Aeronomy has been seeing large bumps in their bandpass during atmospheric experiments with the transmitter on. The bumps in the dome system occur around 438 MHz and are wide (FWHM>5    MHZ).

    On 23apr09 various tests were done with the transmitter on. The transmitter was configured for dual beam mode. The mock spectrometer was run during the testing to record spectra every 500 useconds.

---

The setup and data taking:

• power profile mode:
  
• Sps used to drive signals to the transmitter.
  
• 10 millisecond ipp
  
• 13 length barker code and 4 usec baud = 52 usec rf pulse
• beam = 311 usecs
• the rcvr was blanked for about 200 usecs after the end of the rf pulse.
• The cal was fired for about 200 usecs starting at 8800 usecs from the start of the ipp.
• Local mode with long ipp:
• Transmitter driven from pulses orginating in  hagen's box.
  
• ipp=128 milliseconds
• beam/rf=.1% = 128 usecs.
• rcvr still blanked for about 200 usecs from the beginning.
• The cal was turned off for these measurements.
  

• 512 channel spectra were recorded every 500 useconds.
• The bandwidth was 172.032/8 = 21.504 MHz
• Data was started on a 1 second tick and left running during the entire set of experiments.
• The sps was not synchronized to the 1 second tick so the 500 usec mock spectra do not align perfectly with the 10 millisecond ipp.
  

• A high dynamic range uncooled amp was used instead of the dewar.
  
• The 422-442 filter was in place after the preamp/dewar.
• Data was brought down using the 260 MHz IF.
  
• The band was centered at 430 MHz.
  

---

Tests performed:

The test results:

1. Power split between dome and ch, 10 millisec ipp power profile mode (.ps) (.pdf):
• Page 1: 1 ipp (10 milliseconds) of spectra.
•  20 spectra of 500 usecs each.
• The vertical scale is linear. No bandpass correction has been performed.
  
• The bottom frame has an offset inserted for display. Beam on occurs at the bottom and time moves upward in 500 Usec steps.
  
• The first trace (black) has been partially blanked by the receiver blanking
• You can see the ionosphere in traces 2-5 (500-2.5m or 75 - 375 km range) at 430 MHz.
• Page 2: Averge 100 ipps (to 1 second).
• the top shows the spectra, the bottom frame has an offset between spectra for display.
• The low purple and black spectra include some of the receiver blanking.
• The gold and grey spectra straddle the cal on .
  
• The ripples in the bandpass do not smear out over the 1 second integration.
  
• The small 430 spike in the later spectra is DC from the spectrometer (it uses complex sampling with DC in the center of the band).
• Page 3: Bandpass correction
• The high voltage was brought down to zero and then 30 secsonds worth of spectra were averaged.
• I probably should have excluded the spectra with rcvr blanking and cal.
• The median value between 423 and 436 MHz was used to normalize the spectra to unity.
• The red curve is a 20 second averge spectra while on load. The spectral difference would be the monoplexor and sky frequency response.
• The black bandpass correction spectra was used to flatten the raw spectra.
  
• Page 4: Averged, bandpass corrected spectra plotted in 2.5 millisec groups.
• The vertical scale is db above Tsys.
• Each frame has 5 spectra (2.5 millisecs) with each .5 millisec spectra color coded.
• The center of the bump has moved out to 439 MHz after the bandpass correction.
• The first 2 frames show the bump at 439 MHz increasing with time.
• the bottom frame contains the cal on and part of the receiver blanking.
• Page 5: The ripple period using the ACF.
• milliseconds 3-8 from the averaged spectra were combined. This is shown in the top frame.
• The autocorrelation of the averaged spectra was computed and plotted in the bottom frame.
  
•  The 21 MHz bandwidth gives about 50 ns time resolution.
• The large spike as 4 usecs comes from the 250 Khz ripple in the spectra.
  
• There is also a large spike at .43 usecs.
  
• This could be the reflection of the noise from the dome down to the transmitter and then back up.
2. All power sent to the CH, 10 millisec ipp power profile mode (.ps) (.pdf):
• The power splitter in the dome was set to send all of the power to the ch.
• Page 1: Avg of 100 ipps
• The top frame shows the average spectra overplotted.
• The bottom frame has been bandpass corrected and normalized to Tsys.
• Page 2: Averged, bandpass corrected spectra plotted in 2.5 millisec groups.
• Each frame has 2.5 millisecond of data (5 spectra) with each .5 millisec color coded.
• There are now 2 bumps: 438 and 440 each about 2 MHz Wide.
• Page 3: The ripple period using the ACF.
• The top frame has the spectra averaged over 3-8 milliseconds.
• The bottom  frame is the acf.
  
