The Global Oscillation Network Group (GONG) Project is a community-based activity to operate a six-site helioseismic observing network, to do the basic data reduction, provide data and software tools to the community, and to coordinate analysis of the rich data set that is resulting. GONG data is available to any qualified investigator whose proposal has been accepted; however active membership in a GONG Scientific Team encourages early access to the data and the collaborative scientific analysis that the Teams are undertaking. Information on the status of the Project and the scientific investigations, as well as access to the data, is available on our WWW server at www.gong.noao.edu.
We are happy to report that after three years of network operations, the processed data duty cycle is holding steady at about 87%, the daily sidelobes are virtually invisible, and the overall technical performance and reliability of the network continues to be excellent. We are well on our way to the replacement of the initial 2562 pixel detectors with 10242 detectors (a.k.a. GONG+). The effort should pick-up steam when the production cameras from Silicon Mountain Designs and the high-speed electronics from DNA Enterprises arrive later this fall. The development of the GONG+ Data System has also begun and will proceed in parallel with the production of the upgraded systems. In order to exploit the full scientific potential of the GONG+ data, the Project has begun preparing for an additional phase, GONG++, a high-performance computing capability, to exploit this scientific potential of the high resolution data as we approach maximum solar activity.
Caption: An m-averaged, l-v spectrum, l = 0 to 1000 and the v range is 0 to 8 1/3 mHz, from 8.5 hours of data taken centered on the solar disk center. The data were obtained by interfacing the new camera to the existing GONG data collection system. This arrangement allowed observation of only about 1/16 of the area of a GONG+ image. The diagram shows no spatial aliasing and a high signal-to-noise, with p-mode ridges beginning to merge into the background at an l-value of about 1000. A few faint temporal aliases are attributable to the temporary nature of the data collection system.
The GONG network continued to run well for the third quarter of 1998. Much of the network down-time was attributable to the scheduled preventive maintenance visits which took place at El Teide (24 June-4 July) and at CTIO (4-14 September). An incident of note, which resulted in nearly seven hours of downtime, occurred at CTIO when one of the electronic cards in the data acquisition system failed. The on-site staff discovered the problem, and with help from Tucson, were able to isolate the problem card.
A mix of circumstances at Udaipur caused the telescope to be shut down with unfortunate regularity. During the monsoon season, when little data could be acquired anyway, Udaipur Solar Observatory was suffering prolonged power outages. Although there is a backup generator which will power the GONG telescope, it was felt that it would be best not to run the generator during prolonged periods of power outage, which resulted in the site staff bringing up the instrument each morning and shutting it down each evening. A problem also occurred with the half-waveplate rotator. It appeared that it was not getting power and we suspected a blown fuse. When the on-site staff was able to investigate, they found a good fuse, but a fuse holder that was not making a good connection. That was soon remedied, but about a day of down-time accumulated. The lens slide mechanism, which operates during the daily calibration sequence, is beginning to show signs of age. There have been occasions when it has not moved the calibration lens into its final position by the time image integration has begun. This situation has arisen most visibly at Big Bear, but has not been bad enough even there to cause total loss or corruption of the calibration images. The situation will be addressed more thoroughly during a November PM visit to Big Bear, and at all the other sites as PM trips occur in the future.
During the past quarter, month-long (36-day) velocity, time series and power spectra were produced for GONG months 29 and 30 (ending 980420) with respective fill factors of 0.92 and 0.79. The p-mode reprocessing campaign (data that have been reprocessed or initially processed with the improved p-mode pipeline) added GONG months 4, 5, 6 and 7, boosting the available data set to months 4-30 (950823-980420). The project is also producing time series and power spectra from the intensity images. These products were generated for GONG months 21, 22, and 23.
Ed Anderson, who had been part of the GONG Project since its inception and who was integral to the development of the DMAC's reduction pipeline, left the Project and NOAO in July. Ed has headed north to Flagstaff, the Northern Arizona University campus, and the observatories of the Arizona northland. The Project has also said good-bye to Enrique Chavez, who had joined the Project in January 1996. Enrique leaves us to join Lucent in Denver. We will miss them both and are grateful for their many contributions.
We have completed a revised first-guess table for the peakfinding process, thanks to the efforts of Ed Anderson and Rachel Howe. In the end, the original set of central frequencies proved to be adequate, but the splittings and widths have been revised using values derived from an average of six three-month time series. A final test, which consisted of finding peaks in a three-month series twice using both the original and revised tables, showed no significant differences in the derived frequencies. With the new guess table complete, and with the removal of the temporal deconvolution step (discussed below), regular production of GONG frequencies has resumed. With the rapidly rising solar activity level, the (thankfully brief) hiatus in SoHO observations, and the completion of three years of GONG data, the resumption of regular GONG frequency production will contribute to the advancement of helioseismology and solar physics.
The temporal deconvolution step has been dropped. We had been performing a "brute-force" deconvolution of the first 30 mHz segment of the temporal window power spectrum using code provided by H. Antia. While this was effective and relatively fast for a 36-day spectrum, it was too slow to use for the current 108-day spectra. R. Komm has taken the lead in developing an alternate strategy. First, he tested an FFT-based method, but found that while it was fast enough, it resulted in nearly 25% of the points in the deconvolved spectrum becoming negative. This is much higher than results from Antia's method, and it clearly adversely affects the number of good fits from peakfind to an unacceptably low level. The same conclusion was reached when deconvolving the entire frequency range of the temporal window rather than the first 30 mHz. There was also an attempt to incorporate the temporal window into the model used by the peakfind routine, but this too is unacceptably slow.
