Az, za encoder birdies.
Links to pages:
az swings showed comb with old encoders
Khz comb found at 327 Mhz.
spins show comb is outside the dome.
saw comb at 430 MHz with new encoder
comb strength new,old encoders using 327 Mhz.
061010: strengths of birdies from
old encoder measured in the lab
from azimuth encoders
161011: new encoder rfi measured in screen room.
180601: rfi in 430 rcvr from encoder
180605: turn off vertex encoders one at a time
looking for rfi.
The vertex drive system (az, dome, carriage
house) use encoders to determine the location of the axis.
They are located at:
The encoders are connected to the vertex shelter via shielded
- azimuth: The encoder rack gear runs around the inside of the
platform ring girder. There are two azimuth encoders (azenc1,
azenc2) that are mounted on the top of the azimuth arm and run
in the encoder rack gear. The encoders are fixed relative
to the azimuth arm. Azenc1 is mounted on the dome side of
the azimuth arm, azenc2 is mounted on the carriage house side of
the azimuth arm (next to the vertex shelter). The azimuth
pointing uses azenc1. Azenc2 measures the bending in the
- Gregorian: The elevation rack gear sits at the bottom of the
azimuth arm. The gregorian encoder is mounted on the
gregorian trolleys and rides on the elevation rack gear. It is
fixed relative to the dome.
- Carriage house: Same rack gear as the gregorian (but on the
other side of the azimuth arm). Encoder mounted on the carriage
We 05oct06 we discovered that the az,za encoders
generate a comb at 614 Khz. Some of the comb elements had been seen
before that time (but we didn't know where they were coming from).
oct05: az swings 327:
were done to look at the 327 Mhz birdies. About 7 birdies
had the same az dependence. At the time we did not know
that these birdies were coming from the az,za encoders. This
was determined later when the comb frequencies were measured
with the new encoder. The 7 birdies from oct05 all sat at one
of the comb frequencies.
02aug06: new azimuth encoder
The old encoders were getting water in them and failing from
corrosion. A new waterproof encoder was purchased (same
company). It had a continuous metal case (that we
thought would be better for rfi shielding).
sep06: 327 pulsar observations saw really bad
03oct06: data was taken with the 327 Mhz receiver.
was seen with a spacing of 614 Khz.
04oct06: Using 327
az spins and za motion showed that the birdie was not
coming from inside the dome (since it had and az and za
614 khz comb in seen in the 430 receiver while new encoder is
installed. A2125 did 430 mapping (with the dome) on
04oct06 when the new encoder was installed and again on 05oct06
(night) when the old encoder was back in. The 614 Khz come was
seen when the new encoder was in. It may or may not have been
seen with the old encoder back in (some comb elements were
05oct06: We found that the comb was coming from the
az2 encoder. We replaced the "new" az2 encoder with the "Old"
encoder around 1pm.
05oct06: after replacing the new encoder with the old
encoder, data was taken to see how
the 614 Khz comb is with just the old encoders. Most of
the comb elements disappeared. The ones that had been seen back
in oct05 remained.
05oct06: In the lab
We saw that the new encoder had a 614 Khz comb. We
opened the encoder casing and saw the comb go to > 1000 Mhz.
At 327 Mhz the strongest comb element was at 326.269 Mhz. At 430
Mhz the strongest comb element was at 432.539 Mhz. When we
closed the casing the 327 Mhz comb was strong (> 10 db). It
was probably coming out of the cable.
We opened the case of the old encoder and saw a few
elements of the comb were present. The strongest was at
324.415. When we closed the encoder, we could still see
the strong birdies (though not as strong as the new encoder.
The new encoder has all the elements of the comb while the
old encoder only has a few of these. There is an offset of
about 4 khz between the birdies from the old encoder and the
- 06-07: Old, new encoders, filters.
- at some point in time they stopped making the "Old" encoder
(with the 614 KHz comb).
- The newer encoders had a 1.646 Mhz comb (or maybe 1/2 of
- We started putting filters in the communications lines to
cut down on the rfi (which looked like it was coming out of
the shielded cables running to the vertex shelter).
- 31jul09: measured combs from the encoders using the 327
- az enc1 had comb of 1.646 Mhz, azenc2 failed. it was
replaced and had a strong comb of .614 Mhz..
- independently of what i wrote on the web page, i think we
put an "Old" encoder in az enc2 and it didn't have a filter...
- 161011: az encoder 1 failed. before replacing it we measured
the rfi in the screen room. This is a "new" encoder.
