TOG minutes 1 April 2004

Present: Alef (chair), Foley (minutes) Buiter

Aragon, Balodis, Bertarini, Bezrukov, Campbell, Cassaro, Charlot, Dassing, Foley, Garcia-Miro, Gonzalez, Graham, Gunn, Helldner, Himwich, Hinteregger, Hoefer, Hrynek, Kilger, Kraus, Kronschnabl, Kuper, Lauber, Leeuwinga, Lindqvist, Maccaferri, Martinez, Millenaar, Mueskens, Orfei, Orlati, Parry, Parsley, Pazderski, Porcas, Quick, Ritakari, Romney, Roy, Schilliro, Schlueter, Schwarz, Smythe, Stiepel, Tenkink, Tuccari, Wei, Wunderlich, Zeitlhoefler, Zsolt, de Vicente Absent: (sick) Burgess, Buiter,
A full list of participants can be found at the TOG site

Greetings from Prof Schlüter, the director of Wettzell.

Local arrangements

1. Approval and last minute additions to agenda

No changes were proposed

2. Review of actions from last meeting

Check schedules before observing to check that they can be correlated (Campbell)
Campbell said that it can be done, but Porcas asked who would do this; Romney suggested automating the task. Campbell pointed out that it needs to be done before correlating anyway. Alef suggested moving schedule deadlines to 3 weeks before the session to allow for this, but Graham noted that it was too late to do this for the May session. Charlot suggested that the deadlines could be kept as is, but the PI could be encouraged to deposit schedules one week earlier to allow for checking. Campbell and Porcas said that if the modes were specified when the PC allocated time this could easily be avoided.
Some problems only are seen when the Friends try to drudg the schedule.

For the November session the following was agreed: Deadline for PIs is 3 weeks. Campbell checks schedules in this 1 week and Gunn will run drudg for further testing of the schedules. The stations can start downloading at the earliest 2 weeks before the session. The possibility of an incoming area for schedules and a separate outgoing area was mentioned, but nothing was decided on this.

Action: Friends to contact Gunn if there is a problem
with a schedule
Barrel Rolling
This item is now irrelevant as the Mark5A does not need it
Spare parts (Wunderlich)
With the Mark5A the spares problems with tape units are irrelevant, and when the digital BBCs arrive most of the others are also gone. However we still need the BBCs and there are several critical spares there. Action: Wunderlich make a new list of spares
Scheduling FTTs and NMEs (Paragi)
FTTs and NMEs are to avoid Mondays (Done)
Request Mark5 inspection software (Alef)
Done
Make a Mk5 bandpass inspection tool (Maccaferi)
Done. The tool is chchk, and it can do sampler statistics as well
Arrange real-time fringe tests
Several have already been done. We should schedule one before each wavelength change as a minimum.
Phasecal Analysis (JIVE)
Pogrebenko demonstrated a phasecal analysis system. The algorithms were worked out in Mathcad, but these would be coded so that stations could use them
Stations to do feedback via logs, JOPS and TRACK

Stations to check linearity in BBCs
It was agreed that the linearity checks for BBCs was needed to produce valid calibration and to make sure that the total power to formatters (and so sampler statistics) would be good. Hartebeesthoek and Westerbork reported that they had checked BBCs
Send stations new ANTAB calibration script(Reynolds)
Reynolds reported that this had been done, but that a new improved script was now ready to send
Report on how to use the VLBA radiometer (Graham)
Graham described how to use the VLBA 80Hz noise cal using the FS .
Plan a timescale for Mark5-only operations (Reynolds)
superceded
Stations inform Himwich about what to do for LO settings
Himwich said that he had not had feedback, but he was making a proposal for setting LOs with the FS
Stations tell Reynolds about slew rates
Some had been reported
JIVE contact stations about what to do with their thick tapes
Parsley reported that thick tapes had been disposed of

3. Reliability/Performance of the EVN (Paragi)

The full text is available here

Session 3 2003

Station Comments
Jb Lost the C-band session because of an LO offset
On very low sampler statistics
Nt Very low amplitudes in EC017 - pointing? Piggyback problems
Ur Troubles with C-band receiver, so used uncooled version.VC02 and VC04 USB have low amplitudes due to sampler statistics

Session 1 2004

Station Comments
General Troubles with disk bank switching. Stack 2 lack fringes sometimes
Ef Lost the X-band session (N04X1 & GM052A) because of an LO offset
Wb Polarization purity much improved, except L-band (RFI?) and UHF; 100MHz limitation in C-band
Jb Mk5 cabling problems in 2-headstack experiments
On Sampler statistics problem solved
Mc Lost the K-band session because of an LO offset. IF limitations at L-band
Nt GPS clock file wrong ( ± 10 μsec)
Ur Lost the K-band session because of an LO offset.
Sh N04L1 BBC7 and BBC8 missing.

