18 May 2006
On 18 May 2006 we achieved the first six station fringes at 256 Mbps with Cambridge, Jodrell Bank, Medicina, Onsala, Torun, and Westerbork. There were first ever European 512 Mbps e-VLBI fringes on the Jodrell Bank - Westerbork baseline. The datarate was too low for Arecibo to get fringes even at 32 Mbps. We tested remote restart of Mark5As at the telescopes. After the test, there was a 13-hours science project at 128 Mbps. This went without major problems.

Return to top of page.

21 August 2006
Cambridge, Medicina, Onsala and Westerbork participated in the 21 August 2006 test. The test objective of restarting and rebooting MarkV units was successful, although we found it’s not possible to do this during a run and recover the station without restarting the run. We had the first 512 Mbp fringes (view fringe plots) between three European stations: Cambridge, Onsala and Westerbork. We also got fringes for the first time between Cambridge, Medicina and Westerbork using Medicina’s new 5 GHz receiver.

Return to top of page.

26 October 2006
The 26 October 2006 test was carried out using a new correlator control computer. Starting up correlation jobs was much faster. Initially there were network problems to Medicina and Torun, but these had been solved by the start of the observations. During the test phase we ran long correlation jobs to see how long we could maintain a data rate of 256 Mbps. Since this worked fine, we decided for the first time to run the science experiments at this data rate as well. There were two projects that targeted the X-ray binary LSI +61.303, and the eclipsing binary system Algol. Unfortunately, Onsala had to stop observing because of a storm. Due to various Mark5A related problems, we lost a significant amount of observing time during the night (30% and 20% for the two projects, respectively). Moreover, Cambridge stopped producing fringes after a couple of hours, because of a failing power supply unit in the MkIV formatter. The increase from 128 Mbps to 256 Mbps partly compensated for these losses. Both of the targets were nicely detected in the data analysis pipeline.

Return to top of page.

14 December 2006
The 14 December 2006 test went extremely smoothly with three science projects incorporated. Participating stations were Westerbork, Jodrell Bank, Cambridge, Medicina, Onsala and Torun, with a 256 Mbps data rate easily achieved from all stations. During a short period we experienced problems with connections to Torun and Medicina, but these were resolved without losing significant amounts of data. Problems at the correlator were few and quickly corrected, again without losing much time or data.

Return to top of page.

29 January - 1 February 2007
The 29 January 2007 run was set up as a double adaptive run (29 January - 1 February), in which the first run was to be used to select targets for the second run. For this reason it was essential for the data of the first run be reduced as quickly as possible. The usual 6 stations participated: Westerbork, Cambridge, Jodrell Bank, Torun, Onsala and Medicina, at a data rate of 256 Mbps. A few problems were encountered, but at a technical and operational level the observations went extremely well. No serious problems were encountered, and restarts of the correlator and Mark5 units were few.

The reduction strategy also worked well, and pipeline maps were produced for 14 out of 16 sources by the end of the first run. Unfortunately, none of the sources were found to be in an active state and the second run was cancelled. We did, however, reserve four hours of test time with all telescopes on 2 February. During this test, in which we observed a strong calibrator source, we were able to achieve a one hour 512 Mbps run with five stations (excluding Torun) and produce the first map at 512 Mbps.

Return to top of page.

20 February 2007
Torun, Medicina, Onsala, Westerbork, Cambridge and Jodrell Bank (MKII) participated in the 20 February 2007 test. One science proposal to detect potential calibrator sources for future observations was accepted for this session. The session began with general tests, during which a 512 Mbps data rate was again achieved from five out of the six stations. The science observation was then run at 256 Mbps and, excepting for the loss of approximately 30 minutes of data from Medicina due to a Mark5 unit, all went very smoothly. Cambridge, unfortunately, was not able to produce usable data, but this problem was not related to e-VLBI and has since been corrected. None of the four calibrator sources were detected, indicating that they are resolved.

Return to top of page.

5 Mar 2007
The first e-VLBI test at 1.3cm (22 GHz) was conducted on 5 March 2007. The main goal was to produce first fringes to Metsähovi, but the new dual polarization receiver at Onsala was also checked. Other participating telescopes were Cambridge, Jodrell Bank (MKII) and Medicina. The experiment was a great success with fringes to all five telescopes at a 256 Mbps data rate.

This test proved to be very useful to check the telescope performance before the K-band session observations started. This is normally done in the ftp fringe tests, when a small amount of disk-recorded data is transferred from the telescopes to JIVE and analysed by the software correlator. This time the ftp test failed because Effelsberg experienced high winds and the signal-to-noise ratio was very poor. Using e-VLBI to transfer data to the hardware correlator in real-time, however, performance of the remaining telescopes could be checked.

