A2713/A2765: dps with high freq receivers
21apr13: jumps in
adjacent bands after processing:
04nov12: 1 mock box with glitches.
ripple in polB 1 box for 1 pattern
A2713 is doing double position switching on
sources using the high freq receivers (lband to xband). The mock
spectrometer is used to record the data. A2765 is an extension of
a2713:The setup is:
- both mock groups in use (g0 and g1).
- hi res mode to (86, 57 MHz bw) depending on rcvr.
- offset the bands to get rid of Dc spike
- Use new pshift parameter to downshift everyother butterfly
- 1 second dumps
- 240 seconds on and off source and calibrator.
- 165MHz clock freq was used.
21apr13: jumps i adjacent bands after
A2765 was doing double position switching on
arp220 using 8-9 Ghz of the xband receiver. Both groups of the
mock spectrometer were used to cover the 1 Ghz (with 82/8192 or
10Khz resolution). The data was processed with masdpsp
procedure and then plotted.
There were 3 dps patterns taken.
Chris mentioned that the different bands of
given pattern did not have close to the same values.
the plots show the dps results for
the 3 patterns (.ps) (.pdf):
- Each frame is a separate dps pattern (
- Black is polA, red is polB
- The data has been smoothed by 9 channels
- The green lines are mock group 0, the blue lines are mock
- The solid line is the start of a spectra, the dashed line is
the end of a spectra.
- The first 2 patterns showed a positive slope vs frequency. The
3 pattern did not show this.
- The jumps occurred mainly in the first two patterns.
- PolA and polB had similar jumps and slopes.
- The jumps occurred at mixer boundaries, not group boundaries.
- the 2nd if/lo creates 7 separate freq bands.
- Each of these bands is then sent to a group 0 and group 1
- Since the jumps occurred on mixer boundaries, the jumps are
not coming from group 0 or group 1 of the mocks.
- The problem is probably not in the mock boxes.
- The jumps could come from the mock mixers or the if/lo mixers.
- The dashed blue line shows the end of each mixer band. Reading
left to right
- band 1, low, synth 1
- band 2, synth 5 , jump from band1)
- band 3,synth 2 (no jump from band 2)
- band 4,synth 6, jump from band3
- band 5,synth 3, jump from band 4
- band 6,synth 7, jump from band 5
- band 7,synth 4, small jump from band 6
- The jumps in the 1st two patterns occurred at the same band
- Assuming that the freq slope is wrong it could be caused by:
- Something that has a freq dependence that changes over time
- the calibrator was taken before the source.
- I wonder if a loose connection has a frequency dependence.
04nov12:1 mock with glitches. 1 with
ripple in polB for 1 pattern
04nov12 was the first day. The cband receiver
was used. Some problems seen were:
- b1s1g1 polA (grp 1 second box pdev-108) had lots of 1 channel
spikes in the spectra.
- These were bits flipping in the fpga.You can tell since the
jumps were discrete.
- We've seen this before when hires mode (decimate bw below
172MHz) was used.
- The normal solution to this was to decrease the bandwidth
from 172 MHz to around 165 MHz. That would normally stop the
- They were already using165 MHz clock rate so this was not
- I ran the diagnostics (rdcomp).
b1s1g1 (pdevs-108) failed the diags even at 160MHz.
- We replaced this pdev spectrometer and the diags then ran
- Ripple was seen in 1 band of the
1st dps scan (.ps) (.pdf)
- Page 1: dps spectra for each dps pattern
- The vertical scale is in units of the calibrator.
- Red in polA, green is polB
- The data has been smoothed by 21 channels 10.5 Khz*21=220
- Each frame overplots the 14 spectra of each dps pattern
- frame 1-3 is dps pattern 1 thru 3
- The red vertical lines around 4200 is the bad mock box
(b1s1g1). It was replaced.
- There is lots of rfi 4000-4200 MHz.
- there is a ripple in polB (green) near 4850 MHz of the 1st
dps scan. it is not seen in the other dps patterns.
- Page 2: blowup of 4850 MHz ripple dps 1 polB
- Top: this shows the ratio of bandPass calibrator on/off
- Bottom: ration of srcOnSpc/srcOffSpc.
- polB is strongest in srcOn/Off
- you can see a little of it in polA
- The period of the ripple is about 10.5 MHz. (about 15
meters in air).
- Page 3: median spectral power vs time for the 300 1 second
samples on and off source.
- Top on Source
- bottom Off source.
- PolA seems to wonder more on source than polB and off
- PolA was previously giving us stability problems (before
the elbow in the dewar was replaced).
- The ripple problem is in polB so whatever the variation
is, it isn't causing the ripple.
processing: usrproj/a2712/doall.pro, dbg_121104.pro
- 1 mock box had trouble with bit flips (b1s1g1,pdevs-108,
polA). It was replaced 05nov12.
- 1 mock box (b5s1g1) showed 10 MHz ripple in 1 dps
- It was mainly in the onSource data
- It did not repeat during the next 2 scans,
- cband polA may be wandering around a bit more than polB
(either gain or Tsys).