The sampler statistics problems are always in the sense of too many low bits. Pogrebenko said that there is a fix at the JIVE correlator if the sampler statistics are only slightly wrong - but if they are wrong by more than a few percent this cannot be made. Quick said that he has seen this at Hh when there was a bad connection from VC to formatter. Pogrebenko said that the Jb problem with 2 stack experiments where only 1 was recorded could be repaired in software.

Action: Pogrebenko distribute
the 'lost stack' software to the Bonn correlator

Status of wide-band tests

See also notes above


 Ef Jb Mc Nt On Sh Ur Hh Mh Tr Wb
K y Mk5? y y y y y
½

X y
y y y y y y

½?
C y Mk5? y y y y y y
y ¾
L y Mk5? y y y ¾ y y
y y
These show some limitations, but we could advertise C-. X- , and K-band on the EVN at 1Gbps. L-band cannot be done.

Scheduling

LO changes at Mc and Ur (K-band), Ef (X-band) and Jb, Wb, Ef, Hh (C-band) badly affect reliability. The Graham switch box will make patching problems easier to avoid. Graham suggested a short term solution may be to patch VCs 9-14 to the high patch. It is possible to force SCHED to specify VC9-14 in the high patch range. Himwich pointed out that the connectors are not designed for repeated re-patching and doing this would damage them. Action: All: try to get a patching system to avoid these problems and get it working before May Reynolds: Get modifications into the Walker SCHED package

Users will need JIVE assistance for scheduling 1Gbps experiments (as freq.dat needs to be changed in SCHED), and a new version of the SCHED and the FS should be tested before session 2.

FTP fringe tests

These can at present only be done for 128 or 256Mbps; higher bitrates are not supported in the software. Porcas asked whether we still need fringe test tapes (or disk-packs) as the FTP tests are faster. Paragi said that FTTs still test the full system and can do the highest bit rates and also check sites without high speed connections. The K-band sensitivities were such that a few seconds of data were not enough to see clear fringes. Gunn replied that the FTP tests reflect the system as it is going to be used for the session (and not that of 1 week before) and so give much more confidence. Graham suggested the best compromise was to keep the bit rates low, and only observe short scans; Paragi said the minimum possible values to schedule are 32Mbit/s and 15 seconds. Alef said that we should use both "traditional" FTTs some days before the comming session and FTP transfers immediately before each wavelength.

Phasecal detection (Pogrebenko)

Pogrebenko presented software to extract phasecal signals from Mark5 data streams. This was at current implemented via a dump to disk and then using a Mathcad script. In general a complex bandpass is measured. The software can use autocorrelation data to check behaviour of the LO system and BBCs, and cross correlation data to check on de-correlation - and hence improve closure phases. The 16MHz (complex) bandpasses at Jb showed a problem with poor image rejection, possibly caused by a poor hybrid. Himwich asked whether the autocorrelation checks could be done at the stations, and Pogrebenko replied that it could be included in the channel-checker software, chchk. There followed a discussion of who had used chchk and it was noted that the developers would give a demonstration of this in Wettzell on the following day.

4. Amplitude Calibration (Reynolds)

The full text is available here Reynolds pointed out the importance of: He showed a graph, showing that the general trend is that the amplitude calibration has been improving over the last years; from 30% errors in 2002 to 5% in 2004. He also pointed out that much faster turnaround would be needed as there is now a project (ALBUS) to send calibration data with the data produced by the correlator. This will in general be within 2 weeks of observing, so the rxg files must be at the correlator by then. The Mark4 racks can now produce good quality calibration, but some stations still have problems. Reynolds has now improved versions of the calibration scripts ready to send to stations. These will be sent in April 2004, and these will also support VLBA racks.

Gunn pointed out that the opacity corrections at K-band are still poor, as there are too few bright sources and too few good calibrators. Roy said that he was now working on improving opacity models.

