Leuschner_visit_on_5_25 - david-macmahon/wiki_convert_test GitHub Wiki

Executive Summary

  • Testing at Leuschner Observatory on May 25 showed that our 1 to 2 GHz passband had excessive RFI, resulting in amplifiers that were overdriven.
  • We installed bandpass filters to restrict our band to 1400 to 1630 MHz, rejecting the very high power RFI at both the lower and upper parts of the band.
    • Now we have a 260 MHz bandwidth.
  • We collected an hour of PSR0329 data and team members are looking at it now.

Details

Which Pol is which nomenclature

Our tests show:

PolA = Pol1 = ADC input "I": plotted on Left Side, Blue color on ADC display

PolB = Pol2 = ADC input "Q": plotted on Right Side, Red color on ADC display

Birdie test

We injected a birdie at 1800.000 MHz using the 2nd synthesizer from the Valon, connected to a wire hanging out the door (to act as a transmitter). The dish (for all our measurements that day) was pointed north, near the horizon, in "maintenance mode". The data with the birdie is in a file called FreqTest_may_25_2017.fil. Cursory visual inspection showed the birdie in the right place in the ADC plots and in cursory plots of the file. but it would be good to look at the exact bins in the filter bank file and make sure it's exactly correct, not off by one...

Linearity tests

We made measurements using the Keysight Portable Spectrum Analyzer at various stages in the analog signal chain. We saved many of these spectra on the analyzer (See Spectra recorded on the visit).

We found that the passband was not well defined (there was no noticeable bump from 1 to 2 GHz - the band from 0 to 3 GHz was flat). (we saw this before as well). We found excessive high power levels going into the amplifier just before the fiber driver. We found that one of the polarizations (POL2) had a 10 dB attenuator in the signal path in the focus cabin and the other pol didn't have the attenuator. We removed this 10 dB attenuator. We also removed the 6 dB attenuator we had installed in POL1 just before the snap ADC, So the two bands are closer together in power.

New bandpass filters

We removed the Minicircuits high and low pass filters in both pols, and replaced them with a 1515/230 bandpass filter (previously used by Dave for the RPA, the prototype for the ATA).

Center frequency: 1515 MHz

Bandwidth: 230 MHz (1400 to 1630 MHz)

The passband is now well defined (see spectrum analyzer saved plots for both pols), and sticks out well above the noise. When we removed the 10 dB attenuator (see linearity tests section), the passband went up 10 dB more above the noise (linear!). It is now about 25 to 30 dB above the noise floor (0 to 1400, and 1400 to 3000 MHz).

Action items

  1. Confirm birdie is in exactly the right bin in the filterbank file (see birdie section, above).
  2. Put spectra we recorded at Leuschner with the spectrum analyzer, and this email, on our wiki
  3. Can we find 0329? (if not, see action item 8)
  4. Do another RFI sweep in azimuth and elevation?
  5. Develop code that only outputs band of interest to the file (1400 to 1630 MHz). Then we can shorten integration time by a factor of four.
  6. Add code to measure RMS values on the two ADC's. Do we have enough signal power going into the ADC's? Do we need to add more gain?
  7. Equalize the two pols in software so we can optionally add the two pols together (this also allows us to shorten integration time).
  8. Check integration times - do we know this precisely? (this is important for pulsar detection). Check Simulink design and sample clock rate- to check this, measure how long it takes to write 10,000 spectra. Are we using the correct integration time parameter when we fold the pulses from PSR0329 observations?