Return to top of page.

28 March 2007
An e-VLBI science run took place on the 28 March 2007 involving Jodrell Bank (MarkII), Cambridge, Onsala, Westerbork, Torun and Medicina at a data rate of 256Mbps. There were some restarts due to problems with station units and the correlator, although these were not directly e-VLBI related, as well as time synchronization difficulties which will have to be investigated further. This was the first science run in the L-band, and the first to be completely conducted by an operator and a support scientist without the direct supervision of the technical staff, illustrating the increase in robustness and user-friendliness of the system over the past few months.

Return to top of page.

22 May 2007
The e-VLBI test run on 22 May with Cambridge, Jodrell Bank, Medicina, Torun and Westerbork was successful. A new switch/router at JIVE, new IP address space for some stations and upgraded Mark 5As worked flawlessfly.

The connection between Torun and JIVE was unavailable at the beginning of the test but was re-established halfway through. With the connection re-established, it was decided to restart the correlator and Mark 5s for a clean start. The clock search at 256 Mbps was fine, but most stations could not keep up at the 512 Mbps rate. This was likely due to a change in data transfer configuration after the Mark5As were rebooted.

Return to top of page.

25 June 2007
The 25 June 2007 test was the smoothest to date. With participation from Cambridge, Jodrell Bank (MkII), Medicina, Onsala, Torun and Westerbork stations, the run began with a science observation of almost 12 hours at 256 Mbps, including a 2.5 hour stretch with no problems. Following the science portion, the test continued for another four hours at 512 Mbps, again without any faults or failures, a first ever for six-station e-EVN run at 512 Mbps. (See the press release "First 512 Mbps e-EVN image from six stations".)

The connection to Shanghai was also tested operationally for the first time. Although not directly successful, this was a very useful first step towards including Shanghai in science e-VLBI.

Return to top of page.

21 August 2007
The main goal of the 21 August test run was to test the modified UDP Mark5A code and push the data rate as high as possible. Overnight there were science observations, the first spectral line project with e-VLBI. The array was slightly different from the previous runs; the Cambridge telescope was replaced by Darnhall, and a single dish was used in Westerbork instead of the phased array. Because of the low data rate of 32 Mbps, the correlation jobs were remarkable stable. The 8-hour experiment was correlated in two jobs of 5.5 hours and 2.5 hours. There were minor problems with the telescopes, Medicina being too late in some of the fringe-finder scans, and Westerbork showing too high polarization leakage. Otherwise the experiment was a great success.

Return to top of page.

28 August 2007
On 28 August 2007, e-VLBI fringes between Shanghai-Mopra and then Shanghai-western European stations were demonstrated live at the APAN networking conference in Xi'An, China. Additionally, there was a short time of about ten minutes when there were fringes between Mopra and European stations as well, probably the longest VLBI baselines ever achieved at 12,300 km. See the full press release here.

Return to top of page.

6 September 2007
During this run we continued testing the UDP-based Mark5A software. At low data rates there were no problems. However at 512 Mbps we could not get fringes to all telescopes. It is likely that the error occured in our local network, and this is under investigation.

After the test phase there were two science experiments. Because Cambridge was still unavailable, we used again Darnhall. There was only one-way communication with the telescope, and reaching even 256 Mbps was not possible with TCP. We decided to use the UDP-based Mark5A code for Darnhall, which is the first science application of UDP for e-VLBI. Both experiments were successful, and the second one produced the first Astronomer's Telegram report on e-EVN science observations.

Return to top of page.

8 October 2007
The 8 October 2007 test involved real-time science observations with the Parkes, Mopra and ATCA telescopes in Australia, correlated at JIVE. Data were sent from each telescope at 512 Mbps over three dedicated 1 Gbps lightpaths, provided by a collaboration of AARNet, CENIC, CANARIE and SURFnet. Physically the paths led from the observatories to Sydney, across the Pacific to Los Angeles, Seattle, Chicago, Amsterdam and finally to JIVE in the Netherlands. The UDP transport protocol was used with a modified Mark5A control code developed at JIVE. At the Australian end, data acquisition was done on the LBADR VLBI system (a derivation of the PCEVN) with a Mark5A command emulator, and converted on-the-fly to the Mark5B format. The data were received at JIVE on Mark5A+ units. Data transport via the lightpaths was nearly perfect, and 512 Mbps transfer was easily sustained for more than 12 hours.

Return to top of page.

9 October 2007
We targeted the highest data rates yet in the formatter test on 9 October 2007, using two versions of the modified Mark5A software (both UDP-based). There were numerous problems. However near the end of the run, we managed to get data from Torun to JIVE at 1024 Mbps, the highest to date, with 4% packet loss. The data have been successfully synchronized, displaying green LEDs on the correlator Station Units.