180605: turn off vertex encoders one at a
On 180605 we did a test to see which encoder
was causing the .8391 Mhz comb. The setup was:
- 430 Mhz rcvr
- telescope sitting at: az=227, za=11 deg
- The mock spectrometer. 8192 channels, 26.67Mhz bw, 1 second
sampling, centered at 432 Mhz
- a 26.67/2 offset was added in the first mixing stage, and
then removed digitally i the mocks to get rid of the junk at
Each encoder has a number of cables that go
from the vertex shelter to the encoder (serial line, power, etc..)
We disconnected the power cable (at the vertex shelter) for each
encoder (one by one).
The plot shows the average spectra
with all of the encoders on (.ps) (.pdf) :
- The data comes from the first 150 seconds of the test (while
all encoders were on).
- a 32 order harmonic fit was done and the used to normalize the
- The red dashed lines are spaced every .8193 Mhz (the normal
comb spacing for the new encoders).
The image shows the dynamic spectra
for the test (.png)
- After normalizing with the fit, the median value was
removed from each spectra (to try and get rid of continuum
sources passing through the beam).
- The dashed white lines show when an encoder was turned on or
off (to within a second or 2).
- The green and blue lines at the bottom are a .8193 (blue) and
1.6386 (green) combs.
- The are positioned at the comb in the data
- The comb does not get weaker when we turn a single encoder on
- We see a .8193 Mhz comb in the data
- The comb does not change as we pull the power cable for a
single encoder in the vertex shelter.
- This means either:
- we really weren't completely shutting off the encoder when
we removed the cable... or
- the signal is coming from more than one encoder at a time ..
- the comb has nothing to do with the encoders,
- take data with the same setup.
- Turn off the vertex shelter (the large lever).
- leave it off for awhile, then back on.. for awhile, then off,
- This will tell if the rfi has anything to do with signals
coming from the vertex shelter (except for maybe what the ups
180601: Encoder comb in 430MhzRcvr
5 min on off position switching was done on 3C138
using the 430 Receiver and the mock spectrometer. The setups was:
- rf center freq 432 Mhz.
- mock setup: 32 Mhz bw, 8192 channels, 50 ms sampling, polA,B.
- Data was taken for 5 minutes on source then 5 minutes off
Lots of rfi was seen in the online spectra. To
investigate the source of the rfi:
- A 5 minute average spectra was computed using the off source
- PolA and polB were added together.
- A 43 order harmonic function was fit to the average bandpass.
- The bandpass was flattened with : (AvgSpc/fitSpc - 1).
- an fft was done on the flatten spectrum to look at the comb
The plots show the results of the
measurements (.ps) (.pdf)
- Page 1 average spectra and fit
- top: average spectrum and fit.
- Black is the 5 minute average spectra. the red line is the
- the display was limited to 421-444 Mhz (the rf
- Middle: The flattened spectrum: (spcAvg/Fit - 1).
- The red dashed lines are spaced at .8193 Mhz.
- Bottom: blowup of avgspc/yfit -1
- You see the low level alignment of with the comb.
- There are other rfi spikes present.
- Page 2: magnitude of : fft(flattened spectrum)
- This gives a clearer view of the comb. It shows the number
of cycles in the 32 Mhz bw.
- Top: .8193 Mhz comb spacing. This hits the stronger cycles.
- bottom: Twice the comb spacing: 1.6885 Mhz.
- this includes the weaker comb elements.
- A .8193 Mhz comb was seen in the 430 Mhz data.
- On 161011 the comb from the new encoders was measured in the
screen room (before and filtering was done).
- the encoders had a .8195 Mhz Comb
- The 430 comb is coming from one of the az, ch, or dome
- The comb is probably also present in the 327 Mhz Rcvr.
- We need to do a test (turning them off one by one to see which
are causing the problem.
161011: encoder rfi measured in screen room
Azimuth encoder 1 failed on 10oct16. Before
replacing the encoder, we checked the rfi coming from the encoder
in the screen room:
The setup was:
show the rfi measured (.ps) (.pdf):
- Anritsu spectrum analyzer
- avg 5 sweeps
- 20 Mhz span, .1 Mhz rbw, cfr 316.2
- .1 MHz span, 1 KHz rbw, cfr 316.2
- The anritsu was not locked to the station clock (need to
check the offset).
- probe: 6cm loop (901b H field probe)
- encoder rfi measured with encoder opened and closed.
- the cabling was:
- encoder->encoder cable -> rs485 to rs232 converter
-> rs232 to usb -> computer usb port
- The probe was close to the cable as it exited the encoder.
