Message 6/432 Ari Mujunen Oct 6, 2000 09:03:06 pm +0300 Date: Fri, 6 Oct 2000 21:03:06 +0300 (EEST) Reply-To: Ari Mujunen To: evntech@jb.man.ac.uk Subject: EVNtech: Re: TOG Reports and Agenda / Metsahovi, FS Sender: owner-evntech@jb.man.ac.uk Precedence: bulk Hi all, On Thu, 28 Sep 2000, Mike Garrett wrote: > What I would like is for you to submit a short report to EVNtech > on each of these topics - at least 1 week before the meeting. Please find below short comments on behalf of Metsahovi Radio Observatory and me. > * ICN Programme > What developments are planned to improve reliaibility at the > stations > - Reports (Stations -*) No special actions to improve VLBI reliability in particular have been planned at Metsahovi. However, harmonization of continuum/solar/VLBI usage of our 22/37/43/86 GHz receivers and their cabling/connections is underway. Developments regarding our H-maser (a new maser? two? :-) and its monitoring (a new automatic multi-channel system "Clodi"?) are progressing at their usual (slow :-) pace. Jan Engelberg (our former operations engineer), Kirsi Karlamaa (our former VLBI friend), and Olli Koistinen (our former receiver engineer) have left Metsahovi, and we have only been able to hire a replacement to the post of operations engineer, M.Sc. (Mr.) Pekka Sjoman. Heavily understaffed with regards to continuous VLBI work we haven't regretted too much that the EVN has apparently completely dropped observing at 22GHz and above. > * Formatter Upgrade News (Smythe/Graham - *) > - current status Jouko Ritakari (thanks, Jouko!) has been able to almost complete the installation of MkIV formatter, he is currently trying to find information on how the "MkIV-->VLBA Recorder" patch panel is supposed to be wired / what ribbon cables to use... > - what still needs to be done ? The formatter needs to be tested, and it wouldn't hurt to check that the old VLBA formatter is still functional and that "tapping" IF signals with T-connections doesn't degrade too much the signal seen by the old VLBA samplers. A check that VLBA formatter's DQA module can still be used (also at the same time as the new MkIV formatter) would also be nice. > * 2-head Recording (Smythe/Himwich/Langevelde... - *) > - current status (incl station readiness) We have an old headstack with one dead track which we could use for testing... I believe we also have a spare inchworm motor and LVDT, possibly also the R/W electronics (head interface boards), Jouko knows this. > - what still needs to be done ? Nothing related to the 2nd head has been installed at Metsahovi. > * IF switchbox > > - report on initial investigations (Graham *) As you may remember, my investigations on commercially available switch unit options resulted in horrendously expensive and bulky, complicated devices. My current estimate is that a 3IF (6IF "low" and "high" bands) x 14--16 VCs non-blocking crosspoint IF switch would be best implemented by using a "tree" of IF power splitters to generate 14--16 copies of each input IF signal and then having another "tree" of small SMD GaAs 0--2GHz IF switches for each VC IF output which would select one of those 3/6 available "duplicated" IFs. IF amplifiers, miniature/SMD power splitters, and GaAs (even pin-diode) IF switches are all readily available from vendors like Mini-Circuits, Hittite, and many others. Hittite seems to have particulary attractive small, cheap, wideband GaAs switches with TTL control. To be economical, this device would have to fit on one "rack-size" PCB. I I (nor anyone else at Metsahovi currently) don't have enough expertise to design a multilayer SMD board with almost every connection needing to be 50 ohm striplines, and with special GND plane and guard trace requirements to meet crosstalk specs. To build the power splitter / IF switch combination using traditional metal can modules with connectors and using coaxial cable would result in precisely the same astronomical costs and huge size as what we have seen in commercially available units. Does anyone else have 2GHz 50ohm multilayer PCB (two spatially 90deg crossing tree structures!) experience? > - where do we go from here ? (ALL) Another, more sensible approach (suggested by Jouko) would be to permanently group VCs/BBCs into three blocks and have switching only at block level. (This could be even manual, moving three coax cables is not that bad compared to 14...) However, apparently the "low"/"high" bad requirements make this harder than what it seems initially etc etc. > * Future FS Developments for Astronomy (Himwich/Mujunen - *) These are being prepared separately in an another thread of discussion which I'm expecting to appear on EVNtech sooner or later. My only addition to this is that we need to decide the way in which the ADAM/NuDAM/Metrabyte plain ASCII at 9600,N,8--RS422 protocol gets integrated into the FS. I'm strongly suggesting that for odd I/O bits (like pcal on/off) we really, really, really, really adopt this "ADAM/NuDAM/Metrabyte plain ASCII" protocol. It's simple, you can build your own modules conforming to it, and you can buy commercially interoperable modules from multiple vendors all over the world (if you don't feel like building modules :-). A commercial I/O bit module costs about 200 Euros; I've a prototype of one which we perhaps could sell for 100 Euros---if I get around writing the software for it, this far for Metsahovi we have just bought commercial modules... Also, for this pcal application stations need to check if they really have this TTL ("open collector -- short to ground") input in their Pcal ground unit---the signal is sent as DC level in the same cable as 5MHz reference to the RX unit. We don't have a ground unit at Metsahovi---that means we'll have to build the minimal DC/5MHz part of Pcal ground unit before we can control the RX unit... > * Amplitude Calibration > - report on TPI proc test (Conway -*) > - move to VLBA style Calibration transfer (Garrett -*) This is handled in the EVN FS SW message, coming up separately... > * Logistics (Gunn -*) > - GPS At Metsahovi we need to replace our old H-maser/3xGPS/DCF77/TV/... 