Message 17/486 Michael Wunderlich                        Oct 16, 2000 08:17:20
am +0200

X-Sender: p655miw@mail.mpifr-bonn.mpg.de
Date: Mon, 16 Oct 2000 08:17:20 +0200
To: evntech@jb.man.ac.uk
Subject: EVNtech: Upgrade report Urumqi and Shanghai
Sender: owner-evntech@jb.man.ac.uk
Precedence: bulk                             

China report 2000


Mar. 27: Arrival at Nanshan station. The toolboxes were not there due to
customs problems. They were sent three weeks in advance to avoid things
like that. I examined the JIVE delivery. At first glance there were no
transport damages. I gave a short schedule for the next two weeks.
Documented original cabling and power distribution of the "MK IIIB".
Took photos of the "before" state. With chinese help stripped down
the recorder with chinese tools. I was assured, that toolboxes will
arrive later that day.

Mar. 28: Improvisation was neccessary, but everything I need was in the
toolboxes, which had not arrived yet.
MK3 power supply, NIM crate for head positioner and NIM crate
for MK4 read/write electronics did not fit into the recorder
front. The chinese NIM crate was homemade and had to be reworked
completely to fit the MK3 head postitoner. The new NIM crate
was attached at the recorder rear, the back door and one panel
had to be reworked.
The fan is now at the recorder outside (VLBA-like), to have space
for all the cables. MK4 connector plate and cables were put in place,
vacuum switching was implemented

Mar. 29: Did the modifications on the Honeywell capsten servo
and the tape sensors. No tests for these yet.
Around lunchtime the toolboxes arrived. They have been inspected
very thorougly, but everything was complete.
Hose of the vacuum motor was leaking, had to repair it. Modified
the chinese NIM crate, until everything fitted in. Ground wires
from all the crates to the recorder frame.

Mar. 30: Put in all the cables for power supplies, MK4 R/W and head.
New MK IV I/O board in place and all cables attached. Chinese built
a new I/O panel for the connection to the MK IV formatter.
Put in new tape path with reworked capstan motor and new damper roller.
Unpacked the new head electronic. It was in the toolbox, because at
delivery time in January it was not complete.
One pin of the head connector plug was bent and broke off when I tried
to correct this. I could have used the electronics for Shanghai (which
was O.K.) or yelled for spare parts. I took the head interface used
in the original Urumqi setup, removed the resistors 1K and added the
resistors 2K2 from the new head interface.

Mar. 31: Resoldered head interface and pre-amplifier boards. The
Chinese improvised a workbench with some sort of magnifier and
a light. That worked, but it took several hourse and produced
lots of headache.
Mounted the new (second) head block and attached the cables.
I tried to attach the new headstack, but was not able to
adjust it due to weary eyes.

Apr. 2: Recorder was switched on for the first time.  Inchworm
driver doesn't work due to defective power cable, which was replaced.
Put junk tape on recorder (without heads) to adjust vacuum switch.
That worked. Ajdusted the tape sensors and tried to run the tape.
Recorder dropped vacuum during acceleration. My idea was that the
gain of the control loop was too low. Checked the reel servo.
There were no connectors between motor ground and power ground.
Added them. No change in behaviour.

Apr. 3: Checked the Honeywell logic board. Some of the modifications
had been made earlier, the others were made now. After two hours
the board was complete and didn't work. At closer look I noticed that
the capstan's acceleration was extremely high (with the correct
direction), so that the tape lost vacuum instantly. Checked the
Honeywell board again, all modifications correct, no change.
Removed the board (Urumqi) and put in spare board from Effelsberg.
Big surprise: no change, still didn't work.
Assumed that both boards were O.K. Suspected the capstan servo board.
Reference pulse is there. When servo is enabled with STOP, motor
instantly runs up to high velocity ( no ramping), vaccum is lost.
Suspected that modifications damaged capstan servo board (the integrator),
that there was no feedback.

