* forward declare zmq_msg_t and moved include
* removed zmq as dependency for slsdet (#870)
* Fixed path when building as submodule
* new moench data structure for offline processing
* meonch raw data and zmq process files updated to 7.0.3 version
* implemented config file for Zmq file
* raw data and zmq work with config file, but only with one file/interface
* zmq config change
* added config examples for zmq and rawdata
* update release notes, release versions
---------
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
Co-authored-by: froejdh_e <erik.frojdh@psi.ch>
Co-authored-by: Anna Bergamaschi <anna.bergamaschi@psi.ch>
Co-authored-by: anberga <34126852+anberga@users.noreply.github.com>
* if blocking and handling sync, only master gets blocking acq, slaves get non blocking as they are first and so dont get status or error caught when slaves dont get trigger (due to not connected etc) and acq returns with slaves still in waiting status. so check status of all in blocking acq
* for all dets with sync, ensure atleast one master when starting acq
* docs updated about sync
* fix virtual test when it fails
* catching errors in tests and removing sigchild ignore so servers (process in background) executing commands will not fail (pclose no child processes, if sigchld is ignored) fixed
* uncommented python loading config
* somehow killal slsReciever in second detector test fails even though no receiver running
* fixing script for virtual simlator test:fixed issue with check if process running, fixed moench tests
* fix acquisition finished status to have different status for different modules, but does not have to be error. for eg. jf sync fw (2.4.1 gives idle for master and stopped for slaves when stopping acquiistion)
* fix acquisition finished status to have different status for different modules, but does not have to be error. for eg. jf sync fw (2.4.1 gives idle for master and stopped for slaves when stopping acquiistion)
* jf: if bit 14 in reg 0x5d (electron mode collection bit) is changed, configure chip if v1.1 and powered on. so touch writeregister (setbit/clearbit also calls write register in the end). replace when electroncollectionmode command introduced
* rewrite of status reg bits, waiting state includes both wati for trigger and start frame, blocking trigger only waits if its not in waiting for trigger and run busy enabled, error state connected in firmware
* jf sync mode master could return idle when stopped and so not all modules return the same value and must check for 'stopped or idle', Also must throw if any of the module gives an error
* added contains_only to sls::Result (#827)
* added variadic template for checking if a result contains only specified values
* fix for gcc4.8
* renamed to Result::contains_only
* stop on only the positions
---------
Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com>
* jf sync mode master could return idle when stopped and so not all modules return the same value and must check for 'stopped or idle', Also must throw if any of the module gives an error ( not jungfrau, but eiger could)
* added contains_only to sls::Result (#827)
* added variadic template for checking if a result contains only specified values
* fix for gcc4.8
* renamed to Result::contains_only
* stop on only the positions
---------
Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com>
* getVoltageList, getVoltage /set, getMeasuredVoltage, getVoltageNames /set, getVoltageIndex moved to 'Power' as its misleading
* added cstdint and names slowadc, added division to mV
* changed uV to mV in command line slow adc help. removed all python slowadcs (as it was already implemented as slowadc
---------
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
* updated python bindings for port update from int to uint16_t
* user friendly error message for exception when python arg does not match uint16_t for ports
* merge fix from 7.0.2: new jungfrau fw versions, incremented binary, hdf5 and json versions
* moench: changed dac names and default values to old moench values
* moench: remove interface clk polarity at start up
* moench: default speed is half speed, default values for adc offset and adc phase for different speeds (only half speed confirmed), adc vref voltage to 2.0 like G1
* moench: connected adc pipeline to client
* moench: receiver- default frames per file is 100k and discard partial frames as default
* moench binary in
* using tostring in gui for dacs
* moved frame discard policy as a parameter to be configured with a default depending on detector
* moench: 300 degrees for adc phase in full speed
* silence warnings
* making constructors explicit to avoid unintended conversions
* changed struct to class since we already have public:
---------
Co-authored-by: Dhanya Thattil <dhanya.thattil@psi.ch>
* using fetchcontent to get zmq
* local copy of libzmq
* added guard for policy setting
* removed the need to export by using build interface
* moved pybind11 to FetchContent
* removed zmq hint from cmk script
* Fixed comments
---------
authored-by: froejdh_e <erik.frojdh@psi.ch>
* using fetchcontent to get zmq
* local copy of libzmq
* added guard for policy setting
* removed the need to export by using build interface
* removed zmq hint from cmk script
---------
Co-authored-by: Dhanya Thattil <dhanya.thattil@psi.ch>
Co-authored-by: froejdh_e <erik.frojdh@psi.ch>
* transceiverenable, tsamples, romode for tranceiver and digital_transceiver
* 202 spec instr only for transceiver mode
* removed check for empty in trans readout and clean memory before reading from fifo
* ctb read fifo strobe for all after reading all channels, adding 1us after selecting channel, changing fw date
* updated 10gb transceiver enable
----
* added transceiver (tsamples, romode(transceiver, digital_transceiver), transceiverenable (mask)
* clean memory before reading from fifo (for analog and digital as well)
* read fifo then read strobe (also corresp fw) fixes number of reads (also for analg and digital)-> increases all pipelines by 1
* fixed bug in rearranging digital data in receiver
* fixed bug in streaming size of data after rearranging
* fixed bug in setbit, clearbit,and getbit
* status checks fifo before returning idle (transmitting if data in fifo if transceiver more enabled)
* soem matterhorn specifics that will need to be put into pattern in a month or two. this is temporary.
