For contacting JIVE support scientists (for instance for help with schedule
files etc): ``jive@jive.nfra.nl''.
For distributing a message to other technical friends in the EVN (also
followed by hw/sw developers): ``evntech@jive.nfra.nl''. To
subscribe to this list send a message to
``evntech-request@jive.nfra.nl'' with the word ``subscribe''
in both the ``Subject:'' line and the message body.
For worldwide distribution of FS-related issues:
``pcfs@gemini.gsfc.nasa.gov''. To subscribe, send an informal
message to Nancy Vandenberg ``nrv@gemini.gsfc.nasa.gov''.
For issues related to the EVN Mark III-to-Mark IV upgrade:
``emu@jb.man.ac.uk''. For corresponding EVN VLBA-to-Mark IV upgrade:
``viv@jb.man.ac.uk''. To subscribe to either of these please send
an informal message to Ralph Spencer ``res@jb.man.ac.uk''.
The JIVE WWW home page is located at http://www.nfra.nl/jive.
Alessandro Orfei from Medicina asked me (Mujunen) about a couple of details of thin tape usage. I answered him and the message resulted in additional comments from Himwich and George Peck (NRAO) which I have consolidated below.
I have understood that prepassing is done just like for thick tape, running at high speed (270--330ips?) in forward, releasing vacuum and cleaning the recorder, and then running back in reverse. However, from reading thin tape development memos I get the feeling that initially melting of thin tape edges at high speeds was a problem. So perhaps it is safer to run the prepassing of the first thin tape compatibility test at 160ips only, and only after getting a verification from Socorro that no tape edge damage has occurred, only then switch over to using 270/320/330ips.
Please note that before we need recording a MkIV mode with 18Mbits/s/track, it will be highly unlikely that 320ips will be used in an experiment. VLBA/MkIV modes that use 9Mbits/s/track are recorded at 160ips and 4.5Mbits/s/track at 80ips. These are the modes the VLBA is using all the time, thus the VLBA thin tape ``success story'' does not really apply to 320ips recording, that is, nobody knows yet how well the tapes perform when routinely used at 320ips.
Himwich added:
I talked to Dan Smythe and he said that the thin tape upgrade is
supposed to have solved the edge melting problems (a good thing I
guess because Mark IV requires that it work). I also talked to George
Peck and he felt that there was no need to for a speed limit on
postpasses (and therefore I assume on prepasses as well). He was
worried about flange forcing at high speed, but checked and did not
find a problem. So I think this means that pre- and post-passes can be
done at 330 ips (``sff''/``srw'').
Mujunen adds: Please note that the VLBA also postpasses tapes after experiments. The goal is to smooth out any uneven packing caused by tape starting and stopping during the last tape pass. Unevenly packed tape is more prone to damage during shipment. Apparently postpassing is not required if the schedule didn't stop tape in the middle of the last tape pass.
We didn't adjust ours, but write voltage has always been a tricky thing. In principle flux cells at 56000 bits/transitions per inch are smaller and less write current (voltage) is required. (The same applies when the head wears out and tape comes closer to write coils and the air gap becomes larger.)
I have always wanted to get a nice table/range of voltages to use, but apparently the ``right'' way to adjust the voltage is to search for the maximum SNR in playback spectrum / minimum number of parity errors in playback. (Playback spectrum can be seen by connecting a low-frequency spectrum analyzer to the same connector you use for checking the eye pattern. When a good eye pattern is played pack, you'll see the signal level of range of 0..(bit rate / 2) rise.) I couldn't quickly find what SNR is considered ``good'', >23dB? (Can anybody confirm this?)
Please note that the same write voltage results in different write currents and thus different recording behavior in MkIII, VLBA, and MkIV. (That is, the write resistors in series of the write coils are slightly different in each of MkIII, VLBA, and MkIV.)