- Top: Encoder open, no filter, all cables plugged in
- The blue dashed lines show the .819584 Mhz comb
- the red dashed lines are blown up in the bottom frame
- Bottom: .1 Mhz span, 1 KHz RBW.
- Black: encoder open, no filter, all cables connected
- Green: encoder closed,
no filter, encoder -> rs485/rs232 converter connected but
disconnected after that.
- Blue: encoder closed,
filter installed, same cabling as green trace
- red: encoder
closed, no filter, but cable to rs485/rs232 converter is
- The comb is coming out the shielded cable. We did not see it
around any other encoder opening.
- The strength of the rfi depends on the value being transmitted
across the line.
- We looked at the SSI serial line with a scope.
- the system uses a grey code
- When there were more transitions on a packet, the rfi
- The baud looked to be about 110 Usec (or maybe 1/2 of that).
- When we do rfi tests on the telescope, we should probably
move the telescope (although you can't turn off the encoder
- If the cable from the encoder is not connected to the rs485 to
rs232 converter, there is no rfi.
- The encoder probably doesn't send any data when the cable is
- We did not have to be connected to the computer to see the
- closing the encoder gives 10db of isolation
- putting the filter in gives 10 more db of attenuation
- The comb lands at 316.206 MHz (this includes the anritsu
offset at 326 MHz when unlocked. It's probably less than a few
- The comb spacing is Mhz
- The comb spacing was measured over the 20Mhz.
- This was probably accurate to 100 KHzRbw/24.
- I then divided this number into 316.206 to get the
- I then rounded the fundamental to the closes integer.
- We also found that the cable shield is grounded to the encoder
case (for the cable we were using).
- Need to check if the vertex cable has the shield connected
to the ground on both ends.
10oct06 strengths of birdies from old encoder measured in the
On 10oct06 in the screened room we removed the
case from one of the old encoders. Using the loop antenna (placed on
top of the encoder wires as they leave the case) and the textronix
spectrum analyzer (set to 1 Mhz span) we looked at the frequency
range 300 Mhz to 1.2 Ghz. The procedure was to set the span to 25
Mhz, look for any birdies, then blowup the individual birdies
to 1 Mhz span. The spectrum analyzer was set to average 10 sweeps.
The table below shows the strength o f birdies found. The amplitude
is probably to to about 3db (the baselines were not that flat).
"Old Encoder" birdies. In lab, cover open
The amplitudes have not been corrected for the
response of the small loop antenna used. When searching for
possible birdies caused by the encoders, start at the strongest
values in the table.
- The 614 Khz comb is coming from the az,za encoders.
- The comb from the new encoder is about 1 to 3 khz wide.
- The comb in the new encoder (installed 02aug06) is a lot
stronger than the comb in the old encoders. We can only see a
few elements of the comb in the the older encoders.
- Back in oct 05 elements of the encoder comb were seen (but we
didn't know they came from the encoders). We should verify that
these are really coming from the old encoders by monitoring
these frequencies while turning off all of the encoders.
- The new encoders comb is shifted about +5 Khz from the old
encoder comb (as seen in the lab).
- In the lab, the strongest birdie from the new encoder is at
326.269 (327 band) and at 432.539 (430 band). Using a log
periodic antenna on the platform, the birdie was strongest at
324.408 (this is one of the strongest comb elements seen with
the old encoders back in oct05).
- In the lab, the strongest birdies from the old encoder are at
324.415 (327 band). This is probably the same birdie as the
324.408. In the 430 band the strongest birdie is at 427.640.
This was seen in the "oldEnd,newEnc" data taken at 430 Mhz.
- With the new encoders we've seen the comb at 327 and 430 Mhz
using the telescope. With the old encoders we've seen it at 327
and maybe 430. In the lab the comb goes to > 1000 Mhz.
- The Az dependence of the various comb elements is repeatable
at that frequency. Some comb elements show different azimuth
- The comb gets at least 10 db stronger when the
dome is above 17 degrees.
- at az=270 a peak is at 17.295 (ch at stow)
- the strength of the comb also changes when the dome sits
still and the carriage house is moved from za=2 to za=20
degrees. With the dome at za=17.295 the comb peaks at
ch=stow and ch=10degree. This probably the scattering path
from the ch encoder into the dome (or it could be the platform
tilting and the scattering path from azenc2 into the dome
- We have concentrated on the azimuth encoders, but there is
probably birdies coming from the dome and ch encoders (although
they are the "old" type).