NRAO Home  >  Green Bank  |  Wiki Topic:    GB > PTCS > InformationForProjectTeam > PTCSProjectPlan > AntennaCharacterizationAndCommissioning > AntennaCharacterization02Nov2007

Changes | Index | Search | Statistics | Go

Meeting to discuss tracking performance and future PTCS observing plans in this area

• Date: 02 Nov 2007, location: CV311/GB241
• Attendees:
  • CV: Todd, Fred, Jim, Paul, Brian
  • GB: Kevin, Frank
• Minutes
  1. Todd introduced Paul Ries to the other attendees
  2. Paul discussed what he has been doing since joining the PTCS project this Fall
     • Examined 10-20 maintenance days when the telescope was parked. Took FFT of 10 second periods of accelerometer data to find resonance peaks, compared them to average windspeed, and made histogram and percentile analysis.
     • He finds that the amplitude in the frequency domain scales with velocity squared, as expected.
     • Used the formalism in Greve & Mangum 2007 to convert frequency domain amplitude to millimeters of sinusoidal motion. Plots attached below.
     • In addition to the main resonances, Paul sees a 1.4 Hz peak on the elevation axis accelerometers which goes away at higher wind speeds.
     • Kevin then asked about the uniformity of sampling of accelerometer data, as he has found non-uniformity in the servo datastreams (sometimes 47 Hz, sometimes 49 Hz, etc.) as well as decimation without proper Nyquist filters which will cause aliasing of higher frequencies into the band.
     • Todd noted that the accelerometers have internal timestamps which should be used instead of the timestamp from log collator. See the attached email from Joe about the "basis" option in log_collator.
     • *ACTION ITEMS:* Paul will go back and repeat his analysis using the internal timestamps with a non-uniform FFT in two ways:
       • using the raw accelerometer timestamp values in the FITS files
       • using the acceleroemter timestamp columns interpolated by log_collator
  3. Todd presented results from two half-power tracks from September 2007
     • one at low winds (1m/s) scan 3617 which shows a tracking error rms of 1.5"
     • one at high winds (5m/s) scan 4889 which shows errors up to +/- 8 arcsec
     • Todd noted that some of the wandering could be due to drift in receiver gain. Jim suggested we could get an idea of this by examining data when off source. Todd noted that each half-power scan was preceeded by a 30 second scan off source.
     • *ACTION ITEM:* Todd will compute the effective rms error on the off-source position for these two half-power tracks. Completed. The rms values are 0.06" for Scan 4889 and 0.2" for Scan 3617.
     • Jim noted that the total tracking error will be a quadratic sum of a zero speed error and a wind-velocity-squared-dependent error
     • Jim noted that the value at low winds appears to be better than before (used to be 1" rms in two-dimensions with 2.1m/s wind), which may be due to the new pointing model and flatter track. The value at high winds is similar to before, as expected since the improved track would not have affected performance in these conditions.
     • We should see a degradation of the wind-dependent component that scales upward as wind velocity squared.
     • *ACTION ITEM:* Todd will locate a low-wind example of a half-power track with the old azimuth track for direct comparison.
     • *ACTION ITEM:* Todd or Frank will process a few more half-power tracks and plot the tracking error vs wind speed on the new track.
  4. Todd asked Frank to summarize the results from the first Ka/CCB pointing run
     • Used primary calibrators, took jack scans and Nods. Determined counts/Jy in each of the 4 IF bands. Plotted a gain curve which is nominally flat above 10 deg elevation, but shows peak to peak variation of 20%.
     • *ACTION ITEM:* Frank will write up these results on a wikipage. Completed. It can be found here.
     • Frank talked to Ron Maddalena about Nods and he suggested doing subreflector nods instead. Brian warned that he tried this and sees a large synchronous level change between the nod positions which results in a 30mK square wave in the data. Jim asked whether we could try to minimize this imbalance by adjusting the center position of the nods. Since subreflector nodding is symmetric about the optical axis, as the primary dish tracks the position midway between the two receiver feeds (see Section 7 of MR 2C407), the shift in illumination pattern should be close to the same in the two positions. We agreed that this topic should be discussed in more detail offline.