To search for the maximum SNR / minimum parity errors maybe the most straightforward way is to set the voltage to quite a low value and make a test recording where the voltage is increased gradually, for example:
1000--2000 feet 8V (MkIV, maybe VLBA)
2000--3000 feet 9V
3 10V
4 11V
5 12V
6 13V
7 14V
(8 15V)
Then play back the tape and watch for the footage counter and changes in SNR in spectrum analyzer and parity error counters in MkIII decoder display. (Use MkIII mode A at 4 or 8 MHz and tape speed of 80ips or 160ips, correspondingly.) Use the voltage which gives best SNR and lowest parity errors.
Himwich added: One refinement on this, and it may not be an improvement, is that we have been encouraging the geodetic stations to use the lowest write voltage that gives the minimum parity errors. The error rate is sort of a ``bath-tub'' like profile against write voltage. It starts high, drops toward a low/minimum value, stays at that value a while and then rises again. Since the write voltages needs to be dropped as the heads wear, starting at the low voltage end of the bath-tub bottom gives the longest time between adjustments (and most safety I suppose).
George Peck (NRAO) added: We have found that the write voltage does differ between thick and thin. If the write voltage is too high, it especially effects recordings where just half of the tracks are written during a pass, and the other half of the tracks are written later. Crosstalk between heads can cause the first pass to be partially overwritten by the latter pass. For this reason, we usually check for crosstalk when we have chosen our write voltage. We do this by recording just the even or odd tracks, checking the error rates, and then recording the rest of the tracks in the same head position. Then, an error rate check is made on all tracks to be sure that there was no crosstalk to overwrite the first set of tracks which were written.
Mujunen concludes that by judging these messages it seems appropriate to set the write voltage to the lowest value which produces good parity errors on playback. This will minimize crosstalk and give the longest time before the voltage needs to be adjusted again due to head wear.
I find it easy to remember how thin and thick tape behave with respect to vacuum when I think of a wedge (the head) on top of which I'm stretching the tape with my hands. In this arrangement the vacuum level is represented by the force I'm exerting with my hands---higher vacuum corresponds to more force being used to stretching the tape which then tries to wrap more closely around the ``wedge'' of the head. Another ``mental image'' is to think that the tape is more like sandpaper and then see what would happen to the ``wedge'' (head).
If the tape is thin and thus flexible, it can follow the contours of the wedge/head more closely and thus the head gets a ``sharp'' countour. This is like grinding with flexible sandpaper.
If tape is thick and thus more rigid, it acts more like a flat grinding stone. It flattens out the countour of the head. There is more contact area with the head, i.e. a pointed, ``sharp'' head makes contact with the head mostly with the recording air gap area whereas a ``flattened'' head makes unnecessary contact with ``useless'' head area and SNR is reduced.
However, if you exert more force on the less flexible thick tape, you can make it embrace the ``wedge'' in a sharper angle. This is what we are after with 15in of vacuum for thick tape.
Now it is easy to see what happens with wrong settings.
First, if you forget 5in for thick tape, there is not enough tension to force the tape in pointed manner over the head, the tape makes contact only with the center of head, wears it out quickly (because the contact area is very narrow) and thus flattens the head again, much like if the tape were a flat grinding stone.
Second, if you accidentally use 15in for thin tape, the increased tension wraps the tape more pointedly over the head and the tape starts to slowly wear the sides of the head until there is a prominently high ``peak'' in the middle of the head in the recording air gap area. This actually makes the head better for thin tape recording as the head becomes more pointed. I don't know if this is outward dangerous (say, for the tapes?) but apparently the wear of head sides is faster than wear when the head has reached the ``natural'' contour for a particular thickness/vacuum combination.
From the above you can also deduce why Haystack and NRAO are concerned about head wear with vacuum switching: when you switch from 5in to 15in, apparently thick tape starts to form a slightly less pointed contour, wearing the center of the head. When you switch back from 15in to 5in, thin tape wears a little the sides of the head to get back the ``natural'', more pointed contour. To get precisely the same contour with thick tape as what one gets with thin tape at 5in, apparently more than 15in of vacuum would be required which then would wear the headstack more, as George Peck points out next.