• The 4 usec peak is much smaller and has moved to 3.75 usecs.
• The .42 usec peak is probably the 2 2MHz bumps we see on the right.
  
3. Power split dome,ch,  128 millisec ipp and local mode (.ps) (.pdf):
• The long ipp of 128 milliseconds gave plenty of time to see how long it took the bump to reach its maximum.
• Page 1: Averaged, bandpass corrected spectra for 1st 20 millisecs of 128ms ipp.
• Fifteen 128 ms ipps were averaged.
• The spectra for the first 20 milliseconds are plotted with an offset for display.
• Page 2: Power vs time for various subbands
• The power in various subbands was computed and then plotted vs time from Beam on.
• A 2 MHz bandwidth was used for all bands except for 430 where .5 MHz was used (to get the ion line).
• Top: The entire 128 milliseconds. 439 MHz is the largest (black) followed by 434 MHz (red) and green (432MHz).
• Middle/bottom: blowups in time..
• The * are at the 500 usec sampling.
• 428 is relatively flat, 423 starts to show an increase again.
• Page 3: The ripple period using the ACF.
• The ripple period is now a little bit above 4 usecs (about 20 ns).
  
4. RF off, power splitter dome/ch are equal, 128 millisec ipp and local mode (.ps) (.pdf):
• The rf pulses were switched off while the 128 millisec ipp was being used.
  
• Turning off the rf power made not difference to the strength of the bump.
• At 439 MHz the 2 MHz rbw was still 6 db above Tsys.
  
5. RF off, power splitter to ch, 128 millisec ipp and local mode (.ps) (.pdf):
• With the rf off, all the power was sent to the ch. The 439 bumb, an ripple are the same as that measured with the rf on withthe 20 millisecond ipp in test2.
  

---

Summary:

• A bump in the dome spectra was seen around 437 MHz with a fwhm > 5 MHz.
• The bump moves to around 439 MHz when the bandpass is removed.
• Using the spectrum analyzer, the size of the cal did not change so the bump is probably noise and not a gain variation.
• The bump at 439 is 6db above Tsys when averaged over 2 MHz.
• It takes about 10 milliseconds after the beam is turned off for the bump to reach its maxium.
• The bump remains at its maximum once it gets there.
• Turning the rf off has no affect on the bump.
  
• There is a ripple in the bump of 250 Khz (4 usecs). It could be from a reflection in the waveguide.
  
• Dana computed the waveguide velocity at 440 MHz
• Vgrp=230.7 m/usec, Vphase:390. m/usec
• The waveguide run from the xmter to the dome is:
• about 1300 feet tx to slotted waveguide  (hagens manual).
• 110 feet za=0 to za = 15 degrees.
• 430meters tx to top of dome at slotted waveguide.
• round trip time for 430 meters:  Vphase: 2.2 usec, Vgrp: 3.7 usecs.
• adding another 69 feet would make the Vgrp time be about 4 usecs.
• Unfortunately the velocity for a standing wave is probably the phase and not group velocity.
• If the ripple is a reflection in the waveguide the the extra noise is coming up from the transmitter and is not being picked up over the air.
• The frequency variation of the ripple may be from the turnstile polA,B rejection frequency dependence. If so then the turnstile rejection is not centered at 430 MHz.
• An easy way to check if the ripple is coming from the waveguide is to move the dome down to  0 deg za and see if the frequency of the ripple changes.
  
• On 24apr09 the noise bump from the carriage house was looked at. It had noise on both sides of 430 but was stronger below 430 MHz.
• The beam is turned off when the mod anode goes about 5Kv (below?? or above) the cathode. This will then shut off the electrons boiling off the cathode from being injected into the klystron. Hagen in his manual says that there is still 750 K of noise that goes up to the receiver. This is then knocked down to .75 Kelvin by the 30 db isolation of the turnstile.
• Possible failures causing the problem:
• The mode anode is turning off the beam but the the 5KV drifts to a smaller value after about 10 millisecs giving some residual leakage through the tube.
• The turnstile is not giving the 30 db of isolation so the 750 K is getting through to the receiver.
• If turnstile isolation was not working you'd think you'd melt the monoplexor
• The monoplexor is still working since we don't see a large spike at 430 when the beam is on.
  
• So this is probably not the problem.
• Jon mentioned that corona from something in the deck could also cause this. The question is why the long time to get to full strength once the beam if off.
• Actual failure.. The problem was coming from the transmitter. The mode anode was node completely shutting off the  beam  during the off transmission times. It was a problem in one of the diodes.