lband drifting birdies with 60 Mhz spacing



11jun04: First look at drifting birdie. Harmonics, drift, az dependance, time dependance.
28oct04: measure the drifting birdies at lband and the 430Mhz.
04nov04: The drifting harmonics do not come from the dome airconditioners.
12nov04: switching between sky and load at 430Mhz.
18nov04: 430 data sampled at 1 millisec shows signal jumps very 2.67 seconds.
19nov04: The birdie is coming from the dewar monitoring system.
How the dewar monitoring system works.
18nov04: Summary of what we know
14dec04: birdie seen at 60 Mhz with spectrum analyzer in the dome
17jan05: The 60Mhz birdie, what it was and how we fixed it..

11jun04 First look at drifting birdie. Harmonics, drift, az dependance, time dependance.  (top)

  A birdie that drifted through lband appeared in the middle of may04 (project A1861). There were copies of this birdie spaced by about 60 Mhz. Its drift rate varied from stable to 20 Mhz in 5 minutes. The birdie was not always present.

Data taking setup:

    On 11jun04 data was taken during the day to investigate this birdie. The dome was set to 18 degrees za and the azimuth was swung (at .04 deg/sec) from az=450 to az=90 and then back to az=450. This was repeated twice. Data was taken with the interim correlator. The setup was 4 sbc, 50 Mhz each with 2048 channels.  After hanning smoothing, the resolution was 49 Khz. The sbc were centered at:  1335, 1377.5, 1422.5 and 1465 Mhz. The data was dumped once a second. Two data sets were collected each lasting for 2000 seconds. Each of these covered the azimuth swinging from 450 to 90 and then back to 450 degrees.

Data processing:

    The data processing consisted of :
  1. input 2000 records
  2. Compute a bandpass correction for 500 records at a time. Use the rms by channel on each set of 500 records to throw out data points that had variations in time. Divide by the bandpass correction and subtract 1 (so the units are now Tsys).
  3. For each 50 Mhz, 2000 records display the dynamic spectra. With the cursor, trace each drifting birdie storing the time and channel number.
  4. Given the above time, channel numbers, interpolate the position of the birdie for the entire 2000 records. Shift all spectra to line up the birdies in frequency. Keep 40 channels (+/- 1 Mhz) about the birdie. After shifting, pick a strong record and cross correlate the 2000 samples with this. Shift each of the samples by this amount.
  5. After alignment, compute the average power of 26 channels about each 1 second sample. Remove a baseline that is the median value for the 2000 records of the (81-26) channels that do not include the birdie. Use this data set for the power vs time/azimuth plots.


Spectral density for 11 seconds.

    The spectral density for 11 seconds of data shows the drifting birdies. The y axis is in units of Tsys. The different strips have been offset for plotting purposes. The 3 windows are centered on different multiples of 60 Mhz. The birdie gets up to about 8% of Tsys and looks dual humped. You can see it drifting to lower frequency with time.

The frequency variation of the  birdies with time and azimuth

    The plots: (Freq. vs time and azimuth Data set 1)  (Freq. vs time and azimuth data set 2) show how the frequency of the birdie varied for the two data sets. The data does not cover the entire 2000 seconds because the birdie was not  visible for the entire time period.

The strength of the signal versus time and azimuth.

    The plots (strength of signal Data set 1) (strength of signal Data Set 2) show the variation of the strength of the signal with time and azimuth. The data was shifted in frequency to align it and then averaged over 26 channels that contained signal power.

Dynamic spectra of the aligned data sets.

    The spectra for the aligned data sets are shown below. 81 channels (2 Mhz) is displayed. Birdies that sweep across this image were constant in frequency before the alignment of the drifting birdies. The contrast of the image improves if you are not using too many windows (or print it out on the xp0 laser printer).
data set 1
data set 2
Harmonic 22 1338 Mhz
Harmonic 22 1338 Mhz
Harmonic 23 1398 Mhz
Harmonic 23 1398 Mhz
Harmonic 24 1450 Mhz
Harmonic 24 1450 Mhz
Harmonic 25 1520 Mhz
    In data set 1 harmonic 22, the birdie remains on for about 5 minutes. The birdie look like it has a modulation envelope is weak, increases to a maximum and then decreases again. This is probably caused by the azimuth arm sweeping through a strong signal. In the other images this is not so obvious because the signal is turning on and then off before the sweep can complete.


processing: x101/040611/,

28oct04: Measure the drifting birdies and lband and then 430 Mhz.  (top)