* NOTE: breaking api. rxParameters struct has transceiverenabel and tsamples given from det to receiver
* when dbit list is enabled, the size of data in zmq stream is changed to only the digital bits enabled size. now fixed to also include analog size
* allowing to set 0xffffffffffffffff to pat io control. prevously was used to do a get. fixed also for pat bit mask and pat mask
- start acq list: mixup with master pos #743 : fix that only master starts second and not all (for start acq), typo with pos and masters list
- synced master status running when setting to slave #747: synced master status running when setting to slave
* jf: unsync before setting master/slave and then sync (if it was set) to overcome master going into running state when making it a slave and synced
* add tests for this condition
* updated release notes, updated min fw version requirement for v1.0 boards
* moench, removed chip version, filter resistor, filter cells, currentsoures, gain mode, setttings(modes), dbitphase, maxdbitphase, autocompdisable, comparatordisabletime, made acq start and stop a pulse, removed unused registers
* added parallel command
* remove gain plot for moench
* moench: updated adc invert val
* moench: update adcoffset to 0xf and adcphase to 140 degrees
* removed sync clock in moench
* updated min fw version
* removing config file in moench server
* Revert "Ctb: allow adc mask enable to be 0 for 1 and 10GbE (#750)"
This reverts commit a0f250a487.
* better error message about setting adc mask to 0. Cannot set it to 0 due to ram allocation
* using argparse for parsing command line arguments
* added command line option to specify which servers to run
---------
Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com>
adding rx_roi also in the zmq header for external guis to put the "yellow box".. sending full roi instead of -1, and sending for each zmq port. "(multiple yellow boxes)".
* merge fix from #721 PR (sync) 7.0.2.rc -> developer
* row and column for jungfrau mixed up
* multi module jungfrau sync must do slaves first then master for start acquisition and send software trigger, and master first and then slaves for stopacquisition
* non blocking to slaves first and only then blocking/nonblocking to the master for sending software trigger(jungfrau multi mod sync)
* fixed get/set timing jungfrau when sync enabled, getsync during blocking acquire (for trigger or stop) will get stuck as it should ask the stop server
* switching between 1 and 2 interfaces did not set gui/client zmq port properly. Resulted in dummy streaming forever. fixed
* formatting, refactoring: const & for positions, multi mod M3 stop first master first
* adding missing cstdint for gcc 13
* Refactoring handle sync out, handling synchronization also for softwaretrigger for m3, for start/sync/stop for g2/g1
---------
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
* fixed row and col for moench 2 interfaces
* fix moench getTiming and also allow moench to handle sync
---------
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
adding rx_roi also in the zmq header for external guis to put the "yellow box".. sending full roi instead of -1, and sending for each zmq port. "(multiple yellow boxes)".
* row and column for jungfrau mixed up
* multi module jungfrau sync must do slaves first then master for start acquisition and send software trigger, and master first and then slaves for stopacquisition
* non blocking to slaves first and only then blocking/nonblocking to the master for sending software trigger(jungfrau multi mod sync)
* fixed get/set timing jungfrau when sync enabled, getsync during blocking acquire (for trigger or stop) will get stuck as it should ask the stop server
* switching between 1 and 2 interfaces did not set gui/client zmq port properly. Resulted in dummy streaming forever. fixed
* formatting, refactoring: const & for positions, multi mod M3 stop first master first
* adding missing cstdint for gcc 13
* Refactoring handle sync out, handling synchronization also for softwaretrigger for m3, for start/sync/stop for g2/g1
---------
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
Firmware updated. spi moved to firmware. In Software, configuring, then a pulse to start, wait for done bit and convert the values read from a regiter.
* modified ZMQonline display for stripsel usage. modified dataprocessor to overwrite header.detspec1 with ROI in case of ROI.
