Mercury IPS
The Oxford Instruments Mercury IPS is a superconducting magnet power supply. It is the successor to the IPS. Much of the information on the IPS wiki page also applies to the Mercury IPS.
SCPI Protocol
The IPS IOC now supports the SCPI protocol, which is more feature rich than Legacy mode.
SCPI mode keeps the same PV interface used by the older IPS units wherever possible, to minimise the changes required to instrument python scripts when swapping between old & new controllers. The SNL state machine logic is identical between SCPI & legacy mode.
As already mentioned, SCPI mode provides additional status reporting, much of which is based on the
return string from the READ:SYS:ALRM
command, which is poorly documented in the supplier’s
documentation. The status strings are assumed to all conform to the “Directory of Alarms” section
(17.3) of the Operator’s Manual (Issue 20, July 2018).
Whilst many of the system alarms that we could test, mostly conformed, some differences were
noticed, along with additional, undocumented status, such as “Magnet Safety”.
It has not been possible to test all alarm scenarios with the IPS unit. Some messages are
undocumented and were ‘discovered’. Our best guess is that the IPS Manual presents some level of
truth, but as such it has not been possible to fully ascertain that all the expected message
strings are correct - they may need to be adjusted later on, if/when they arise.
Tested and verified to date:
PSU board open circuit
PSU board short circuit
Temperature board open circuit
Temperature board short circuit
The support module exports an aSub record subroutine to facilitate handling of the responses to
READ:SYS:ALRM
, which is not feasible with a StreamDevice protocol handler.
See the section on System Alarms below for more details.
CONTROL and CONTROL:SP
These records have been removed from the SCPI variant database as the SCPI command set does not manage the panel lock in the same was as legacy and Magnet Group advised to remove this feature, as they always want to be able to control the IPS via the front panel.
SYSTEM:HWFAULT status is derived from the status bits via a calc record.
This collates the various possible hardware faults into a single record,
which in the legacy protocol is a single value (Xm
bit 4).
SWEEPMODE:PARAMS and SWEEPMODE:SWEEP
These records have been removed as they are meaningless and underivable in SCPI protocol.
_SWEEPMODE:SWEEP
In the legacy version, this used to be the readback from n
of Mmn
part of Examine command
return.
0 output constant
1, 2, 3 output changing
The SCPI protocol doesn’t directly offer this, so it has to be derived via the ACTIVITY
record
(DEV:GRPZ:PSU:ACTN
)
Direction from Alex Jones: “A response of HOLD
or CLMP
would be equivalent to n=0
in the
response of the X
command in the old protocol. Responses of RTOS
or RTOZ
would be equivalent to
n=1
.
There is no equivalent for the m=0,1
fast/slow ramps, but we do not use this feature anyway.”
System Alarms
System alarms are derived by interrogating the IPS with the READ:SYS:ALRM?
SCPI command.
This returns a comma-separated string of active alarms, or an empty string if no alarms are present.
The string is parsed by an aSub record, which sets the relevant bits in a binary register.
The bits are mapped to individual status records, which are then used to set the relevant alarms.
The precise protocol was determined empirically by sending the command and observing the response
directly from the instrument.
Discrete records are generated for each board type and each alarm by the
scpi_system_alarms_discrete.template
and substitutions files.
The base response is: READ:SYS:ALRM:
which may (or may not) be followed by a board identifier,
a tab character (9) an alarm string and a semicolon.
In ABNF:
response = "READ:SYS:ALRM:" *(error)
error = board_id TAB error_message SEMICOLON
board_id = 1*(ALPHA / DIGIT / ".") ; e.g. MB1.T1
error_message = 1*(ALPHA / DIGIT / ".") ; e.g. "Open circuit"
TAB = %x09 ; tab character
SEMICOLON = %x3B ; semicolon ";"
The board identifiers are provided as macros:
Macro |
Default |
---|---|
BOARDID_MAG |
MB1.T1 |
BOARDID_10TMAG |
DB8.T1 |
BOARDID_PRESS |
DB5.P1 |
BOARDID_LEVEL |
DB1.L1 |
Development Notes:
Alex Jones has looked at some of the differences between the SCPI and legacy command set and has
summarised some useful information, as quoted below:
For the quench and overheat status, these (and many other issues) are dealt with by “Alarms” for
the SCPI protocol. See the Magnet board section on p. 162 in Issue 20. These are read by
READ:DEV::PSU:STAT
in some undefined hex format for just the magnet-related alarms, and
READ:SYS:ALRM
for a list of everything.
