We tested a setup for:
- running puppi in baseband mode with 213 MHz
bandwidth while also..
- running wapps in spectral dump mode as well as the
- For HI absorption in our galaxy, we needed 1420 MHz in
the puppi band
- A wide bandwidth was also wanted for scintillations
- The absorption occurs during the burst, so we
needed baseband data.
- For baseband samping we needed an anti aliasing filter
for the lbw data.
- we found a bandpass filter 140 MHz bw, centered at
1600 MHz that we put in the first IF (1-2 Ghz) that goes
- It was 30 db down 1505 and1697.
- This crosses the 200 MHz boundary (1400 -1600)... so
we increased the clock from 200 to 213 MHz
- the band 1491 to 1704 will now fold down to bandband
(7 folds) so the final band is flipped..(IF1 is not
- The center of the 213 MHz band in IF1 is 1597.5 Mhz
- The filter was inserted after the 1st 10db pad and
before the first amp int the puppi chassis.
- We needed more power so we removed the 10db pad an
put in a circulator.
- The center of the 7th nyquist zone 1491-1704 is
- We wanted to put 1420 MHz sky freq here..
- the 1st IF is centered at 1500 Mhz
- We mix the requested sky freq to 1500. IF1 1597.5 is
1420Mhz Sky freq, So IF1 1500 is 1322.5 sky Freq. This
is the rf freq for cima
- The wapps then selected all of lband using the 2nd
- centering 1322.5 Sky in the center of the 1Ghz 1st IF,
gave an IF1 sky range of 822.5 -> 1822.5. More than
- puppi was run in command line mode (cima did not
- adjust the wapp power levels
- setup puppi for:
- bw='-213', raw mode, cfr 1420 MHz, 32 channels
- adjust puppi power using command line
- start things going
- you need to manually stop puppi.
Testing the setup:
- we took 1 10 sec scan while transmitting a birdie at 1423
MHz (into the lbw receiver).
- a second 10 sec scan was taken while transmitting a birdie
of 1433 MHz
- These would check if the frequency was centered correctly,
and that the channel spacing was correct.
Looking at the data:
- The raw files were copied from the gpu's to megs3.
- the idl routines to read raw puppi data were used to look
at the data.
- 2^20 length spectra were computed for the 2 gpu
channels that had the 2 different birdies
- gpu07 chan 0 for 1423
- gpu06 chan 2 for 1433
- DC was removed, the spectra were computed and then
The plots show the
spectra for the two tones (.ps) (.pdf):
- The bandwidth of each channel was : 213/32=6.65625 MHz
- 2^20 samples gave .157 seconds of data.. a freq
resolution of 6.3 Hz.
- Page 1: 1423 Tone. gpu07
- top: Spectra for polA,B (black,red). smoothed to 812.5
hz (so plot file didn't get too big)
- The birdie on the right is the 1423 tone, the
birdie on the left is an alias (probably from the pfb
- middle: spectra averaging polA and polB.
- The small bump at 1420.4 MHz is probably galactic
- Bottom: hires plot of 1423 birdie. 6.3 Hz resolution.
The birdie is centered where it is supposed to be
- Page 2: 1433 Tone . gpu06
- Same plots as page 1. The 1433 is sitting in the correct
- There is no aliased birdie .. probably because the tone
is not close to the edge.
- Running puppi with a clock of 213 MHz in baseband mode
gave the correct results for two birdies in different
- So the frequency center and the span is correct
- The birdie close to the edge of a gpu channel showed
appreciable aliasing in the pfb channel. I wonder if this is
normal , or the increased clock caused this?