* Removing changes to the slsReceiver, moved to branch jf_zeromq_display_rxroi
* rx_hostname and port combo to one, or hostname to all, or a vector of hostnames and ports, ignoring none or empty, then verifying no duplicates for the host port combo including from shared memory
* extracted function for rx_hostname (#694)
* c++14 revert
* unique hostname-port combo for port, hostname, rx_tcpport (#696)
* verify unique combo for rx_port as well
* check unique hostname-port combo also when setting control port, hostname, rx_hostname and rx_tcpport
---------
Co-authored-by: Erik Fröjdh <erik.frojdh@gmail.com>
Co-authored-by: Erik Frojdh <erik.frojdh@psi.ch>
* eiger: hardwareversion, fix firmware version unable to read version scenarios, check to see if febl, febr and beb have same fw version
* feb versions can be picked up only after feb initialization
stop should really stop even if receiver had crashed, so check rx status after sending stop; also ensuring restream in acquire happens only if thers a callback
* fix to access to shared memory that doesnt exist
* fix for freeing shm and then setting hostname from API
* exception error message moved to private function
* refactoring to avoid allocating intermediate string
* modified ZmqSocket to expose the SO_RCVBUF and SO_SENDBUF parameters. Modified DataStreamer.cpp to change the SENDBUF to 1MB when HWL is <10. Modified Datastreamer.cpp so that when HWL is changed, socket is unbind/rebind. Added rebind to ZmqSocket.cpp.
Changed slot UpdatePlot() in qdrawplot.h from privat tto public, added plot->UpdatePlot() after every axis range change, so the plots are updated immediatly and not at the next image.
Added onlinedisp_zmq program which connects to the receiver ZMQ port and show images and histos. Compiled against ROOT 6.22/02. Added examples files.
* added setbuffer size also for gui, moved hardcoded values to a macro, removed unnecessary return of ok or success, added actual zmq exception message to could not create sockets error
* zmq: changing buffer size done within hwm
---------
Co-authored-by: mozzanica <l_mozzanica@mpc2012.psi.ch>
Co-authored-by: Dhanya Thattil <dhanya.thattil@psi.ch>
* eiger: adding mask to read/write registers. useful for setting quad parameters as they might have different values for left and right fpga registers.
** fix quad position
* fix quad flipping
* formatting
* first tries with a process intead of thread for rx_arping
* Moving delete pointer of udp socket to stopReciever,so rx_arping can only be set when udp socket is closed
* refactoring and formatting
* unused variable processId
* ignore sigchild to prevent zombie from child processes being killed
* added none or 0 to unset bad channels
* free function split to get int array from string of arguments for badchannels
* missed a file
* allowing list for badchannels in command line
* added badchannels in python
* added size check
* more comments in Detector.h and added more tests for facny command line badchannels
* removeDuplicates accept any container, added tests
* corner cases: 1:5,6,7 or 5,6,7 or just 1:5
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
* copied jungfrau server to moench and adapted
* fixed image size and num packets
* read n rows allows 16
* commneted out configure_asic_timer at server startup. To be removed later the ASIC_CTRL_REG and storage cell options
* moench:removing the decrement (which was in jf) in read n rows to register
* removed lblsamples from gui
* loop for stop, 10 retries
For example, if module A got a "stop" during an acquisition, it stops to an 'Idle' state. If module B gets a "stop" before an acquisition, it will return it is in 'Idle' state and continue to start the next acquisition, but module B then waits for "ready for trigger" synchronization from module A, which it will never get. Since module B missed the asynchronous stop command, the workaround in the client is to send another "stop" command (upto 10 retries) till it returns Idle.
* added a lock when stopping or starting state machine in eiger server
* added 5ms right after acq started to make sure it really started
* added 100ms right after acq started to make sure it really started
- qt4->qt5
- in built qt5 6.1.5 because rhel7 is not upto date with qt5, removed findqwt.cmake
- made a fix in qwt lib (qwt_plot_layout.h) to work with 5.15 and lower versions for qrect constr.
- qt5 forms fixed, qt4 many hard coding forms switched to forms including qtabwidget, scrolls etc, fonts moved to forms
- docking option enabled by default, removed option to disable docking feature from "Mode"
- added qVersionResolve utility functions to handle compatibility before and after qt5.12
- qtplots (ian's code) takes in gain mode enable to set some settings within the class, with proper gain plot ticks
- ensure gain plots have no zooming of z axis in 2d and y axis in 1d
- removed placeholder text in qpalette in main window form as its not supportd until 5.12 (so using qt5.9 designer insted of qt5.15 to cope)
- tab order
Servers:
- fixed some error messages that were empty for fail in funcs (mostly minor as if this error, major issues)
- removed getClientServerAPIVersion in server (not used)
- removed rxr side (clientversion compatibility check), removed enum as well as it is now done on the client side.