In light of these tests, and since deconvolution primarily affects the estimated widths and amplitudes and not the frequencies, we have decided to eliminate temporal deconvolution of the spectrum prior to peakfinding, but instead, determine a correction factor for both the widths and amplitudes. This development has begun and has resulted in two important bits of information. We again verified that the frequencies are unchanged, and we found that the widths are increased (and the amplitudes decreased) by a nearly constant factor approximately equal to the duty cycle of the window. We are further testing the second conclusion by performing an analytical investigation, and by comparing the time series with different fill factors. We will refine the width and amplitude correction while the mode parameters are being produced.
In anticipation of GONG+, C. Toner has developed a method of merging images obtained simultaneously at widely separated sites. Currently, we merge spherical harmonic coefficients which would preclude the usage of GONG+ data for local helioseismology, which is one of the main scientific objectives of the GONG+ upgrade. The new approach will register images for each site-day into heliographic coordinates, and detrend them with a two-point backwards difference filter. Simultaneous images will then be averaged together with equal weights. For traditional helioseismology, the merged images are then decomposed into a time series of spherical harmonic coefficients which are corrected for the average image degradation by division by the merged network modulation transfer function (MTF). Tests with current GONG data produce a power spectrum that is virtually identical with the power spectrum produced by the current merging algorithm, and the process is computationally efficient, eliminating the need for "real" (expensive) image restoration. We're not out of the woods however, the GONG+ images will provide a new challenge since the higher spatial resolution will require high accuracy and precision angular registration between the sites.
As we begin to track solar cycle variations in anticipation of solar maximum, we have looked at the frequencies for four three-month GONG time intervals and have found that they display an interesting variation in a2 and a4.
Figure 2 shows the variation of the magnetic flux density, while Figure 3 illustrates the average latitudinal distribution of the flux for the four time periods. Notice that for interval 1, at the end of the previous cycle, the flux is concentrated close to (and a little south of) the equator and is distributed asymmetrically, while for intervals 2 to 4, at the beginning of the new cycle, the flux is distributed at higher latitudes. The activity of intervals 1 and 2 is comparable. The magnetic flux is much stronger at interval 3 and increases further at interval 4.
Figure 4 shows the frequency differences (i.e., changes in a0) between
Intervals 1 and 2 (diamonds), Intervals 1 and 3 (stars), and Intervals 1 and
4 (triangles), averaged over l, from 1-150, and in frequency bins of about
70 microHz, clearly demonstrating the well-known correlation between changes
in the frequencies and changes in solar cycle activity.
The degeneracy between modes of the same l and n, but different m (azimuthal order), is lifted by effects that break the Sun's spherical symmetry, the most important of which is rotation. The coefficients of the Legendre polynomial series are commonly used to parameterize the frequencies, where the odd-order coefficients, a1, a3, ..., are used in rotation inversions and reflect the advective, latitudinally symmetric part of the perturbations caused by rotation. The even-order coefficients, a2, a4, ... , which are much smaller than the odd-order coefficients, are sensitive to latitude-dependent properties. a1 and a3 showed no discernible changes, but the a4 coefficients, which are plotted in Figure 5 for the four time intervals, illustrate an interesting variation. The curves suggest a relationship to the latitudinal distribution of the magnetic flux as well as the total magnetic flux, but we will need more data to confirm this comparison. We also see changes in the a2 coefficients, which for the periods we have examined, have values about half the size of the a4 coefficients and of opposite sign. This behavior is different from what was observed during the previous solar minimum.
For the intervals 1 and 2, the independent SOI and GONG measures of a2 and a4 are nearly coincident, which gives us confidence. The changes in a0, a2, and a4 are already highly significant, and as solar activity picks up, we look forward to following these variations.
The camera development team has been operating one of the Silicon Mountain Designs 1M60_20 test cameras at the Tucson GONG facility. The temporary image sub-sampling interface, which provides a 242 line by 256 pixel portion of the camera's output to the existing data acquisition electronics, has provided a means of evaluating not only the camera's performance but also alternative observing methods and optical components.
DNA Enterprises of Richardson, TX, has been selected as the vendor to provide the video data acquisition system for the instrument upgrade, and will deliver a system incorporating two Texas Instruments TMS320C80 digital signal processors on a single VME board. The computational capability of this product is an astonishing 480 megabytes per second and should easily handle the camera's output data rate of 120 megabytes per second. We are eagerly awaiting the first prototype system for testing, which is expected to arrive sometime before the end of the year.
Modifications to the existing instrument synchronization system necessary for the installation of the new camera have been implemented and tested. The existing Cohu video cameras currently convert the video synchronization signals generated by the GONG electronics into pulses used by the waveplate speed control system. Because the new camera is not equipped to handle this conversion, a prototype circuit board has been built to emulate this function and the results are quite satisfactory.
The volume of raw data collected each week at each site is expected to be about 70 Gigabytes. Our plan is to use a pair of 18 gigabit hard disk drives to temporarily store accumulated data during the day and then transfer those files to the DLT tape drives (each with a capacity of 35 Gigabytes) at night. The purpose of the pairing scheme is to permit redundancy of data storage and to provide additional data capacity in the event of equipment failure.
We expect to have procured all of the equipment for the GONG camera upgrade by next summer and we will be deep into the system verification effort. Recent tests indicate that an l value of ~ 1000 can be reached with the new GONG+ instrument in average seeing conditions. Full disk images will be available as soon as the data acquisition system electronics arrive later this fall.
John Leibacher