Hawkesbury Radio Astronomy Observatory

Vela Pulsar Observations

Observing PSR B0833-45 with a Small Aperture Antenna


Observation Information

A selection of sub-panel images from the 'Virtual Control Panel' (VCP) is uploaded to the 'Daily Observations' webpage during, and after, a daily observation.

Note: The images here are representative examples only. The latest software version display may differ in detail.

HawkRAO Observatory Status Panel

This message box signals the current status of the Vela observations (updated every 5 seconds). Status modes include:

HawkRAO Period Search Panel

This panel shows the result of the last observation.

The 'Best S/N Scan' graph in the top-right of the panel is a plot of S/N (Sigma) versus fold period generated during the search for the observed pulse period during Epoch Folding.  There are three traces of the calculated S/N for each period search step - a raw trace (thin white), a moving average of 5% scan width (thick white) and a moving average of 10% scan width (medium grey).  To indicate S/N the thick white 5% moving average trace remains white below the 4σ level, turning green between 4σ and 6σ, and purple above 6σ. The adopted 4σ and 6σ levels are minimum levels of significance for citizen scientist and professional results respectively.  That is, only observations which contain at least some green in the moving average trace are statistically significant enough to be regarded as a positive detection of a pulse.

The 'Period Search' panel is refreshed every 15 minutes (although the data only changes once a day after an observation).

The Effects of RFI on Period Searches

Some attempt has been made to mitigate against impulsive RFI by simple blanking of the power data (square-law detected) at values greater than the 3σ level. In addition a harmonic filter is applied to the detected power data. At the HawkRAO location there is interference which has fluctuations near to the spin frequency of Vela.  Due to the narrow pulse width of the Vela pulse most of the pulse energy is contained in higher order harmonics.  It has been found that rejecting the fundamental and second harmonics of the Vela signal by applying a high-pass filter (Fc = ~23 Hz) the Vela pulse energy is largely passed whilst almost completely eliminating the interference - improving both the S/N and the accuracy of the 'Best S/N Scan' algorithm accuracy.  To further improve S/N a harmonic filter is applied where only a small band of frequencies (width ~0.1 Hz) around each of the harmonics numbered 3 to 32 are passed.

Transit Plot

As an indicator of strength of the Vela pulsar signal as it passes through the beamwidth of the fixed-pointing antenna during the observation period, a transit plot is produced with pulse phase on the vertical axis (0 - 360° from top to bottom) and observation time on the horizontal axis (nominally 0 - 120 minutes left to right for the current observations).  The resolution in the vertical scale (pulse phase) is the number of time bins in the profile (100 for the current observations), while the horizontal resolution is in minutes - creating a 120w x 100h pixel image.

Scan Quality

Each day's observation result is assigned a 'Scan Quality' [SQ] '+/-' rating based on the nature of the 'best S/N period' scan plot which sorts out the good scans from the poor scans depending on how narrow and symmetrical the curve shape is (narrow shape/symmetrical = good: broad flat shape\asymmetrical = poor) .  Typically poor scans are caused either by excessive levels of RFI...

...or simply a low average signal level from Vela over the observation time span.

Example 'Good Period Scan' [SQ+] Example 'Poor Period Scan' [SQ-]

Those observations with a [SQ-] result are excluded from the 'Glitch Monitor' plot (and associated statistics).  The SQ process detects and excludes excessively corrupted data, which, at the time of writing, is about 15% of the observations.

HawkRAO Glitch Monitor Panel

Each daily observed topocentric spin frequency (observed period-1) is entered into a database and compared with the predicted topocentric spin frequency and a spin frequency offset calculated (ppm).  A mean and standard deviation for all spin frequency offsets observed to date is calculated and 5σ and 6σ levels determined.

Note that the sigma (σ) levels adopted for the Glitch Monitor are different to the Pulse Profile sigma levels. This is because the Glitch Monitor drives an alert function and so requires a higher certainty.

The 'Glitch Monitor' panel result is refreshed every 15 minutes (although the data only changes once a day after an observation).

The horizontal green line is the 5σ level for glitch detection and the horizontal purple line is the 6σ level derived from the spin offsets observed.

Each day's spin frequency offset is plotted as blue dots (·). In the example to the right the offsets are mostly less than ±0.5 ppm from the predicted spin frequency with a small number of outliers. The green 5σ and purple 6σ lines are the levels of spin frequency offsets (in terms of ppm) used for determining whether a spin offset measurement signals a glitch event or merely an expected variation due to noise. That is, only spin frequency offsets ('glitches') above the 5σ and 6σ levels are statistically significant enough to be regarded as a positive detection of a glitch (see table below).

The Standard Deviation (SFO σ ) and the Mean (SFO μ) of the spin frequency offsets are shown in the bottom-left blue box.

The values for the 6σ and 5σ glitch alert levels are shown in purple and green respectively at the top-left.

The daily S/N (dB) is also plotted as yellow dots (·).  The Mean S/N (S/N μ) is shown in the bottom-right yellow box.

It is useful to be aware of the relevance of the 5σ and 6σ glitch alert levels in terms of the probability of any one spin offset measurement being a 'real' glitch and not just an expected measurement variation.  This can be examined in terms of the probability of a result being a 'defective' in Six Sigma parlance. Short term sigma (σ ) is used here as there are multiple measurements, but a single measurement setup.

Defective Rate vs Sigma Levels
Sigma Level (σ) Chance of a 'Defective'
1 1:1.5
2 1:3.2
3 1:15
4 1:160
4.5 1:741
5 1:4,292
5.5 1:31,250
6 1:294,118

Starting from the 3σ level we would expect one result to be outside the 3σ level every 15 observations.  Obviously not very useful for sorting the 'real' results from the expected 'noise' results.

