1. Achieving a Quantum Leap in Observation Density Bill Petrachenko, NRCan

2. 2 The purpose of this study is to:  Investigate if it is technically feasible by 2010 to increase the number of observations per day at each site by an order of magnitude  Today, each site makes ~200-450 observations per day  The goal is to make ~3000 observations per day at each site (or about one observation every 30s).

3. 3 Secondary Goals: We want to achieve our main goal of increasing observation density by an order of magnitude while still having:  Adequate signal sensitivity to detect all sources in the ICRF and VLBA Calibrator list  High enough “Delay” precision to meet the 4 ps target of VLBI2010  The ability to do all this with an array of low cost ~12m antennas

4. 4 Why do we want to do this?  Solutions for IVS products will improve in the following ways:  It will be possible to sample the troposphere more uniformly  It will be possible to model the troposphere with more temporal and spatial structure  It will be possible to model the reference oscillators with more temporal structure  It will be possible to increase the number of degrees of freedom (provided, of course, that the parameterization does not increase too quickly as well). More degrees of freedom will lead to:  More robust solutions  An increase in precision due to Sqrt(N) averaging

5. 5 Why do we want to do this? (cont’d)  A large number of sources per day opens up the potential for  The use of a single observing strategy to fulfill all IVS goal (ITRF, EOP, ICRF) which in turn will increase the self- consistency of products  The observation of hundreds of sources on a daily basis, (which firmly anchors all measurements to the ICRF)  The monitoring of thousands of sources on a regular basis to support Astronomy and spacecraft tracking phase reference applications

6. 6 What two factors are needed to increase the number of observations per day?:  Shorter integration times  Shorter switching times between sources

7. 7  How can we achieve short integration times while simultaneously achieving high delay precision using weak sources and 12m antennas?  1. Use an optimized RF frequency sequence that enables phase resolution at a low SNR.  2. Use high bit rates  How can we achieve short inter-source switching times  3. Use an antenna with high slew rates 4. Use an antenna with a favorable mount configuration.

8. 8 The questions are:  How can we achieve short integration times while simultaneously achieving high delay precision using weak sources and 12m antennas? 1. Use an optimized RF frequency sequence that enables phase resolution at a low SNR. 2. Use high bit rates  How can we achieve short inter-source switching times ? 3. Use an antenna with high slew rates 4. Use an antenna with a favorable mount configuration.

9. 9 Why do we want to resolve phase?  The phase delay observable is very precise  E.g., at SNR=7 (minimum detectable signal), Xband phase delay precision is better than 3ps.  Unfortunately, the phase delay is also corrupted by an integer cycle ambiguity which needs to be resolved before it can be used.  It is true that the phase delay ambiguity can be resolved (even for S/X band) if high enough signal levels are used.  Our goal: Find a way to resolve phase even at low SNR so that very high delay precision is available even for smaller dishes observing weaker sources.

10. 10 How are we going to make this happen?  Use a low-cost decade bandwidth front-end like that developed for the Alan Telescope Array (ATA). This will allow continuous frequency coverage over a very wide bandwidth  Take a number of IF channels and space them optimally within the full frequency range. [Optimization will be done with respect to minimizing the signal level required to resolve the phase ambiguity.]  Use single band delays to resolve phase differences between bands, and then use these phase differences to resolve the phase in each band.

11. 11 Let’s investigate concrete examples:  Consider separately two continuous RF ranges:  2-15 GHz (i.e. an extended S/X)  8-33 GHZ (i.e. an extended X/Ka)  Within each of the two RF ranges consider either 3, 4 or 5 bands  Each individual RF band will have the following characteristics:  A continuous 1 GHz bandwidth  2 bits per sample at the Nyquist rate  Full two polarization processing  To make the scenarios more realistic, allowance will be made for as much as 200 MHz uniformly distributed RFI per band  To ensure success, 5 sigma confidence will be assumed for each case where phase needs to be resolved

12. 12 Now, we do an exhaustive search stepping each band across the entire frequency range (in 0.5 GHz steps) searching for the band configuration where phase can be resolved with the lowest signal levels.