Action: Graham, Himwich and Gunn: produce improved K-band corrections (particularly opacity). Action: All: Check BBC linearity up to 25000 units and zero -level offsets.

Continuous calibration in VLBA racks

Graham described how he had made 2 new FS programs specifically to use the 80Hz calibration diode used be the VLBA. See how to use the VLBA 80Hz noise cal using the FS . Tuccari asked what extra hardware was needed, and Graham said that an interface box was needed. Pazderski said that one had been made at Torun for local use and another for China.

Alef reminded the meeting that although Conway had supported the calibration effort he was stopping this now, although he was writing a manual. Alef asked for volunteers to continue the work.

5. Automatic flagging (Reynolds)

For the stations that use automatic flagging this works well and improves dramatically data quality for phase-referencing. Some telescopes cannot provide this data. When good flagging is available it greatly improves the EVN's final data. Wb has problems as it is a tied-array. Pazderski said that Tr has recently improved its software to give good flagging data. Himwich said that Ur has been working on improving their antenna control software. Noto plans to improve it soon. Action: Noto improve flagging data There was a discussion of the interaction between noise source firing and changing source. In particular whether the noise source should always be fired during POSTOB. When we move away from tape the scan length could become very long. Romney said that a 2-level scheme for noise source firing would be best. Alef also it would be good to include a warning in SCHED if noise source switching happens too infrequently.

6.MK5 (Smythe)

The full text is available here Quick commented that Conduant now supply 160GB Hitachi disks. Ritakari said that the latest Hitachi disk technology was in the 160, 250 and 400 Gbyte series. He also pointed out that there are already inexpensive chips for SATA port multiplication (up to 15disks).

Several stations only keep tapes drives for the benefit of the VLBA correlator.

Himwich said that the FS would keep track of Mark5 version numbers.

Graham mentioned that Conduant were shipping some Mark5 systems with old firmware versions. Firmware should also be kept updated, although this needs some special cabling and software as the PROM cannot be removed.

There followed a discussion on the best way of implementing SATA in the Mark5. It was concluded that the current plan of keeping a parallel ATA (PATA) connector would severely limit future expansion. The meeting agreed that the Mark5B must have a SATA connector, possibly in addition to the planned PATA connector, to allow for future expansion

Action: Alef contact EVN chairman to express the need for SATA connection capability in Mark5B.

Purchasing and Logistics

Eb and On are ready for disk-only operations. Mc, Ur, Nt Hh and Wb will be ready with enough disk packs by May 2004. There is a scarcity of shipping boxes. Bonn has boxes and foam and can help in case of problems with purchasing shipping boxes. Action: Smythe ask Conduant to send shipping boxes with each disk pack kit

Mark5B at the VLBA (Romney)

A Mark5 pre-meeting had taken place in Bonn on Monday 29, to prepare a recommendation for the TOG (Alef, Brisken, Campbell, Graham, van Langevelde, Porcas, Romney, Tuccari, Whitney). Romney gave a summary (the full text is available here) Parsley stated that the JIVE correlator would only shift to Mark5B playback when Mark5A was no longer needed. Mixtures of tape/Mark5A/Mark5B should be avoided. There was a discussion of the price tag. Hinteregger said that a complete Mark5B upgrade would be about $8000, which was debated by Smythe. Action: Alef take the recommendations of the pre-meeting as agreed by the TOG (for all EVN stations to change to Mark5B) to the board of directors. 

7. Field System (Himwich)

The full text is available here Maccaferi asked about the need to change recorder type; Himwich replied that the Mark5 commands would work without this, but automatic bank switching would not be used unless the FS knew that the recorder was a Mark5. The JPL has a program chooser to select which recorder is being used.

8. Sched Developments (Reynolds)

The full text is available here

A new version of sched will be released in April or May, and it will support Mark5-only stations, including ftp-data transfer.

Action: All: Check their SCHED catalog entries and send updates to Reynolds

9. UHF (Tuccari)

The full text is available here

F04U1 was an dual-pol observation of the 860-892MHz band, done to evaluate new UHF receivers. All had fringes, but most stations (except On) had problems with RFI.

Tuccari also suggested that stations should consider adding filters to reduce RFI

Porcas reported that narrow-band user observations were planned for 805, 851, 866 and 999MHz.