George Peck (NRAO) added: The worst thing to do is to use the wrong thick tape vacuum, because the thick tape apparently changes the headstack contour more quickly than thin tape. We believe that it is also bad to run the thin tape at high vacuum. As Ari pointed out, this would cause a more pointed contour on the headstack. We have been receiving some brand new headstacks which are contoured in this way, and they might work well for a short time, but after the thin tape starts to flatten out the pointed contour, they do not work well until a new contour is established.
I don't know what vacuum all of you are planning to use for the thick tape. Our attempt at switching between thick and thin used 15 inches of water for thick, and 5 inches of water for thin. You might have better results than we did if you use a higher vacuum for thick tape. The disadvantage, of course, is that the high vacuum surely wears the headstack much more quickly.
Himwich: The existing *hdcal.snp schedules should work with thin tape, but I think this has already been discovered.
Yes, if you calculate (135ips / 80ips * 33333bpi) = 56250bpi which is
(about? does anybody remember the nominal value which would be the ``right''
one to use in FS ``bit_density'' command...?) the required bit density on
thin tape, then you can see that (56250bpi * 80ips) = 4.5Mbits/s.
(The 120ips figure is a historical misnomer: the button '120' on the
MkIII recorder has actually always delivered 135ips, as the internal tacho
generator was adjusted a little higher.)
Himwich added:
56250 for Mark III/IV format is correct. For VLBA format the density
is 56700. The difference is due to the fact that the VLBA format is
NDR and splices header bits in without losing data bits. The
``bit_density'' command thinks in units of total bits on the tape
including parity and headers as opposed to only data bits. The speeds
for both Mark III and VLBA formats on thin tape are 80ips and
160ips. This is big improvement over thick tape where not only were
the bit densities different, but the speeds were as well.
A note though: some correlators, but not the VLBA, can handle low density on thin tape, but this is used rarely and I would it expect to all but disappear.
equip.ctl file for thin tape?I'm not sure which parts of the system use this value, I vaguely remember that the VME control computer of the VLBA drive uses it to estimate the position on tape by comparing the relative speeds of reels. We changed it from 268 kA (thick) to 160 kA (thin), can anyone verify if 160 is the correct nominal value?
Himwich added:
We are advertising 152 as the value, but this only needs to be set for
VLBA drives (in equip.ctl). Mark IV drives don't care.
George Peck (NRAO) added: The VLBA uses 153 for thin tape. It probably wouldn't matter if the VLBA uses 153 and everyone else uses 152. This value is used by the firmware to calculate the values for the takeup and supply pack detectors. I'm not sure, but it may affect the low tape point.
FS provides vacuum switching support starting from release 9.3.13.
The recommended version is 9.3.15 or later, since the ``drudg''
component of FS9.3.13 had a bug potentially affecting collecting Tsys
data.
Orfei wrote: A note about our cut-out threshold: it is not possible to set below 4" H2O and when a switching is done from high to low a slight different vacuum level respect to 5" is set unless the vacuum door is open and closed again.
And Mujunen added: Yes, we experience similar effects. The cut-out threshold depends on fine-tuning the pressure switch as described in thin tape upgrade documents and I believe Michael set ours too to about 3.5--4in. This far the switch hasn't activated falsely at 5in, so I think 4in is probably ok. (I'd expect that at 3--4in the tape hardly will run at all.)
Our vacuum motor also likes some ``warm-up'' time at low vacuums, that is, after loading the tape at 5in vacuum gradually increases from about 4.4 to 5.3in during 2--3 minutes. (The vacuum gauge isn't apparently any real precision instrument either, try tapping it slightly with a finger... :-)
Himwich added: The tape drives should be warmed up for a few hours before use anyway, but maybe now we should include a suggestion to let them warm up with a tape mounted (and preferably at the vacuum that will be used for drives that switch).
George Peck (NRAO) added: We also have noticed that the vacuum pressure rises during the first few minutes after the tape is loaded. For this reason, we try to set our levels when the motor is warm, but we haven't ever had any problem with loading the tape and moving tape immediately after the load.