- versionAPI.h
- GITBRANCH changed to RELEASE
- dates for all API changed to "sem_version date". Scripts to compile servers modified for this. Empty "branch" name will end up with developer for sem_version.
- Version class with constructor taking in the long version (APILIB date). Other member functions including concise(to get sem_version for new releases and date for old releases),
- bypassing initial tests, also now bypasses the client-rxr compatibility check (at rx_hostname command)
- previously, compatibility between client-det was ensuring both had the same detector API (eg. same APIJUNGFRAU)
- Now, compatibility only checks APILIB (client side) and detector API(eg. APIJUNGFRAU) (detector side) have same major version. It only does backward compatibility test. Rest is upto user to ensure.
- If server is from an older release, it will compare dates like previous implementation (APIJUNGFRAU from both client and det)
- - previously, compatibility between client-rxr was ensuring both had the same APIRECEIVER
- Now, compatibility only checks APILIB (client side) and APIRECEIVER (rxr side) have same major version. It only does backward compatibility test. Rest is upto user to ensure.
- If rxr is from an older release, it will compare dates like previous implementation (APIRECEIVER from both client and rxr)
- removed APIGUI, evalVersionVariables.sh, genVersionHeader.sh (not needed or not used)
- clientVersion, rxrversion and detectorserverversion all return strings and not integers (in hex) anymore. Depending if it has semantic versioning, it will print that or the date if it is too old.
- fixed in python (strings for versions)
- check_version function in detector server changed to "initial checks" as it only checks server-firmware compatibility and initial server checks. Client compatibilities are moved to client side.
- --version gives sem_version and date? Is date needed as well. The clientversion, detserverversion and rxrversion API gives only sem_version (no date)
- - formatting
- eiger server: fix for fw workaround where stop acquisition processing done signal does not come up, by removing reset in stop acquisition and waiting for2 seconds for feb done processing signal to go down, if it doesnt, throw if status is not idle.
- error messages not setup for some eiger server errors
- quad fix (chip signals to trim quad, both left and right registers can be different)
- minor logical error of no consequence (stop acquisition returns a different enum than expected)
- removed feb reset in stop acquisition as it caused processing bit to randomly not go high (leads to infinite loop waiting for it to go high). This is anyway done at prepare acquisition and set trimbits.
- left AND right registers monitored for processing bit done
- febProcessinginprogress returns STATUS_IDLE and not IDLE
- In feb stop acquisition, if processing bit is running forever, checks for 1 s, then if acq done bit is high, returns ok, else throws
- feb stop acquisition returns 1 if success and fucntion in list calling it compares properly instead of STATUS_IDLE (no effect, but incorrect logic)
- chipsignals to trimquad should only monitor right fpga (not both as it will throw)
- fixed error messages of readregister inconsistent values
- setmodule and read frame was returning fail without setting error messages (leading to broken tcp connection due to no error message)
-
* badchannels for m3 and modify for g2 (file from single and multi)
* m3: invert polarity of bit 7 and 11 signals from setmodule, allow commas in bad channel file
* badchannel file can take commas, colons and comments (also taking care of spaces at the end of channel numbers)
* tests 'badchannels' and 'Channel file reading' added, removing duplicates in badchannel list, defining macro for num counters in client side
* fix segfault when list from file is empty,
* fix tests assertion for ctbconfig (adding message) for c++11
* fixed badchannels in m3server (clocking in trimming)
* badchannel tests can be run from any folder (finds the file)
* initital implementation
* datetime replaces with sls::Duration in Python C bindings
* using custom type caster
* fix for conversion to seconds
* added set_count in python
* common header for pybind11 includes
authored-by: Erik Frojdh <erik.frojdh@psi.ch>
* g2: new hdi values, write hdi value to reg, set slave/master to reg, able to set master from server config file, server command line and client
* print versions for virtual as well
* udp socket refactor from reuss: closing socket before throwing for bind error in constructor, closing socket at destructor and not at shutdown to allow other thread to read with a -1, so the object can still be accessed
* check for size of packet for every detector
* nullptr to unique ptr to call its destructor, when deallocating resources moved out of shutdown
* minor
* incorrect dimensions for virtual hdf5 parameter set
* fix for corner case bug in hdf5 virtual parameter dataset when frames caught is not a multiple of framesperfile
