Puppi uses an
interleaved AtoD converter. Two AtoD samples are taken for each
clock cycle.
SBW observations (on 7jan13) showed an
image of the
2140 rfi at about 15 db below the rfi (
.png).
- The band center is 2030 MHz (the band is flipped in
freq)
- the 2140 rfi also shows up as an image at 2030 -
(2140-2030) = 1920 MHz.
- Paul demorest pointed out that this image could be created
by a problem in the interleaved sampler.
A simple simulation was done to see
how sensitive interleaved sampling was to amplitude
errors (.ps) (
.pdf):
- The simulation used:
- 16384 points with a 110 MHz sine wave (amplitude 1).
- Gaussian noise was also added.
- The sine wave was broken up into even and odd samples.
- The odd voltage samples were then multiplied by various
errors (1.0,1.05,1.1,... to 1.45).
- The spectrum was then computed and plotted.
- Page 1:
- Each panel shows the spectrum with the rfi (at 2140) and
the birdie (at 1920).
- These spectra are not flipped.
- The vertical scale is db below the peak of the 2140 rfi.
- Page 2:
- Plots the image strength (db below rfi peak) vs voltage
error (1+xaxis).
- The image strength at error=0. is the noise floor.
Summary:
- Amplitude scaling errors on the interleaved sampling will
create an image.
- a/d offset errors added a DC spike.
- To get a 15db image you need an very large amplitude
error (1.4 times the correct amplitude)
- Tests were also made modifying the phase of the odd
samples.
- This created images, but they were a lot less than the
amplitude errors.
- Hanning smoothing the data before transforming can get rid
of most of the image.
processing: x101/130114/puppisim.pro