Summary of discussions with V. Barnes during March 2001 CMS week.
Posted on the Web: 2001.03.14
Source of all misunderstandings and errors is
the scribe.
HCAL Source Calibration System
There is a number of sources in use in CDF experiment at FNAL
and different test facilities. They are either cesium 137 or cobalt 60
sources of varying intensity.
The proposed intensity for CMS HB/HE calibration is 2 mCi, and for
HF calibration is 10 mCi (although the HF team seems to prefer a source as
strong as 20 mCi).
The foreseen speed of the source displacement is 10 cm/sec. With this
speed the scanning time per calorimeter segment is not too long and
reasonable photostatistics is expected per data point. Of course it is
always possible to decrease the speed.
The number of source drivers per HCAL subsystem is: 6 for HE, 4 for HB,
and 4 for HF.
One PC can control up to eight source drivers. Each driver is controlled
by 8 bits sent through PCI TTL I/O card (max 8 cards) to the corresponding
Control box.
There are three distinct tasks for the control of the source system:
- Program and control source movements;
- Read out the electronics and
- Collate the source position with the readout values
These tasks are clearly visible in
a scheme designed
during March 2001 CMS week brainstorming session.
The programming of the source driver has to:
- define which tubes (calorimeter segments) should be scanned.
There should be an option to do scans in a given range
of tubes.
It is prudent to also load the foreseen length of each tube in order
to stop the wire propagation when it reaches the end of a megatile.
- position the indexer bar (max 110 tubes) into a proper position.
The indexer position is read out by legend counters and it is possible
that they give a wrong value. The safe way around it is to move
the bar to its limit positions and make sure that the values are
correct.
- control the start/stop/reverse of the source propagation through the
calorimeter.
The Control box is actually sending control signals
to the piezo valves in the driver that control flow of compressed
nitrogen.
- monitor the reel counter.
Reel position (0.1 mm precision) and the indexer placement
are sent to the controlling PC through RS 485 line (through up to
eight comm. ports). These readouts happen with 12 Hz frequency; max
frequency is 15 Hz.
- define status bits (e.g. alarm in case of stuck wire).
- send these data to the "Event builder".
Tasks of the read out electronics:
- HTR cards read the data from HPDs at 40 MHz and stream them to the
in-crate processor.
- Depending on the data quality and the power of the processor (both to
be determined) either an averaging for a certain number of time frames
or histogramming should be performed.
- Sending the data to the "Event builder".
With the wire speed of 10 cm/s, one millimetre precision can
be obtained if the processor sends data with 100 Hz frequency.
The data flow is therefore limited to few Kbits/sec.
To keep in mind:
It would be very useful if we read out only a limited number of HPD
channels to
reduce the data flow through DAQ/DCS system and burden on the in-crate
processor. On the other hand, it is useful to monitor another
channel(s) in order to iron out possible pedestal shifts. We should
also consider a possibility that the indexer sent the wire into a
wrong tube. Final answer still has to be found, after checking the
quality of the 40 MHz signal with QIE based electronics chain.
We should be able to program which channel is going to be read out,
to switch corresponding HTR card to "data streaming" mode and program
the in-crate processor to do whatever operations appear necessary.
Tasks of the "Event builder":
The "event builder" is a place where the source position and the electronics
readout are put together and a strip chart is generated. In CDF this is a
separate PC. For CMS application it might physically be in the PC
that controls the source, or the DCS PC where the PVSS can deal with
trending and alarms. It should be noted that the DCS PC should only have
"Event" and "Database" managers running and avoid overloading it with a
need to deal with graphics. GUI should be done elsewhere.
Synchronization:
The proposed mechanism of data transfer between different machines is
through Distributed Information Manager (DIM). As it is a client-server
program, in case of fluctuating network load, it might not be able to deliver
all the data with the same delay. It is proposed to attach time stamps to
each data point sent by either in-crate processor or the source controlling
PC.
As there several times more data-points than the source positions, the
data are usually presented as a function of the event number.