DFD integer transfer function for CIC and HBF

[Pappion]

A designer gave me Integer Transfer Functions for CIC and HBF filters. I'm new to this package and don't see how to enter the TF or edit the filters output.

I tried the DFD Build MRate Filter from TF but performance is lacking and there is no way to verify it's an CIC.

I get half the SINAD I should vs. your stock CIC. The stock CIC does give some spectal smearing of the fundemental (FFT view). CIC is not as bad as home brew Moving Avgerage Filte which gives great SINAD but a lot of smearing.

Also the MRate Filtering for Single Block (array)  has an Extension Type padding Control. I need to crop the extension to get good SINAD. Many of the filters have dead saple on one or both ends and need croping. Why can't I get a selection to do this, or at least an integer values for array subset?

[Myr.i.am.]

Hey Pappion,

Would it be possible to get a few screenshots of what you're seeing in order to better understand the issues you're having?  Also, do you happen to be using any hardware?

Thanks,

Myriam

Applications Engineer

National Instruments

[Pappion]

No hardware yet. I plan to use an HSDIO to capture ADC stream output. I am using a designers simulated input, and another designers integer coefficents.  I can't post most of the details.

Attached doc shows spectral smearing from my custom Moving Avergae filter with LabView generated sinewave. Of course it gets much larger for more iterations and decimation stages.

Odd how the FFT shows it in vi too. Sinad is much better, but cropping adds smear?

[Myr.i.am.]

Hi Pappion,

I apologize, I'm still having some trouble understanding your question.  I can definitely help you out with direct LabVIEW questions, but I'm not very well-versed in signal processing.

In regards to the issue with cropping, wouldn't you expect the resulting graph to look different if you removed data points from the signal?

Myriam D.

Applications Engineer

National Instruments

[Pappion]

Expect? It was 20 years ago I did something similar with Power-of-2 FFTs. The FFT fundamental was thinner the more samples that were thrown at it. Your FFT is not x2 power samples. It's skirt width is very dependant on the samples being nearly an interger multiple of fundamental. I should think the window would have helped with the non-integer discontinuity, and larger samples should also help. All in all the FFT display is just a side show.

More important is the amount of cropping needed for non-corrupt SINAD measurements. Variables?: filter type, order, samples, for one or both ends? I don't have the time or patience to test everything.

Back to the HBF. Different filters can be selected, but I can't enter my designers integer coefficients.

Back to the CIC. Orders can be selected but I can't enter designers integer coefficients. I also can't readback the coeffiecients.

In summary: The DFD package is worthless to me.

I am now confident my custom filters & decimators can find ADC Full Scale, per AES17, and it's SINAD can be verified.

The good capture files can be seperated from the junk, labeled, and sent on to the designers.

Also this justifies their efforts to provide me with Matlab DLLs, instead of DFD.

[Myr.i.am.]

Hi Pappion,

To obtain more information about this, I think it would be best to contact NI Support directly.  Please see my private message for more information.

Thanks,

Myriam

Applications Engineer

National Instruments