13. 13 05 10 15202530 Freq (GHz) Optimized RF Frequency Sequences 3 RF Bands, 2-15 GHz 4 RF Bands, 2-15 GHz 5 RF Bands, 2-15 GHz 3 RF Bands, 8-33 GHz 4 RF Bands, 8-33 GHz 5 RF Bands, 8-33 GHz

14. 14 Number of IF Bands RF Range (GHz) Integration Time (s) 12m x 12m Minimum Flux (J) 12m x 32m Minimu m Flux (J) dtau (ps) at Min. Flux 32-1560.1770.0662.11 42-1560.1130.0422.05 52-1560.0880.0322.47 38-3260.3820.1420.45 48-3260.2040.0760.62 58-3260.1350.0510.61 Performance of Optimized RF Frequency Sequences

15. 15 Is this sensitivity high enough to detect the sources we want to use?  Consider 3 source lists:  Dave Schaeffer’s List  Core of regularly used geodetic sources  ~100 sources  Minimum flux = 0.09 J [Total Flux]  ICRF List  700+ sources  Minimum flux=0.09 J [Total Flux]  VLBA Calibrator List (from Leonid Petrov)  3500+ sources [800+ with full flux information]  Minimum flux=0.04 J [Small component flux]

16. 16

17. 17 Source List Name Min. Flux (J) Min. Integration Time (s) to phase resolve the Weakest Source Average of All Minimum Integration Times (s) in each List 12mx12m12mx32m12mx12m12mx32m Dave Schaeffer’s List 0.095.50.77.24.03 ICRF Sources0.095.50.77.11.02 VLBA Calibrators0.0428.23.94.10.57 Integration Times for Three Source Lists (assuming the Best Frequency Sequence)

18. 18 The performance quoted in the previous slide assumes the use of 5 RF bands, each producing data at 8 Gbits/s, resulting in a total data rate of 40 Gbits/s. Is it realistic to expect that by 2010 disk capacity and rate will reach the level of performance required for a 40 Gbit/s system?

19. 19 Disk and RAM Projections for 2010 First EpochSecond Epoch Projection for 2010 Disk Capacity* 2 Gbytes 1996 200 Gbytes 2003 20 Tbytes Disk Record Rate* 7 Mbytes/s 1997 42 Mbytes/s 2002 250 Mbytes/s Cost of RAM (per MByte) $10.89 1996 $0.16 2005 $0.0025 *Mujunen and Ritikari, 2004

20. 20 Implications for VLBI  The projected capacity for a 16 disk VLBI record system will handle  3.5 days of 20% duty cycle recording at 40 Gbits/s  Projected data rate for a 16 disk VLBI record system will be  32 Gbits/s  Because of the low cost of RAM, it will be possible to store data into a RAM buffer at 40 Gbits/s during 20% of the time when the antenna is on source and then write it to the disk at the slower rate of 8 Gbits/s during 80% of the time while the antenna is slewing

21. 21 Drive speed End to End Slew Times (s) (assuming 5s for accel. and decel.) No. of sources per day (assuming a 5s Integration Time) 360°/90° Drive 180°/180° Drive 360°/90° Drive 180°/180° Drive 3°/s125656641234 6°/s653512342160 9°/s452517282880

22. 22 First Steps to Achieving this Goal:  Start simulations immediately to verify benefits of significantly higher observation density  Use simulations to develop and test new observing strategies to make best use of the higher observation density.  Begin to prototype as soon as possible  Acquire high slew rate antennas with broad-band feeds  Develop a new frequency selective RF-to-baseband converter  Develop appropriate digital back ends and recorders  Verify that phase delay ambiguities can be resolved  Finally, test the benefits of increased observation density [NOTE: If new hardware is incompatible with existing correlators, test data can be processed with a software correlator.

23. 23 A Powerful Observing Strategy  For a global array, assuming each station can achieve 2880 observations per day, a simple scheduling algorithm will produce twice as many (2*2880=5760) source observations per day. src1 src2 src4 src3

24. 24 A Powerful Observing Strategy (cont’d) With 5760 observations per day, daily observations can be divided in the following way  50% of the time, observe the most stable sources (10 times daily)  288 sources can be observed every day  This establishes a firm connection to the ICRF  40% of the time, observe the next most stable sources (5 times daily)  460 sources can be observed each day  This monitors potential substitutes for the most stable list  10% of the time, monitor other sources (5 times daily)  115 sources can be observed each day  In this way, ~3500 sources can be monitored on a rotating basis each month, i.e. 12 times per year.

25. 25 A Promising Future  It has been demonstrated that it is reasonable to expect that by 2010 all the elements will be in place to operate a VLBI system capable of ~3000 observations per day at each site.  It has also been demonstrated that this can be accomplished without sacrificing other important features, e.g.  High delay measurement precision  Adequate sensitivity to detect all sources in the ICRF and VLBA Calibrators List.

26. 26 A Promising Future (cont’d)  Furthermore, all this can be accomplished using low- cost 12m antennas (although the assistance of larger dishes is beneficial for the weakest sources.)  The use of low-cost antennas opens up the possibility of a comparatively large dedicated network made up of 32 antennas for example, with as many a four antennas replaced on a rotating basis by larger existing antennas.

27. 27 Conclusions  A large increase in the number of observations per day at each site should be feasible by 2010.  This will bring many performance benefits including the potential for a unified observing strategy very powerful for a self-consistent treatment of the ICRF, ITRF and EOP