Page 2: Blowup of 3 cycles of peak,pulse,average power
an
duty cycle
Page 3: Measure how long it took to do things:
Top: peak power
the X axis uses the time stamps written to the power meter
logfile
The dashed red lines are the computed length of time each
ipp
step should have taken.
Middle: the duty cycle:
The long noisy 20 sec duration was the coded long pulse
(20
to 40 secs), then the ipp steps began (at 40 seconds).
Bottom: the time step between samples returned by the power
meter.
Discussion:
The power meter setup:
430 MHz
13 db defined pulse
2000 averages (buffers) are requested.
The .5 MHz sampling of the 10MHz video IF is done.
A buffer of 125 samples is acquired and then sent to
the
pc.
It is sampled at 2 usecs. So one buffer takes 256 usecs.
There are actually 2 samplers: one looking at the
signal and a second looking with a pad in front of it so the
system can
covert the -60dbm to 20 dbm range.
after 2000 buffers (512 milliseconds) the pulse info is
computed:
the sampled data are sorted by amplitude
The first point is the peak power.
We've set the pulse threshold to 13 db so the program
searches for the 1st point 13 db below the peak (or first
point). All
points from here to the start are in the pulse.
The average of these points is the pulse power.
All points from the 13db point to the end are out of the
pulse.
They are used to compute the duty cycle, and the average
power.
Looking at the time between samples (bottom plot):
It always took .6 seconds independent of the number of ipps
in
.6 seconds.
The difference between the sampling time (512 ms)
and
the measured time (600 ms) is the overhead.
overhead is about 20%.
To increase the signal to noise we should probably switch the
filter to 2 MHz. This would not affect the pulse too much and it
would
increase the signal to noise.
There is a 2nd mode called pulse profiling mode. It uses
interleaved under sampling:.
the user specifies the number of points in a sweep (say
10000).
The routine taken 10000/128 buffers of data
After each buffer the device increments the a/d clock freq
by
20 ns. This allows for the interleaved sampling.
Notes:
Data from the various experiments is stored at: /share/aserv00/pwrmet