* reafctoring for readability and error prone hard numbers
* gui message doesnt show if it has a '>' symbol in error msg
* minor refactoring for readability (size_t calc fifo size)
* refactoring listening udp socket code: activated and datastream dont create udp sockets anyway, rc<=- should be discarded in any case
* wip
* refactoring memory structure access
* wip: bugfix write header + data to binary
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* portRoi no roi effecto on progress
* fail at receiver progress, wip
* segfaults for char pointer in struct
* reference to header to get header and data
* refactoring
* use const defined for size of header of fifo
* updated release notes
* remove pointer in callback for sls_receiver_header pointer
* rx same name arguments in constructors
* rx: same name arguments in constructor
* rx: removing the '_' suffix in class data members
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* diff undo for clang later
* wip
* Wip
* const string&
* gui message doesnt show if it has a '>' symbol in error msg
* minor refactoring for readability (size_t calc fifo size)
* refactoring listening udp socket code: activated and datastream dont create udp sockets anyway, rc<=- should be discarded in any case
* wip
* refactoring memory structure access
* wip: bugfix write header + data to binary
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* portRoi no roi effecto on progress
* fail at receiver progress, wip
* segfaults for char pointer in struct
* reference to header to get header and data
* refactoring
* use const defined for size of header of fifo
* updated release notes
* remove pointer in callback for sls_receiver_header pointer
* rx same name arguments in constructors
* rx: same name arguments in constructor
* rx: removing the '_' suffix in class data members
* merge fix
* merge fix
* review fix refactoring
* gui message doesnt show if it has a '>' symbol in error msg
* minor refactoring for readability (size_t calc fifo size)
* refactoring listening udp socket code: activated and datastream dont create udp sockets anyway, rc<=- should be discarded in any case
* wip
* refactoring memory structure access
* wip: bugfix write header + data to binary
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* portRoi no roi effecto on progress
* fail at receiver progress, wip
* segfaults for char pointer in struct
* reference to header to get header and data
* refactoring
* use const defined for size of header of fifo
* updated release notes
* remove pointer in callback for sls_receiver_header pointer
* passing reference header for callback instead of copying it
* gui message doesnt show if it has a '>' symbol in error msg
* minor refactoring for readability (size_t calc fifo size)
* refactoring listening udp socket code: activated and datastream dont create udp sockets anyway, rc<=- should be discarded in any case
* wip
* refactoring memory structure access
* wip: bugfix write header + data to binary
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* wip
* portRoi no roi effecto on progress
* fail at receiver progress, wip
* segfaults for char pointer in struct
* reference to header to get header and data
* refactoring
* use const defined for size of header of fifo
* updated release notes
* refactoring from review: fwrite, static_cast
* separating pattern levels from command name: command line done
* separated patten level from command in examples and default pattern files in servers
* command line and server works
* python: patnloop not verified, wip
* works except for patloop (set, and get does not list properly)
* minor
* fixed tests
* added 3 more levels for ctb and moench
* wip
* minor err msg
* minor
* binaries in
* separating pattern levels from command name: command line done
* separated patten level from command in examples and default pattern files in servers
* command line and server works
* python: patnloop not verified, wip
* works except for patloop (set, and get does not list properly)
* minor
* fixed tests
* added 3 more levels for ctb and moench
* wip
* minor err msg
* minor
* binaries in
* python working
* import fix
* changed fw version for ctb and moench. binaries in
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
* prevent mem size check before Detector constructed (for loading config files), allowing freeing shm for submodules
* back to not allowing free submodules - consistency issues
* vicin default changed to 800, only setting vthx directly allows to set dac even if counter disabled, else disable counter, setallthresholdenergy if an energy is -1, get module value, fix that reg was repaced by isettings
* vth3 disabled for interpolation enable, interpolation disable sets counter mask to what it was before (updating old mask whn setting counter mask except for setting all counters for interpolation enable) and enabling vth3 if counter was enabled
* refactor and test for previous commit
