this out and compare to matlab, and keep me posted.

[PatCarbo]

Ian,

Thanks for looking at this issue.

Our signal processing engineers use mainly Matlab but the products we make use LabVIEW.

I appreciate MathScript availability which allows for quick algorithm implementation.

We suspected the filtering to be different and your code snippet will help us validate LabVIEW code against Matlab scripts.

Regards,