Apr. 4: Changed components in Capstan servo as far as possible.
No change in behaviour of integrator.
Rest of Team China 2 showed up around lunchtime. Gave a short
report of the status. We decided that Les should help with the
capstan. Studied the schematics for two more hours, then found
out, that the feedback could have the wrong polarity. We discussed
the several versions of this board and the jumpers to change
the signal phase. My impression was that there were no jumpers,
but Les found them instantly. I didn't recognize them when I did
the modifications. Put them in the correct place, transport worked.
Adjusted the reel motors and ran thin tape. Tape path looked good,
tape packing was fine.

Apr. 5: Assembled the complete head assembly. Head 2 (old): Metrum
K46 stepped head, Head 1 (new): SpinPhysics 47581 triple cap.
Adjusted heads, mounted the electronics. Head 1 came loose due
to wrong screw. Had to do it all over again. Mounted the head
assembly and tried to move the heads.
Head 1 won't move. Swapped inchworm cables. Head 2 won't move.
Tested inchworm driver on the bench. Fuse blown. Opened the
housing and replaced fuse. PC boards full of dust, typical
symptom for arcing. Switched on again, arcing between transformer
housing and driver 2.
Inchworm driver was manufactured by Signatron. All the cables
on the rectifier board were not soldered properly. The space between
transformer housing and p.c.board was less than 0.5 mm. Put in
an insulator plate, but FET's were already damaged.
Ordered spare parts for Shanghai to repair the inchworm driver there.
(Ask Gino for the relay race!). Agreed with Ed to fix the software
to run with one head only (the read/write head).
So we made the MK 3 B a new MK 4 B.
Build a 'wet air kit' to profile the new head over the night.

Apr. 6: The tape lost vacuum last night. One transistor for the
takeup reel was defective. There were no spare transistors.
I took the driver transistor (same make) to replace the defective
one and replaced the driver with a funny chinese transistor.
Transport worked again and shutteled the tape for several hours.
Connected the recorder to the field system computer. Head MAT
reports 'Head moving' and head wasn't. Recorder MAT wouldn't talk.   
Replaced cables, no change. Put in Urumqi spare MAT, had to use
same address switch. No change. Put in Effelsberg spare.
Worked with short cable, didn't work with original cable.
Connected MAT to MK3 controller. Didn't work then. Swapped
MAT's between recorder and BBC. Both communicate. Swapped
MAT's again, address switch came loose in pieces. Seems that
it was the problem. Replaced switch, MAT worked.
Tried to find the head positioner problem. Replaced A/D board
with Effelsberg spare. Different problem then. Reported O.K. but
head was moving all the time. Finally we found that the 5 MHz
for the head positioner was missing. Connected cable, positioner
worked.
First attempt to calibrate the head with the old MK3 formatter.
Was a failure. There were lots of interference, eye pattern
hardly to be seen. Tried to fix this. Several attempts, no change.
Data was written, because MK3 decoder displayed time and date in
bypass mode.
Stopped calibration here and put on prerecorded tape. No proper
eye pattern visible due to massive interferences. Interference
was gone, when write cables were removed. Eye pattern visible now,
but not clean, like wrong equalizer or bad frequency response.
Parity error rate was high.

Apr. 7: Checked read electronics. Looked good, eye pattern is better
now. Removed MK3 formatter, put in MK4 formatter. Wouldn't talk.
Les found loose components before mounting. Checked all cables,
called Arno in Bonn.
After several hours of investigation and a lot of phone calls
it seemed that the RS232 chip on the single board computer was
damaged due to misplaced cables. We dediced that the formatter
was not usable and agreed with Arno to send spare boards to
Shanghai, so that Gino could replace them later.
We decided to put in the old MK3 formatter again and to write
down all the changes for Gino. MK3 formatter worked.
Connected it to the recorder and had interferences again, worse
than before.
Lots of noise all over the place. Tried to fix this with ground
wires, but no big change there.
Later that night we decided that the terminal was not usable as
it was at that moment.