* pump probe only has vth2 enabled, handles both pump probe mode and interpolation mode as well
* wip
* refactored pump probe and interpolation and added to setmodule
* check dacs and trimbits out of range for setmodule (not just threshold)
* binaries in
* m3: pump probe and interpolation mutually exclusive
* minor
* added the possibility to save settings file for m3 and eiger
* added save trimbits to gui
* update release notes
* python wip
* moved location of trimbits save option in gui
* python works
* updating getModule with all its parameters in the server side
* updating binaries
* roi structure expanded to have ymin and ymax
* compile with 'detector roi'
* wip
* wip, rx_roi, rx_clearroi
* wip rxroi
* rxroi wip
* wip rxroi
* merge fix
* wip
* rx_roi works, impl wip, test
* tests in, impl left
* wip, rxroi impl
* wip, rxroi impl
* wip
* setrx_Roi works, getrx_roi, wip
* rx_roi impl done
* wip, rxroi
* wip, getrx_roi rxr ports
* fix ports
* wip
* wip
* fix positions on server side
* wip
* numports wip
* wip
* jungfrau top inner interface row increment
* x, y detpos, wip
* removed eiger row indices flipping in gui (bottom flipping maintained)
* wip
* wip, jungfrau numinterfaces2
* jungfrau virtual works
* eiger, jungfrau, g2 virtual server works
* eiger positions fix, wip
* binaries in
* minor printout
* binaries in
* merge fix
* merge fix
* removing getposition
* setrxroi wip
* set upto port
* get messed, wip
* roi multi to module works, wip
* wip
* roi dont return -1
* added rxroi metadata in master file
* added rxroifromshm, not yet in detector
* rx roi in gui with box, also for gap pixels (gappixels for jungfrau mess)
* fix for segfault in gui with detaching roi box in gui
* wip
* m3 gui: slave timing modes should be discarded when squashing
* fixed m3 virtual data, and fixed counters in gui asthetics
* m3 roi works
* wip, g2
* wip
* handling g225um boards, and showing roi for gainplot as well
* udpate python functions
* fix for 1d and a2d roi written
* fixed actual roi written to file
* no virtual hdf5 when handling rx roi
* test
* minor
* binarie in
* start acq for master m3 was sent twice (non blocking), removed redundant code, check that there is only one master
* m3 can have more than 1 master (when many master modules used independently)
* fix for singe mod m3 or other dets
Use already installed version of the slsDetectorPackage. Assumes that the library has already been built and installed either on a system wide location or pointed to by CMAKE_PREFIX_PATH
* inconsistent copy with generalData and implementation members, especially for m3 (non default rxr generic values), issue caught on second configure with non m3 default values, eg tengiga 0
* removing test
* wip, adding m3 functions: polarity, inerpolation, pumpprobe
* added interpol, polarity, pump probe, analog pulsing, digital pulsing
* tests
* binaries in
* update release
* added python polarity enum
* fixed python and minor readability in mythen3.c
* binarie sin
* added all the m3 funcs also in list.c and enablingall counters for enabling interpolation
* binarie sin
Fixed by checking for help action before using the detector
added test that checks that for all helps this doesn't crash
Disabled Timer tests by default since they take ~2s
# Setting DAC names for CTB
* Introduced new shared memory for CTB only
* Prepared for additional functionality
* Works from C++ and Python
Co-authored-by: Dhanya Thattil <dhanya.thattil@psi.ch>
- progress looks at activated or enabled ports, so progress does not stagnate
- (eiger) disable datastreaming also for virtual servers only for 10g
- missing packets also takes care of disabled ports
* test for rx_arping
* arping ip and interface from client interface
* apring thread added to thread ids
* clean code for thread for arping
* removing the assumption that udpip1 fill be updated along with udpip2
* review, replacing syscall(sys_gettid) with gettid()
* removed Makefile for moench and integrated the build in CMake
* broke out tiff reading and writing to its own library
* moved tiff includes to include/sls
* moved tiffio source to src
* removed incorrectly used bps
* cleanup and tests for tiffio
* removed using namespace std from header
* some fixing for moench04
* Program for offline processing renamed
Co-authored-by: Anna Bergamaschi <anna.bergamaschi@psi.ch>
A delay of 100ms has been added between the generation of the stop pulse and the resetCore function call. This should give enough time to the detector to readout and streamout the ongoing frame before the internal logic is reset (even after the transmission is delayed with txndelay_frame).
Conflicts:
slsDetectorServers/jungfrauDetectorServer/bin/jungfrauDetectorServerv6.1.0
A delay of 100ms has been added between the generation of the stop pulse and the resetCore function call. This should give enough time to the detector to readout and streamout the ongoing frame before the internal logic is reset (even after the transmission is delayed with txndelay_frame).