1pps clock difference monitoring system. The old one was provided to us by the local metrology agency in exchange for H-maser data. It is based on an obsolete ISA 1pps mux board, an old HP counter, an obsolete GPIB board, obsolete TSR DOS software, and an obsolete 386 PC running DOS. Several channels of the mux are already broken. What we want is a multichannel device capable of measuring/averaging every 1pps rising edge of every input channel. We have created a prototype of a single-channel counter which we will be expanding into an 6--8 channel rack-mounted device we decided to call "Clodi". 1pps clock difference data will be collected from this "Clodi" device by Linux software which we have decided to run on a separate (FS-like) PC (running FS Linux, i.e. the same Debian/GNU/Linux). This "station time management Linux PC" already runs Efos-9 H-maser internal parameter logging software and 'xntpd' network time (with low-res RS232 connection to a HP time-recovery GPS receiver). This is the system where we will realize daily GPS--H-maser log file FTP to correlator servers. We do have ordered a TAC-2 from 'http://www.cnssys.com/tac32/' so we may introduce redundancy for our HP GPS by interfacing our TAC to 'xntpd' network time server. "Clodi" will certainly include hardware for 1pps interfacing to a PC but I'm not promising 'xntpd' modifications to support this hardware... It may happen, but for now I'm completely satisfied with the 30ms or so 'xntpd' network time accuracy provided by the "stock" RS232 'xntpd'<-->HP GPS interface. > - fibre developments, pilot expts (Parsley -*) > - off the shelf VLBI DAS (Smythe -*) We at Metsahovi (at least Jouko and I) heartedly welcome this "COTS" direction of development of the next evolutionary step of VLBI hardware. Internally, we have dubbed various commercial technologies as viable alternatives to implement the next "Mark V" system. Our favorites include (but are not limited to :-) the following: 1) Off-the-shelf high capacity tape storage. (Or maybe disk storage, or both. Of course, in a time span of say, 5--6 years it tends to take to create a "Mark n" system I wouldn't be surprised to see fiber being offered in abundance and this would render all ways of recording of raw VLBI data completely obsolete.) 2) Focusing on selecting technology trends, chip families, bus solutions instead of particular computer models, chip types, etc. Those 5--6 years we need to develop systems like this will certainly obsolete _every_ particular component we would have used today, 4Q2000. With a little bit of unluck we may even see a demise of computer architectures / buses / technologies / chip families. 3) Fully exploiting the trends in consumer / wireless market segments. For instance, the LOs in the range of 100--2500MHz have an enormous selection of chips made primarily for low-power, low-cost cellular market at 800/900/1800/1900MHz. Many (but not all!) of these perform thru the full frequency range, making them very suitable for an IF band of 1--2GHz. 4) Experimenting with the best fit of "horizontal vs. vertical" partitioning of the system. This means partitioning into separate channels vs. partitioning along the signal path. The VLBA system (esp. the formatter) is an example of how extreme functional partitioning ("one step in signal path per board") leads into an excessive amount of bus interface electronics. On the other hand, in the current recorder the number of (identical) channels is very high (36 x #heads) which would badly need higher track capacity to reduce the # of channels. 5) Exploiting the developments in digital systems, primarily meaning the widespread use of efficient VHDL (or Verilog) synthesis tools of programmable logic devices which in turn are becoming bigger and faster. To effectively use these PLDs/FPGAs you absolutely need to use the latest high-level logic description language tools. (This can be compared to writing software in either assembly language or in a high-level language like C/C++/Perl/whatever.) 6) Implementing computing / high-level software functions with general-purpose computers, not with rare "one-time" processor boards. 7) Distributing power supply management across the system instead of huge centralized power supplies. 8) Migrating from ECL logic, differential ECL signalling, and coaxial cabling to LV-CMOS, LVDS, and (possibly shielded) twisted pair ribbon cabling. Many other items will certainly come to mind as time passes and the project progresses... In any case I'm really looking forward to rapid advances in this area. Best regards, Ari Mujunen Metsahovi Radio Observatory Helsinki University of Technology Finland