On Thu, 19 Jan 2006, Josh Stein wrote:
In the undulator hall of the LCLS we expect to monitor (what I
consider) a large number of temperature points (on the order of about
300). My past experience with thermocouples has been local installations
of between one and eight data points. It would be easy to just scale
that solution by a factor of 40, but I have to believe there are more
cost effective solutions.
Josh--
The DFELL has a large number of temperature sensors: nearly 600, including
those which we are currently installing as part of the new booster
installation.
The main technologies in use:
"1-wire" network with DS1820 sensors and a 1-wire to RS-232 ASCII host
adaptor ( http://embeddeddatasystems.com/page/EDS/PROD/HA/HA5 ). Our EPICS
interface is using devAscii/devSerial to talk to the host adaptor which
handles the 1-wire communication. The sensor cost is low and you can
connect up to 200 sensors per host-adaptor/serial-port. But the 1-wire
network can be VERY fussy (sensitive to ringing and reflected signals)
(and it actually requires 3 wires: power, signal and ground). Once we got
it working, it has been fairly stable, but it took a fair bit of effort to
get it there. My predecessor here purchased about 500 sensors and we
currently have about 450 in use divided over 4 host adaptors. Measurement
takes about 0.75 seconds per sensor per host adaptor so an effective
update rate for 100 sensors on a single host adaptor is about 80 seconds.
Resolution is 0.5 deg C. with accuracy of a few degrees.
LM35 ( http://www.national.com/pf/LM/LM35.html ) These are cheap and easy
with a linear analog signal of 10 mV/deg C. Just put 5 V in and connect
the output to any ADC for EPICS interface. Our only difficulty was to
decided how to mount and cable. We ended up using a TO-220 with a
transistor socket and a piece of perf board with leads soldered to it and
heat shrinked. We are just putting up to 100 of these into service so
don't have any long term evalution but they seem to be good.
RTDs. We have a few of these for high resolution, high accuracy, critical
sensors. The RTD itself is moderately priced, but the
transducer/transmitter can be expensive. (You can build your own for much
less then they are available commercially, but for the few we needed,
decided it was easier to buy.) The transmitter gives you 4-20 mA, 1-5 V,
or 1-10 V linear to temperature, so an EPICS supported ADC is all you need
to interface. Good stability and reliable with good accuracy and
resolution for our range of interest (20-100 C) as compared to a TC. (A
thermistor could also have been used instead of an RTD.)
Let me know if you'd like more info on any of these or a few other
approaches I've looked in to. I've done far too much research into
temperature sensors <g>.
Regards,
