PC-DAQ trouble-shooting
Y. Unno

This is for LBIC+CDP+HAC combination.  


1) HAC latch-up:  This is the single biggest source of trouble, since
each time the hybrid is powered off and on, the HAC goes on a latch-up
mode and need to be reset. Sending "reset control block" resets the
latch-up, but one sending may not always reset the HAC, several or
sending "test pattern" may reset it.  (A novice player tends to assume
one "reset" sending has reset the HAC... not true.)

Lesson 1: Don't belive the "reset", EVERYTIME confirm the Vdd current 
          that the HAC is not in the latch-up mode. 

2) Selection of calibration line: Depending on the hybrid assembly, 
not all calibration line is connected.  Most of the case, the cal.
line "0" is connected, but sometimes, the cal line "2"... (There 
are four cal. lines, "0", "1", "2", and "3", which can be selected
cal_0.cb, cal_1.cb, cal_2.cb, cal_3.cb control block data file, or
cal_all.cb for activating all calibration lines.)  This cal line
selection is done by loading and executing the FIFO1 (with the
cal. cb file in the window) in the "DSP card" panel.

Lesson 2: First, find the correct cal. line by trying each cal.
          selection. Watch the scope sensing the HAC singal lines
          in the 50-way cable between the bias card and the hybrid.

3) Instruction sequence:  Depending on the measurement to follow, a 
proper "control block" has to be sent to the HAC.  (A novice player
don't know what is the correct/proper "control block" nor the proper
sequence.)  Here is the sequence (if wrong, let me know):

   "DSP card" panel:

	FIFO3 - reset HAC latch-up, (see above)
	FIFO1 - calibration line selection 

   Then, "Detector plane" panel:

	Whatever the action you want, e.g., "Strobe delay scan",
        "Qinj", ...

4) Resetting the parameter:  Initially, most of the parameters are
read from the "detector definition" file.  When a "scan" is made, the
value of the "scanned" parameter is just the last value of the scan. 
When you proceed to a next measurement, you have to set the "parameter"
by hand to a proper value (or, start from the scratch, i.e., cancel
the program, and execute a new process.  Then, don't forget to confirm
the HAC latch-up.)  Cautions are after "Strobe delay scan", "Vref scan", 
...

Lesson 3: Check/confirm the parameter values before your next action

5) LL-card bottom "jumper".  This will adjust the phase of clock and the
HAC signals in the LL card.  This phase may vary with the temperature of
the LL card/chips, cable length, ...  If this is out of phase, you have
"0"'s in the data.

Lesson 4: Move the "jumper" location and confirm that the LL card is 
sampling the HAC signals correctly, e.g., by running "test pattern", 
run "readout" mode and check the read.dat for the proper activation of
"test pattern" channels.

6) LL-card temperature: The chips on this card become very hot.  It is
not clear whether a cooling (e.g., blowing air with a fan) is required...
It seems like the phase of the clock and the HAC signals is affected 
by the temperature and would be better to blow air.  (Without the
air, the jumper location was 5; with air the location gets be 3 or
4. At 5, no data.)

7) Software bug???  "Calibration" run kills the VME-DSP communication in
the KEK PC-DAQ system: 266 MHz Pentium II + NI VXI interface.  This is a
"soft" kill, i.e., the program does not stop and almost no indication of
trouble.  However, after this "crash", no run produces a correct data,
even the run which was successful previously.  Other runs, "Qinj",
"Strobe delay", ... are not affected.  (Very initially, the program
crashed (memory protection error) for any runs.  This was fixed by a
patch by Peter P. )

8) Detector bias: When the hybrid is connected to a detector, and 
channels are fired because of an offset to the hybrid input, there would
be a peculiar chip on-off pattern.  Put a small voltage to overcome the
offset. If noise/occupancy is a problem, you need to put a full bias to
make the channel quiet.

  (I think above are the essential troubles...) 

  If one of the above fails, the data is a failure, e.g., "0"'s.  (the
data is the "success" count out of the "events".)  It took long time
to sort out all combinations.  

					Y.Nobu UNNO
					Tsuyoshi ICHIKAWA