    On 28oct04 the drifting birdies were measured using the lband wide receiver. The dome receiver was switched (within 1 minute) to the 430 Mhz system and the drifting birdies were seen at 430 Mhz. When we shifted back to lbw, the birdies and disappeared. The plots show the birdies at lbw and 430 Mhz (.ps)  (.pdf). Using the difference between the lband harmonics, you can predict the frequency of the drifting signal at 430Mhz.  The signal is about 20db stronger at 430 Mhz than lband. The problem is that the 430Mhz receiver has a filter of about 20Mhz. Occasionally you will see the birdie in the lbw system and not in the 430Mhz system. That is because the birdie at 430 can fall outside the 20Mhz bandwidth of the receiver system. When searching for the drifting birdie with a spectrum analyzer/antenna you should use 430 Mhz since it is so much stronger. You will still probably need to use the lbw system to identify the drifting nature of the birdie.
processing: x101/040128/

04nov04 The drifting birdie is not from the dome AC units.   (top)

    While the lband 60 Mhz harmonic birdie was drifting, the dome air conditioning units were both turned off. The birdie remained. So the air conditioners can be ruled out as a source of the birdie. The image shows a plot of the drifting birdie while the dome AC's were off (.gif).
    processing: x101/041104/

12nov04: look at birdies with 430 rcvr on and off load.  (top)

    On 12nov04 the birdies where seen in the 430Mhz receiver. While monitoring the birdies, the receiver was switched from sky to load and then back to the sky. Tsys on the sky is about 60 K while Tsys on the load is probably 330K.
    The image shows the dynamic spectra when switching between sky and load (.gif):     The birdie went away when switched to load. This could happen if the birdie was coming in through horn, or the system temperature on load was too high to see the birdie.

    The plot shows the spectra as second number 20 (.ps)  (.pdf) while on the sky. The signal is 25 Kelvins. The 1 second integration and 25Mhz over 1024 channels gives deltaT/T=.0064 . For a Tsys of 330 K, deltaT is .0064*330=2.1 Kelvins, so a 25 Kelvin signal is easily detectable.

    Since the birdie went away while on load, the birdie is coming in through the horn. This is also supported by the harmonics at 430 and lbw. If the birdies were in the IF then you couldn't use the harmonic numbers to predict the 430 frequency using the lb spacing.


18nov04:  430 data sampled at 1 millisec shows signal jumps very 2.67 seconds.  (top)

    On 18nov04 the birdie was visible in the correlator using 1 second dumps. The setup was switched to run the wapps in pulsar mode with 25 Mhz, 512 lags, and 1.024 milliseconds per dump. 300 seconds of data was taken. The images show the dynamic spectra for different smoothing of the integration times:     The 2.67 second spacing of the jumps was the key to figuring out where the birdie was coming from. The dewar monitoring system (the system that reads the dewar temperatures, voltages, currents... see below for a description.) takes 2.67 seconds to cycle through a single dewar. After 2.67 seconds, the same "address (temp,current, etc)" is being reread (but now from a different dewar).
    The frequency jump is 4.39 Mhz at 430. This would be a jump of .628 Mhz at the 60Mhz fundamental. This jump is not seen in the lband data because it is too weak. The duty cycle of this "upper freq" is only 5%. If fast dump data was taken at lband, it would be seen as a .628*24=15.1 Mhz jump.
    (note: the 19nov04 data showed that the amount of this jump varies).
    processing: x101/041118/

19nov04: The birdie is coming from the dewar monitoring system.  (top)

    The fast dump data from 18nov04 showed that the birdie was jumping every 2.67 seconds. This is the same time period that the dewar monitoring system needs to cycle through 1 receiver. On 19nov04 data was taken with the lband receiver. When the birdie appeared, the dewar monitoring program was stopped. The birdie was still there!!
    I then moved to the 430 Mhz receiver. The signal is 20db stronger than lband and you can see the frequency jumping every 2.67 seconds. Data was taken with 1 seconds dumps, 25 Mhz bandwidth and 1024 channels.  The dewar monitoring program was stopped for about 50 seconds. The image shows the dynamic spectra when the dewar monitoring program was stopped.     The plot shows that the birdie is coming from the dewar monitoring system. During 140 milliseconds of the 2.67 dewar monitor cycle, something is being done that causes the birdie to jump up in frequency. This could be a particular mux address being accessed or maybe switching between dewars when the match of some circuit is changing.
    Turning off the dewar monitor program does not make the birdie go away. It just stops it from jumping.  When stopped, the dewar monitor program is no longer cycling through the event that causes the jump. At lband you see no change when the program is stoppped because the jumped to higher frequency is too weak to ever see (because lband is 20db weaker than 430, and the higher frequency has a 5% duty cycle).
    processing: x101/041119/

The dewar monitoring system.  (top)