10. MkIV Decoder (Smythe)

The full text is available here

There is new firmware ready for the Mark4 decoder, which can extract up to 8 phasecal tones ( either as one channel with 8 tones, 2 channels with 4 tones or 4 channels ( 2 tracks with USB and LSB) 2 tones. There will also be a calculation of state level statistics. Himwich said that the FS would support this when the new firmware was installed. R. Cappallo is ready to send PROMs out, but stations can download it and burn the PROMs themselves. Wunderlich stated that it is still possible to buy this type of PROM.

Action: Winderlich request the firmware code

11. IF switchbox (Graham)

The full text is available here

The patch command has never actually changed anything except at the DSN. There is now a cheap and effective device (in the prototype phase) to switch patching from software, with 70dB channel separation. The estimated price is €1300 if stations build it themselves.

12. Digital BBCs (Tuccari)

The full text is available here

There was an meeting on 31st March at which the DBBC operational group was set up. Although the overall project is now defined the details need to be agreed on and presented to the EVN board in May. The plan is to have 4 prototypes ready by the end of 2005. These will have only standard components - no custom devices, and have an upgrade path.

The provisional list is (partly assemled in a pre-meeting in Bonn on Monday 29 before the TOG).

Gunn asked how many channels were planned for the DBBCs; Tuccari said that the current plan was 64 channels (split over 4 IFs) for each of 2 VSI outputs giving a maximum data rate of 2Gbit/s. Pazderski asked about the FPGAs to be used and probable component costs, but the cost is not yet clear as the design may yet change.

13. e-VLBI status (Parsley, Smythe, Ritakari)

JIVE (Parsley)

Parsley outlined the GEANT project and reported that there were now six 1Gbit/s lines from Amsterdam to JIVE. The international links to Amsterdam are 10Gbit/s. There is one 1Gbit/s local loop (Westerbork-JIVE). There are two dedicated Mark5 systems connected for eVLBI and the option of cabling up another four. Most of the network hardware was 'plug-and-play' with no special tweaking. The first image (2 baseline) milestone has been reached.

The next plans for real-time eVLBI were being made; the buffers are already large enough (160msec) for the expected maximum delays inside europe (10-20msec). The main bottleneck is moving the correlator to telescope observation time. The major items for the next 3 months are:

Experience has shown that the current generation of Mark5 units will not support 1Gbit/s eVLBI but 512Mbit/s is a realistic goal, as this has been attained on benchtops and local loop tests.

Haystack (Smythe)

The full text is available here

32Mbit/s eVLBI has been demonstrated from Westford-Goddard, Westford-Onsala and Kashima-Westford. There are plans for fast links to Kokee Park and Regensburg for intensive UT1 observations in near real-time. There is a 3year grant to work on real-time protocols for eVLBI and there is also collaboration with the DRAGON project. The best results to date were obtained using the UDP protocol.

Parsley asked about how to get UDP transfer in the Mark5A program, and was told to contact Lapsley for details.

Other developments (Ritakari)

A full version is available here

Ritakari pointed out that eVLBI should be developed from 'off-the-shelf' components, as there are already several reliable UDP protocols for Linux. Lab tests for normal desktop motherboards give about 500Mbit/s, but soon motherboards with native 1Gbit/s (not connected to the PCI bus) would be readily available. The problem with the current generation was the PCI-bus arbitration which limits speeds to roughly 400-500Mbit/s.

14. Station + correlator reports (as time permits)

As the station and correlator reports had already been been distributed there were only 2 items.

Robledo (Garcia)

The sub-reflector has been repaired and there are now options either to track for best efficiency at that elevation or to remain in one fixed position. The default do date has been to remain fixed. Graham pointed out that the usual case for Effelsberg was to move the sub-reflector to optimize efficiency. Garcia wanted to know what the TOG proposed, and it was agreed that Robledo should track the sub-reflector.

Westerbork (Foley)

The full text is available here

Foley outlined the plans for a fully digital tied-array at Westerbork, in particular how this would reduce possible causes of crosstalk and interface directly to the Mark5B system in future.

15. AOB

None

16. Time and place of the next meeting(volunteers please!)

Proposed meeting date in late Nov. 2004 - Jan. 2005 - possibly at Jodrell Bank

TOG mini-workshop 2 April 2004

The aganda of the mini-workshop is here.