Some further information re UID naming:
The UIDs can be either the nickname for the card (which we can set to anything we want) or related
to the slot number and signal type. From the spreadsheet, the positional UID for the magnet
temperature sensor will be MB1.T1
, the UID for the level meter will be DB1.L1
and the UID for
the magnet supply will be GRPZ
for all 4 of the systems.
The 10T system will have an additional temperature sensor DB8.T1
and a pressure sensor DB5.P1
.
Devices connected to the motherboard are prefixed: MB1
Devices connected to a daughter-board are prefixed: DB<slot #>
A comparison of availability of legacy vs SCPI features
Function |
Legacy |
SCPI |
---|---|---|
Trip current readback |
|
Not available |
Trip field readback |
|
Not available |
Ramp mode reporting (fast/slow) |
Yes |
Not available |
Status reporting |
Yes |
Detailed |
He Level reporting |
No |
Yes |
N2 Level reporting |
No |
Yes |
Pressure reporting |
No |
Yes |
Front panel control lock |
Yes |
No |
PSU Status:
Using the SCPI command: READ:DEV:<UID>:PSU:STAT
It is important to note that the STATus word should be examined at the group device level,
not the individual PSU level.
It is also very important to mask out (ignore) all other bits in this 32-bit word (i.e. ones
not defined in the list given below):\
Status |
Bit value |
---|---|
Switch Heater Mismatch |
00000001 |
Over Temperature [Sense Resistor] |
00000004 |
Over Temperature [Rundown Resistors] |
00000002 |
Over Temperature [PCB] |
00000008 |
Calibration Failure |
00000010 |
MSP430 Firmware Error |
00000020 |
Rundown Resistors Failed |
00000040 |
MSP430 RS-485 Failure |
00000080 |
Quench detected |
00000100 |
Catch detected |
00000200 |
Over Temperature [Sense Amplifier] |
00001000 |
Over Temperature [Amplifier 1] |
00002000 |
Over Temperature [Amplifier 2] |
00004000 |
PWM Cutoff |
00008000 |
Voltage ADC error |
00010000 |
Current ADC error |
00020000 |
Compromises with SCPI command set and EPICS
STS:SYSTEM:LIMIT
Is an mbbi
and has has 5 possible values:\
0: “Normal”
1: “On +ve V Limit”
2: “On -ve V Limit”
4: “Current too -ve”
8: “Current too +ve”
The limit flags came from the legacy command: Xmn
with the index denoted by the n
value.
SCPI does not provide this information.
IOC Test Framework:
With support for the new SCPI based IPS command set, there are now two sets of StreamDevice
protocols. The appropriate protocol is implemented by use of a macro
(PROTOCOL
= SCPI
| LEGACY
) defined prior to running the IOC.
The test framework has been adapted by splitting the existing legacy tests into common tests and tests specific to either control interface. For instance, the legacy command set knows nothing about cryogen levels, which the SCPI command set does.
The IOC support provides two protocol files:
Legacy: OxInstIPS.protocol
SCPI: OxInstIPS_SCPI.protocol
The lewis emulator and IOC test framework for the Mercury IPS is located within the IPS support module. Manually running the tests is achievable using the following approach:
cd
toC:\Instrument\Apps\EPICS\support\IPS\master\system_tests
run_tests.bat -t ips_scpi -a -f
- this will run all the SCPI specific testsrun_tests.bat -t ips -a -f
- this will run all the legacy specific testsTo run a specific test, use something like:
run_tests.bat -t ips_scpi.test_WHEN_inductance_set_via_backdoor_THEN_value_in_ioc_updates_0__0_12345 -a -f
Note that the -a flag simply prompts as to whether to run tests or simply run the emulator.
The -f flag forces a fast fail on error and no further tests are run.