* Setting pattern from memory (#218)
* ToString accepts c-style arrays
* fixed patwait time bug in validation
* Introduced pattern class
* compile for servers too
* Python binding for Pattern
* added scanParameters in Python
* slsReceiver: avoid potential memory leak around Implementation::generalData
* additional constructors for scanPrameters in python
* bugfix: avoid potentital memory leak in receiver if called outside constructor context
* added scanParameters in Python
* additional constructors for scanPrameters in python
* M3defaultpattern (#227)
* default pattern for m3 and moench including Python bindings
* M3settings (#228)
* some changes to compile on RH7 and in the server to load the default chip status register at startup
* Updated mythen3DeectorServer_developer executable with correct initialization at startup
Co-authored-by: Erik Frojdh <erik.frojdh@gmail.com>
Co-authored-by: Anna Bergamaschi <anna.bergamaschi@psi.ch>
* Pattern.h as a public header files (#229)
* fixed buffer overflow but caused by using global instead of local enum
* replacing out of range trimbits with edge values
* replacing dac values that are out of range after interpolation
* updated pybind11 to 2.6.2
* Mythen3 improved synchronization (#231)
Disabling scans for multi module Mythen3, since there is no feedback of the detectors being ready
startDetector first starts the slaves then the master
acquire firs calls startDetector for the slaves then acquire on the master
getMaster to read back from hardware which one is master
* New server for JF to go with the new FW (#232)
* Modified Jungfrau speed settings for HW1.0 - FW fix version 1.1.1, compilation date 210218
* Corrected bug. DBIT clk phase is implemented in both HW version 1.0 and 2.0. Previous version did not update the DBIT phase shift on the configuration of a speed.
* fix for m3 scan with single module
* m3 fw version
* m3 server
* bugfix for bottom when setting quad
* new strategy for finding zmq based on cppzmq
Co-authored-by: Dhanya Thattil <dhanya.thattil@psi.ch>
Co-authored-by: Dhanya Thattil <33750417+thattil@users.noreply.github.com>
Co-authored-by: Alejandro Homs Puron <ahoms@esrf.fr>
Co-authored-by: Anna Bergamaschi <anna.bergamaschi@psi.ch>
Co-authored-by: Xiaoqiang Wang <xiaoqiangwang@gmail.com>
Co-authored-by: lopez_c <carlos.lopez-cuenca@psi.ch>
* Modified Jungfrau speed settings for HW1.0 - FW fix version 1.1.1, compilation date 210218
* Corrected bug. DBIT clk phase is implemented in both HW version 1.0 and 2.0. Previous version did not update the DBIT phase shift on the configuration of a speed.
The new server has been compiled
Co-authored-by: lopez_c <carlos.lopez-cuenca@psi.ch>
Disabling scans for multi module Mythen3, since there is no feedback of the detectors being ready
startDetector first starts the slaves then the master
acquire firs calls startDetector for the slaves then acquire on the master
getMaster to read back from hardware which one is master
Before building from source make sure that you have the [software wiki](https://slsdetectorgroup.github.io/devdoc/dependencies.html) installed. If installing using conda, conda will manage the dependencies. Avoid also installing packages with pip.
Please do not update to any xxxx.xx.xx.dev0 tags. They are not releases, but tags for internal usage.
Use only releases with tags such as x.x.x or x.x.x-rcx.
## Documentaion
Detailed documentation can be found in the [software wiki](https://slsdetectorgroup.github.io/devdoc/index.html) and on the [official site](https://www.psi.ch/en/detectors/software).
### Documentation
Detailed documentation on the latest release of 5.0.0 can be found in the [software wiki](https://slsdetectorgroup.github.io/devdoc/index.html) and on the [official site](https://www.psi.ch/en/detectors/software).
## Installation
### 1. Install binaries using conda
Conda is not only useful to manage python environments but can also
be used as a user space package manager. Dates in the tag (for eg. 2020.07.23.dev0)
are from the developer branch. Please use released tags for stability.
We have three different packages available:
* **slsdetlib** shared libraries and command line utilities
* **slsdetgui** GUI
* **slsdet** Python bindings
### Binaries
Binaries for the slsDetectorPackage are available through conda.
```
#Add conda channels
#Add channels for dependencies and our library
conda config --add channels conda-forge
conda config --add channels slsdetectorgroup
conda config --set channel_priority strict
conda install slsdetlib #only shared lib and command line
pybind11 packaged into 'libs/pybind'. No longer a submodule. No need for "recursive" or "submodule update".
**2. Compile without script**<br>
Use cmake to create out-of-source builds, by creating a build folder parallel to source directory. This would create a debug build with address sanitizers.
* **Older versions**:
pybind11 is a submodule. Must be cloned using "recursive" and updated when switching between versions using the following commands.
#. Tftp must be already installed on your pc to use the bcp script.
#.Kill the on-board servers and copy new servers to the board.
#.Copy new servers to the board. See :ref:`how to upgrade detector servers<Detector Server Upgrade>` for more detals. A reboot should have started the new linked servers automatically. For Eiger, do not reboot yet as we need to program the firmware via bit files.