• • • Brian asked what was the range of wind speeds during this run? Frank was unsure.
    • *ACTION ITEM:* Todd will compute the wind statistics for this run. Completed. See results on TPTCSPNT_071007 and TPTCSPNT_071021
  1. Ideas for next PTCS observing runs in November & December
     • Jim notes that future pointing runs and half-power tracks should use the gold-standard calibrators in the PCAL catalog, as described in Project Note 58.
     • *ACTION ITEM:* Frank will either extract a subset catalog, or change the current script to search for the Gold flag column.
     • Todd noted that the Quadrant Detector is an important resource that has not yet been utilized well. He realized recently that no data were recorded for it in /home/gbtlogs between April 16, 2007 and October 26, 2007 when Frank and Joe reactivated it at 15:32 UT.
     • Todd reminded the group of the correlation he found between the QD data and the total power oscillation seen in a focus scan (see plots under item 3 of the wikipage on FeedArmVibration). Kevin expressed interest in these data.
     • *ACTION ITEM:* Todd will email Kevin the date/time of interest for the correlated vibrations seen in the QD and CCB data. Completed.
     • Todd asked what steps are needed to put us on the track of calibrating and eventually using the QD to correct Mustang data and other high-frequency data.
     • Brian suggested doing scans (daisies, circles or nods) to excite the motion and compare to receiver data. For example, he noted that the current Nod procedure has a pre-smoothed trajectory to try to minimize excitation, so that could be modified.
     • Brian said that it will be quite useful to have someone looking at QD data regularly, and that if PTCS can sensible results and get coefficients, he could read the QD data into his Mustang processing pipeline.
     • Kevin noted that this sounds like a classic System Identification problem. Todd agreed and suggested that Kevin can probably help us out a lot in this effort.
     • After the meeting, Brian sent this link which shows the effects of servo error during on-the-fly Nod scans. He also added this description, as well as some data from a more recent observing run on TKA_09OCT07 (the four plots attached below). Todd added a plot showing servo following error during scan 170 and during scan 70. These plots confirm that the servo following error is much worse on the beam switch transition in scan 170.
     • After reading Brian's email, Todd remembered his finding from last year that the wiggle in Nod scans correlates with wiggles in the inclinometer readings. Also, the wiggle is consistent with azimuth velocity reversal. That is, it happens after the first beam switch when the sources is in the south (tracking with positive azimuth velocity), while it happens after the second beam switch when the source is in the north (tracking with negative azimuth velocity). See this wikipage for more details.
     • This finding underscores the importance of activating friction compensation ASAP before Brian's next Ka/CCB observing run, because it should help the GBT performance during velocity change. For example, see this plot of servo following error obtained during a Nod scan vs. friction compensation gain. Todd was unsure whether friction compensation had been turned on yet although everyone involved (Tim, John, Joe, Todd) had decided on October 10th that it should be done on the "next" maintenance day.
     • *ACTION ITEM:* Todd will ask Joe to enable friction compensation at the next opportunity.
  2. Todd asked Kevin to tell the group his work plans, as not everyone present may be aware of them
     • short-term: track down and eliminate the servo resonance, as described in this project document
     • long-term: lay groundwork for implementing a model-based servo (LQG or H-infinity) for when we upgrade the servo control boards from analog to digital
  3. Next steps for Paul, beyond the "GBT at rest" analysis
     • In addition to speed, wind effects will be a function of elevation, and the angle between azimuth and wind direction.
     • Paul notes that on days with moderate to high winds, that there are periods of low winds as well, so picking out these may be the most efficient to analyze.
     • Frank notes that there is an old program that provides a quick way to get maximum wind for a whole hour.
     • *ACTION ITEM:* Frank will dig out an old program and send it to Paul. Completed. A description is here: Weather summarizer
• Further details on Brian's scans attached below
  • wind was less than 1.5 m/s
  • In both cases, the source is in the south and tracking with positive azimuth, thus Todd would expect to see the wiggle after the first beamswitch but not the second.
  • TKA_09OCT07 scan 70 = 2007_10_09_06:31:16.fits, az=173.5, Elev=43.2, source 0241-0815, does not show the standard wiggle, but does show more overall wander than scan 170
  • TKA_09OCT07 scan 170 = 2007_10_09_08:58:39.fits, az=219.8, elev=34.6, source 0241-0815, shows the standard wiggle after the first beamswitch

-- ToddHunter - 03 Nov 2007

Attachment:	Action:	Size:	Date:	Who:	Comment:
tka09oct07-s70j9.png	action	9401	03 Nov 2007 - 16:09	ToddHunter	scan 70, 43 deg elevation, 38 GHz channel
tka09oct07-s70j12.png	action	8998	03 Nov 2007 - 16:09	ToddHunter	scan 70, 43 deg elevation, 28 GHz channel
tka09oct07-s170j9.png	action	9233	03 Nov 2007 - 16:09	ToddHunter	scan 170, 35 deg elevation, 38 GHz channel
tka09oct07-s170j12.png	action	9184	03 Nov 2007 - 16:09	ToddHunter	scan 170, 35 deg elevation, 28 GHz channel
1102graphs.pdf	action	91922	04 Nov 2007 - 22:33	ToddHunter	Paul's plots of accelerometer displacement vs wind
brian.text	action	1022	04 Nov 2007 - 23:20	ToddHunter	Brian's description of CCB Nod results
scan170.pdf	action	50701	05 Nov 2007 - 14:48	ToddHunter	Plot of servo following error during scan 170
scan70.pdf	action	49774	05 Nov 2007 - 14:57	ToddHunter	Plot of servo following error during scan 70
joe.email	action	747	05 Nov 2007 - 20:21	ToddHunter	Joe's email about "basis" feature of log_collator
srnod.email	action	5860	07 Nov 2007 - 02:10	ToddHunter	Brian & Ron M. email thread on SR nod spillover
SRNod.gif	action	4605	07 Nov 2007 - 02:12	ToddHunter	SR nod figure created by Ron in email thread above