low pass digital filters
30oct13
The mock spectrometer has a low pass
digital filter after the a/d, digital mixer section. The filters are
8x overlap filters that are symmetric about DC. The decimation can
be 2 through 1024. The coef's for a particular decimation:
- are 16 bit fixed point numbers
- there are 8xdecNum coefs in total
- There are 4 files, each containing DecNum coef. These are
mirrored about DC to give the 8*decNum coef.
The filter types available are:
- rectangle
- hamming
- hanning (this is the default that we use)
- blackman
- bartlett
The filters are generated with jeffs program
(pnet_mkdlpf_coeff).
IDL routines:
Some idl routines used with the dlpf
processing:
- pdevklpf.pro - make coef for a decimation and optionally store
it on disc
- pdevinplpf.pro - input coef for a given decimation from disc.
03oct13:plot filter shapes and aliased
power. dec=2..15
Plots were made of the filter shapes (in the time
and frequency domain) for decimations 2 through 15 using the filter
coefficients.
The plots show:
- Page 1: filter shape in the time (top) and frequency
domain bottom)
- Each of the filters is over plotted in a a different color.
- top: time domain
- This plots the filter coefficients.
- Bottom: frequency domain db plots.
- Each filter was normalized to its maximum.
- The x axis scale is the unfiltered bandwidth [-.5,
-.5] * samplingFrequency
- The dashed vertical line shows the band edge after
filtering.
- The picture plots 3 times the filtered bandwidth to show
the filter rolloff.
- Example: Decimation = 10
- x scale -.5 to .5 of sampling frequency
- filter bandwidth 1/10 -.05 to .05
- display 3*filtered bandwidth (-.15 to .15).
- Page 2. The filter shape with the aliased power over plotted.
- The horizontal axis shows the filtered bandwidth ( in units
of the sampling frequency).
- The black plot is the spectral density
- The red trace includes the aliased power.
- Each frame shows a different filter type
processing:
x101/pdev/dlpf/simDlpf.pro
<-
page up
home_~phil