Return to top of page.

11 December 2007
On 11 December 2007 we conducted an e-VLBI test during which we, for the first time, obtained real-time fringes from five telescopes at nearly 1 Gbps. The test started with six telescopes at 512 Mbps (Westerbork, Onsala, Torun, Medicina, Jodrell Bank and Cambridge). After making sure the 512 Mbps mode worked without problems, we tried higher bandwidths. Unfortunately, Onsala could not participate, as the motherboard/CPU of the Mark5 at Onsala has not been upgraded. As can be seen in the first plot, good fringes were obtained from all five telescopes. The second plot shows the aggregate data rate into JIVE.

Although we are still some way from advertising this as an operational observing mode, it clearly has a lot of potential. With it we will be able to use available bandwidth nearly to the limit. We are now working on a scheme to avoid dropping headers, which will make it easier for the correlator to keep the data streams synched, and we will investigate the possibility to create a closed loop in which the drop rate is automatically adjusted by the availability of bandwidth.

Return to top of page.

21 January 2008
On 21 January a new script was exercised for downloading and installing the jive-mark5 code, and some problems with Onsala and Medicina (both of which upgraded their Mark5s recently) were solved. After this, 512 Mbps data rate was tested from all stations (Mc, Wb, On, Tr, Cm and Jb2) and found to work reliably. It was followed by a run at 1 Gbps with packet dropping. The fibre links were filled to near-capacity, but the correlator could not keep up due to data headers being discarded with the dropped packets.

Return to top of page.

24 January 2008
On 24 January a new version of the correlator control code was tested. This version, implemented by JIVE, had modifications to the packet dropping algorithm so that only data packets are dropped at the sending side while headers remain intact. There was apparent improvement, with the correlator able to synchronise the data streams immediately while only dropping 4.5% of the packets. Due to telescope constraints there was only one baseline that produced fringes (Torun-Medicina), but the correlation jobs were succesfully run for several hours at 1 Gbps with three stations (Mc, On, Tr). Click here to see aggregate bitrate traffic to the correlator.

Return to top of page.

5 February 2008
On 5 February 2008, Arecibo rejoined e-VLBI operations after a long absence. At this time, connectivity is still limited by a shared 155 Mbps connection to the mainland United States, but this situation is expected to improve very soon. 64 Mbps produced immediate fringes. 128 Mbps was only reached for very short periods, no doubt because of competing traffic. However, by dropping one in five packets at Arecibo, steady green LEDs and good fringes were obtained at an effective transfer rate of approximately 100 Mbps.

After this success, we observed 3C395 (image) with six stations (Medicina, Westerbork, Onsala, Torun, Cambridge and Jodrell Bank) at approximately 1 Gbps, with only formatter data from Onsala. The Cambridge microwave limits useful transfer to 128 Mbps, but from a correlation point of view, this was a robust 1 Gbps run sustained for several hours.

Return to top of page.

14 February 2008
On 14 February 2008 we conducted the first successful real-time data transfer between the Effelsberg radio telescope and the EVN correlator at JIVE. Formatter data was sent at a modest data rate of 256 Mbps to prevent disturbances to an ongoing geodetic data transfer to the Bonn correlator. The data transfer ran smoothly for about an hour with green LEDs at the EVN correlator for the duration.

We also conducted a separate Mark5B-based e-VLBI test of JIVE Mark5 control code. This code was used to direct the transfer of Mark5B data from Westerbork to a Mark5A+ at JIVE via UDP, at conservative rates of 128Mbps and 256Mbps. There were no green LEDs at the station units, and testing with official Mark5A code indicated an A+ configuration issue. In principle, however, things are looking good for Mark5B-based (B to A+) e-VLBI.

Return to top of page.

8 April 2008
On 8-9 April 2008, the first production 512 Mbps e-EVN observations took place. Participating telescopes were Cambridge, Jodrell Bank, Medicina, Onsala, Torun, and the Westerbork phased array. Two science projects were observed: one normal non-triggered and one triggered.

The correlation job on the first project ran for nearly 13 hours, the longest uninterrupted correlation job at JIVE, including disk operations. The first 45 minutes of the second project were lost because of a control computer problem unrelated to e-VLBI, and another hour of Mc data was lost due to a problem with a Mark5A. In spite of this, the run went very well and we collected about 8.5 hours of data on the X-ray binary Cyg X-3, the preliminary results of which were published in The Astronomer's Telegram #1476.

You can view a plot of the data throughput for both science runs here.

Return to top of page.