For the 4σ level the expected 'noise' results has dropped to one result in 160 observations.  For the current number of observations (>140) it would be expected that there might be 1 'noise' result in the observations. Observations are done daily, and so a 'noise' result can be expected once every 23 weeks. As the observations are going to be conducted over at least several years, this level of significance is not sufficient.

For the 4.5σ level the expected 'noise' results further drops to one result in 741 observations and so a 'noise' result can be expected once every 2 years. As the observations are going to be conducted over at least several years, this level of significance is marginally sufficient.

For the 5σ level the defect rate drops to one 'noise' result in 4,292. For a daily observation a 'noise' result can be expected once every 11.75 years.

For the 6σ level the defect rate drops to one 'noise' result in 294,118. and a 'noise' result can be expected once every 805 years.

For the designed purpose of the 'Glitch Monitor' panel an alert threshold level of at least 5σ would seem appropriate given the time between Vela glitches is about 3 years.

NOTE: The typical mean of the spin frequency offsets in the above example is (SFO μ) is less  ±0.1 ppm, showing that the predicted and observed periods are in good agreement. However, as time goes on, this agreement may drift as the predicted period is derived from an ephemeris generated at the beginning of this year. This is because the Vela pulsar 'glitches' regularly (with major glitches every 2 or 3 years), but also 'micro-glitches' more frequently than that.  This will show up as the spin frequency dots clustering around a line which either slopes up (where the ephemeris prediction is lower than actual) or slopes down (where the ephemeris prediction is higher than actual). Caution must be taken that a drift in the sampling clock frequency is not the cause.  Currently the clock has 0.5 ppm TCXO stability and so can be expected to remain within ±0.2 ppm.  This uncertainty can be removed by replacing the TCXO with a Rubidium Frequency Source - which will be done in the future.

When a glitch occurs, Vela retains a varying small proportion of the spin-up jump after a period of days to months. Therefore, although Vela 'spins down' like a normal pulsar does, there are periodic residual 'spin-ups' which reduce the spin-down rate from the value calculated from observations spanning a limited number of days (e.g., 30 days).  The plot to the right shows the cumulative spin-ups over nearly 30 years.  The slope of the line is ~1.91x10-9/day.  This relationship could be applied to ephemerae produced for epochs more than a year or so in the past to make a rough correction.

More generally it needs to be recognised for young glitching pulsars (like Vela) the instantaneous rate of spin-down can vary from the ephemeris prediction depending on where in the inter-glitch time period the ephemeris was derived. For example, for Vela, the following graphic shows the spin frequency second derivative df/dt () variation over nearly 40 years...

The upward sloping line shows a slow decrease in the spin down rate - a consequence of of the decreasing rate of angular momentum loss as the pulsar spins slower.  To paraphrase, the slower the pulsar spins, the slower it gets slower.  If the ephemeris was generated at Epoch A (just after a glitch) the value of the spin frequency second derivative would be -1570 x 10-15 Hz s-1, while if it was generated at Epoch B (just before a glitch) it would be -1559 x 10-15 Hz s-1.  The difference of 11 x 10-15 Hz s-1 may not seem much, but it is Hz s-1, so after 1000 days the difference would be 1000 x 24 x 3600 x 11 x 10-15 Hz = 0.085 ppm.  While this is just barely enough to be significant at the current resolution of spin frequency measurement, in the future it will become more significant when higher S/Ns are obtained by an increase in antenna aperture.  In any case it it wise to be generally aware of differences in second derivative spin frequency values between ephemerae from different epochs.

Initially the X-axis is set to cover from the beginning of 2017 to the beginning of 2018 (MJD 57800 to 58200 = 400 days).  It is hoped that Vela will 'glitch' within that time frame.

Vela Information Panel

This panel shows the current Vela Pulsar position in terms of UTC time of next transit and time until next transit, as well as azimuth and elevation for the HawkRAO location.

The azimuth and elevation field texts are in green if Vela is above the horizon at the HawkRAO location and additionally in brackets (...) if Vela lies within the antenna beamwidth.  The text is red when Vela is below the horizon.

 The current Modified Julian Day is shown.

The remaining time until the next transit of Vela at HawkRAO is displayed as well as the next transit time.

NOTE: The start of data acquisition is earlier than transit and begins when Vela enters the beamwidth of the antenna.

This panel is updated every minute and is accurate to ±1 minute.

Aggregate Pulse Profile Panel

This panel shows the summation of many profile results - indicated by the '∑ nnn Profiles' annotation.  It should be similar to the profile results of professional observatories (who can use bigger antennas and get good profiles in one observation).

The current aggregate pulse profile shows the 'scattering tail' due to the ISM for Vela @ 436 Mhz.  This 'scattering tail' becomes more pronounced as the observation frequency is lowered.

Comparing this profile with professional profiles obtained at higher observation frequencies (e.g., 1400 MHz) shows that this aggregate pulse profile is consistent with expected results.

Also shown is the current 'Uptime %'.  It is hoped to keep this above 90%.

The 'Aggregate Pulse Profile' panel is is refreshed every 15 minutes - although the profile only changes once a day after an observation and, as it is an average, its shape changes very slowly.

Note: The 'bump' which precedes the main pulse (from ~30 to 45 time bins) is of unknown origin and subject to further investigation.

HawkRAO 'Virtual Control Panel' - Source of Images

The above result images were produced by HawkRAO software which implements a 'Virtual Control Panel' providing manual and automatic operation of the data acquisition and analysis process...

NOTE: The software shown here is hard-coded specifically for the HawkRAO hardware setup and is not available for general use.  Please don't ask as a refusal may offend.