Frequent Asked Questions
Q: What is the main design goal of the MRO recording system?
A: The objective was to develop a low-cost replacement for the old Mk4 and VLBA recorders using off-the-shelf components.
Q: What is the status of the project ?
A: At this moment (end of April, 2002) we are ready to start volume production. Three prototypes of the VSIB board and one prototype of VSIC board have been tested and seem to be working perfectly well.
Q: What is the data format used in the MRO recording system?
A: The MRO recording system is designed to capture raw sampler data and store it into normal Linux files. The data is partitioned into files so that each file contains one second of data and starts at the sample coincident to the hydrogen maser 1PPS marker pulse.
Q: How does the MRO recording system capture data ?
A: The principle of operation is very simple. When the VSIB board gets the "start" command, it waits until the next 1PPS marker. When the 1PPS marker is found, the board starts to store data into a huge ring buffer (typically 256MB-512MB) in the computer main memory. The main processor only has to store the data that appears in the ring buffer to the hard disks, there are no time-critical tasks.
Q: How do we reduce the data rate, if we want to save disk space?
A: The MRO recording system supports the following ways to reduce recorded bandwidth:
- Reduce sampling rate from 32 MHz to 16, 8, 4, 2....
- Use only one bit for each sample instead of two.
- Reduce the number of recorded sidebands. Instead of recording all the sidebands record only the eight USBs or the first four USBs.
These are independent, it is possible to choose any combination of the above.
Q: Why record sampler data instead of formatted data in the Mark5 fashion?
A: Formatting and deformatting are low-level functions associated to the tape recording. If the tape recorder limitations (high error rate and 4-8 Mbit/s speed per track limitations) are removed, formatting serves no useful purpose.
Q: Are there special hardware requirements (motherboard type etc.)?
A: It is better to use a modern motherboard to get the full 512 Mbit/s speed. We have tested several motherboards and got the best results using an low-cost Asus board with the new Nvidia chipset. The expensive server-type motherboards were slower, because they do not allow one expansion board to reserve all the PCI bus bandwidth.
Q: Is the MRO recorder less expensive than the Mark5 ?
A: We estimate that the production costs of the VSIB input board and the VSI converter board are 500 euros each. The cost of 512 Mbit/s system would be 1000 euros (plus one office PC and disks) and the cost of a gigabit recorder would be 1500 euros (plus two office PCs and disks).
However, we would like to point out that the most important benefit ouf our recording system is not the low cost, although it makes the rapid deployment of the system possible. Standard file formats, standard hardware, expandability and a clear growth path are much more important in the long run.
Q: The maximum speed of one MRO recorder unit is 512 Mbit/s compared to the Mark5 1Gbit/s, why ?
A: There are several constraints in the PC technology. The most important is the speed of the disk subsystem, which is around 500-600 Mbit/s. The PCI bus and the Gigabit Ethernet both limit the speed of one unit to less than one gigabit per second, so we thought it's better to divide the load to two computers.
Q: Is it possible to convert the design to 64-bit, 66MHz PCI to improve throuhput?
A: Yes. The PLXtech PCI9656 chip uses the same local bus and programming interface as the old PCI9054 we have used. Changing the PCI bus interface impacts only a small part of the VSIB interface board.
Q: Is the MRO recorder compatible with the Mark 5A?
A: Yes, we have designed hardware support for the Mark5A recording modes and can record data in all the Mark5 formats. Data can be recorded with the MRO recorder, transferred to the correlator via Internet and played back with the Mark5 playback unit. Example files of Mark5 compatible data are available at ftp://www.metsahovi.fi/pub/temp/vsib
However, we cannot produce Mark5 compatible hard disks, because the Boulder Instruments data format is proprietary and undocumented.
Q: What is the relation of the MRO recording system to the PCEVN project?
A: The PCEVN project uses four computers equipped with the VSIB input board and a data distributor that is currently being developed in JIVE to achieve gigabit per second recording speeds. However, the MRO recorders can be used for other purposes, for example to record Mark5 compatible data with the VSIC data converter.
Q: What is the role of the "Metsahovi Milking Machine" alias the VSI converter?
A: Originally the "Milking Machine" started as a VSI test pattern generator with two input connectors to get real data from the VLBA samplers. Our original idea was to build a Torun-type autocorrelator without having to build a correlator board. In the design process we noticed that almost all the signals used in VLBI use differential signaling and it's possible to build a universal "two 40-pin flat cable connectors-to-VSI" converter. When we noticed that people have strong opinions about retaining the old Mark4 formatters, we added firmware to convert Mark4 IOB, Mark4 WDB or VLBA formatter output signals to VSI.
Q: Is the MRO recording system compatible with the Japanese gigabit VLBI ?
A. Yes, two MRO recording units can be directly connected to the Japanese ADS-1000 sampler. In this configuration, the MRO recorders can record two hours of continuous gigabit data.