Apr. 8: The interference in the read electronics could not be
removed. The reason for this was the old MK3 formatter. Its outputs
can not be disabled. The write voltage was switched off, but signals
kept 'leaking through'.
This was no problem in the old MK3 configuration,
where you have one head for reading and one head for writing.
Here we only had one head, both for reading and writing (due to
the defective inchworm driver), so there were always interferences
from the write data when you tried to read the tape.
This would work with the MK4 formatter (outputs can be disabled),
but that one was unusable.
So this configuration in principle never could work.
Found out this, we recabled for two heads and MK4 formatter and
hoped that we could repair at least the inchworm driver in Shanghai.

Apr. 10: Checked the JIVE delivery. Put in the new read/write modules
and cables. Repaired the Urumqi inchworm driver. Most of the FET's    
in the second driver were defective.
Tried to mount the new head block. Didn't work, because the metal ring
in the shaft was tilted. Had to use the Shanhai spare part.
Spare parts for MK4 formatter arrived in Shanghai, customs problems.

Apr. 11: Repaired headblock with spare inchworm. Tested both Shanghai
and Urumqi drivers. Mounted new headstack and calibrated it.
Put in new head electronics and mounted the complete head assembly.
Bad news about the formatter boards. They have been declared as
'computer' parts, so the process of getting them out of customs
would take at least three days.
Connected the MK4 power distributor to the VLBA recorder. Had to
replace one connector, which was designed for the old MK3 recorder.
With this module in the VLBA, the write voltage can not longer
be controlled by software, the VLBA style power supply would have
been better.
The MK4 formatter produced data.
We were not able to calibrate the recorder.
The head positioning problem is still there.
I replaced all the components in the analog conditioner and in
the VME transport module, there was no change. Checked all the
cables. Finally I activated the LVDT oscillator all the time.
It looks like a software bug in RECON, but a look into the
source code showed no reason for this. Ed made a software fix
to start the head calibration. It was impossible to write data
because data from the formatter didn't show up at the VME write
module.

Apr. 12: Still missing tracks. Reason was a bad connector in
the formatter cable. Replaced it, but still problems.
Gino had to leave for Urumqi, so we tested the formatter boards
we brought to Shanghai. They worked. There was a different problem
with the Urumqi formatter, bad for Gino.
We continued with the new MK4 formatter and found more bad
contaces in several connectors. We assume that they were oxidized.
After several plugs and unplugs they worked.
Did the head calibration twice with similar results.
During write/read test with thick tape the parity error rates were
high. Problems to reach 15 in H20 with thich tape. Reason still
unknown. High vacuum set to 13.
There was no eyepattern on the write drivers, data was decodable
in bypass. Several components were missing on the board, so that
the BNC output was not active (and never had been tested).

Apr. 13: Populated the missing components in the WRM's and there
was an eye pattern now. Checked the read module and found missing
capacitors. The frequency response and the gain was all wrong.
After completion the eye pattern was much better. Called Gino
in Urumqi. The same problems there.
Still high parity error rates and the eye pattern better, but not
completely O.K. The reason was that in VLBA recorders the sequence
for the three equalizers is different from the MK4. Rearranged the
equalizer acordingly. Eye pattern very good and error rates much
lower then. Shutteled a thick tape for nearly two hours, error
rates even lower, now acceptable.
Problems again with missing tracks during write. Read electronics
O.K. with prerecorded tape. Unmounted the head assembly and
checked the head interfaces (write only) again. While checking 
the connectors the probe broke off, so the connector was ruined.
Had to solder a new one.

Apr. 14: Tried to write data again. Still missing tracks, different
ones that before. Reason: connectors are designed for flat cable,
thin cables for write interfaces were soldered to them and had been
fastened with epoxy. Glue crawled into the connectors, so that
contact was intermittant. There was no time to fix this problem
completely. We decided to do the fringe test with missing tracks
to find out about the MK4 formatter. After the fringe test
we agreed with the Chinese to recable the recorder to VLBA and
send the defective parts back to Bonn for repair.


Recommendations:
Both for Urumqi and Shanghai: Keep a good set of spare parts, like transistors
For recorder and its components. Make sure that the documentation is up to
date.
Make sure that staff keeps track of developments, memos, etc.