    The dewar monitoring system measures the dewar temperatures, bias voltages, currents, etc... It is installed on 9 of the receivers. For each dewar there are about 28 analog readings to be made. The system conists of     The digital i/o lines (mux/rcv addresses) go from the hp34970 to the muxbox. From the muxbox they fan out to the 9 receviers. The two bits that are used for the temperature diode select (16K,70K,omt) go first to the postamp chassis and then to the bias box. The other 3 bits go straight to the bias box. These i/o lines are common to all recievers. They are not buffered in the muxbox.
    To read a receiver, the hp34920 first selects the address of a receiver. This controls the muliplexor in the muxbox. It then cycles through the 28 addresses for the data to read. All 9 receivers switch together and send their analog signal to the muxbox. The muxbox then passes the analog signal from the selected receiver back to the hp34970 a/d converter.
    It takes about 2.67 seconds to cycle through all of the addresses of a single receiver. It takes about 24 seconds to read all of the receivers and start over.

Summary of what we know:  (top)

From 11jun04 lband measurements. From 28oct04 measurements. General:


    19Nov04: The birdies are coming from the dewar monitoring system (writeup to come).

14dec04: birdie seen with spectrum analyzer at 62.457 Mhz in the dome.  (top)

    On 14dec04 the birdie was being monitored with lband. It appeared at 15:52 (same time as previous day!). The textronix ybt250 spectrum analyzer was used to search for the birdie in the dome with a small whip antenna. The birdie was seen at 62.457 Mhz using the spectrum analyzer. It was up to 30 db above the noise floor using a10 Mhz bandwidth.  The image shows a screen dump of a dynamic spectrum of the birdie.

17jan05: The 60Mhz birdie, what it was and how we fixed it.. (top)

    The 60 Mhz biridie was rfi that produced harmonics close to 60 Mhz. They were seen in the lbw, alfa, and 430 Mhz receivers (the 327 receiver frequency range did not include 300 or 360 Mhz so it was not seen there).  These birdies were first noticed around may04. They were time variable. The interference would show up for awhile and then disappear only
to reappear later. Typical times would be 10 to 60 minutes when it would be always present. Another characteristic was that the birdies would drift in frequency. They could drift by 20 Mhz in 5 minutes and then settle down and remain constant. The  birdie drift  and the 60 Mhz spacing were used to identify the birdie as belonging to the 60 Mhz harmonics.

    High time resolution data taken while the birdies was present showed that the birdies  jumped by a few Mhz every 2.7 seconds. The jump would last for about 140 milliseconds and then it would return to the old frequency. 2.7 seconds is the time that it takes the dewar monitoring system to cycle through all the readings of a single receiver. Further tests showed that if the dewar monitor system was not cycling through the receivers, then the birdie remained but this jumping in frequency stopped. This pointed to the dewar monitoring system as the culprit.
    The dewar monitoring system is built into the post amp chassis's and the bias boxes for each receiver. The cables that control the receivers all run in parallel from the control device out to the post amp chassis and bias boxes. Using a portable spectrum analyzer and a small antenna we were able to see the birdie in the dome at 60 Mhz (although it was still variable).  After awhile we noticed that changing the cal configuration (correlated cal, uncorrelated cal, crossed cal..) could make the birdies start or stop (although not all the time).  This cal configuration is also in the post amp chassis. Eventually we narrowed the problem down to the sband wide (sbw) post amp chassis.

    The cal configuration control was using some high speed cmos chips. Some of the gates on the chips were not used and they had been left floating. Some of the configurations of the cals setups would also leave some high speed gates floating. It turns out that the floating gates were oscillating. The oscillations were many volts. Putting the antenna close to these chips showed oscillations 50+db above the noise floor ( the noise floor was probably a few 1000 K) and they continued through the 0 to 200 Mhz range.

    The solution was to ground all of the unused gates on the high speed chips (a better solution would be to remove all the
high speed chips since they are not really needed). Resistors were also added so that none of the cal configurations allowed any of the high speed gates to float. These changes on sbw got rid of the 60 Mhz birdie.

    When the chips were oscillating, they would couple to all of the lines leaving the sbw post amp chassis. The dewar monitoring cables would then fan out to all receivers and act as radiating antennas. The 140 millisecond long frequency jump every 2.7 seconds was probably caused when one of the addresses in the dewar monitoring read back was selected. This address must have been affecting the oscillating chips and caused the oscillation to jump in frequently

    Sbw was causing the oscillations, but all of the receivers have the same hardware circuitry (although the wiring for sbw looks  a bit of a mess!!). We are in the process of making the changes to all of the high speed chips in all of the post amp chassis so that this problem does not crop up in another receiver.