.. code-block:: bash
# Option 1: from detector console
# kill old server
ssh root@bebxxx
killall eigerDetectorServer
# copy new server
cd executables
scp user@pc:/path/eigerDetectorServerxxx .
chmod 777 eigerDetectorServerxxx
ln -sf eigerDetectorServerxxx eigerDetectorServer
sync
# Options 2: from client console for multiple modules
* This is crucial when registers between firmwares change. Failure to do so will result in linux on boards to crash and boards can't be pinged anymore.
* This step is crucial when registers between firmwares change. Failure to do so will result in linux on boards to crash and boards can't be pinged anymore.
#. Bring the board into programmable mode using either of the 2 ways. Both methods result in only the central LED blinking.
@ -50,8 +27,13 @@ Upgrade
Do a hard reset for each half module on back panel boards, between the LEDs, closer to each of the 1G ethernet connectors. Push until all LEDs start to blink.
* Software:
..code-block::bash
# Option 1: if the old server is still running:
sls_detector_put execcommand "./boot_recovery"
# Option 2:
ssh root@bebxxx
cd executables
./boot_recovery
@ -79,11 +61,24 @@ Upgrade
#update front right fpga
bcp download.bit bebxxx:/febr
#update kernel (only if required by the SLS Detector Group)
#update kernel (only if required by us)
bcp download.bit bebxxx:/kernel
#. Reboot the detector.
..code-block::bash
# In the first terminal where we saw "Succeess"
# reconfig febX is necessary only if you have flashed a new feb firmware
reconfig febl
reconfig febr
# will reboot controller
reconfig fw0
.. note ::
If the detector servers did not start up automatically after reboot, you need to add scripts to do that. See :ref:`Automatic start<Automatic start servers>` for more details.
As it is still in developement, the rbf files must be picked up from the SLS Detector Group.
If the detector servers did not start up automatically after reboot, you need to add scripts to do that. See :ref:`Automatic start<Automatic start servers>` for more details.
Gotthard II
-------------
Download
^^^^^^^^^^^^^
- detector server corresponding to package in slsDetectorPackage/serverBin
If the detector servers did not start up automatically after reboot, you need to add scripts to do that. See :ref:`Automatic start<Automatic start servers>` for more details.
For v7.x.x of slsDetectorPackage and older, refer :ref:`zeromq notes for cmake option to hint library location. <zeromq for different slsDetectorPackage versions>`
Build using in-built cmk.sh script
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
..code-block::bash
The binaries are generated in slsDetectorPackage/build/bin directory.
[-j <Number of threads>][-k <CMake command>][-l <Install directory>]
[-m][-n][-p][-r][-s][-t][-u][-z]
-[no option]: only make
-b: Builds/Rebuilds CMake files normal mode
-c: Clean
-d: HDF5 Custom Directory
-e: Debug mode
-g: Build/Rebuilds gui
-h: Builds/Rebuilds Cmake files with HDF5 package
-i: Builds tests
-j: Number of threads to compile through
-k: CMake command
-l: Install directory
-m: Manuals
-n: Manuals without compiling doxygen (only rst)
-p: Builds/Rebuilds Python API
-r: Build/Rebuilds only receiver
-s: Simulator
-t: Build/Rebuilds only text client
-u: Chip Test Gui
-z: Moench zmq processor
# display all options
./cmk.sh -?
# new build and compile in parallel (recommended basic option):
./cmk.sh -cbj5
# new build, python and compile in parallel:
./cmk.sh -cbpj5
#For rebuilding only certain sections
./cmk.sh -tg #only text client and gui
./cmk.sh -r #only receiver
.. note ::
For v7.x.x of slsDetectorPackage and older, refer :ref:`zeromq notes for cmk script option to hint library location. <zeromq for different slsDetectorPackage versions>`
Build on old distributions
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
If your linux distribution doesn't come with a C++11 compiler (gcc>4.8) then
it's possible to install a newer gcc using conda and build the slsDetectorPackage
Receiver processes can be run on same or different machines as the client, receives the data from the detector (via UDP packets).
When using the slsReceiver/ slsMultiReceiver, they can be further configured by the client control software (via TCP/IP) to set file name, file path, progress of acquisition etc.
Detector UDP Header
---------------------
| The UDP data format for the packets consist of a common header for all detectors, followed by the data for that one packet.
#. UDP source port is hardcoded in detector server, starting at 32410.
#.**udp_dstport** : UDP destination port number. Port in receiver pc to listen to packets from the detector.
#.**udp_dstip** : IP address of UDP destination interface. IP address of interface in receiver pc to listen to packets from detector. If **auto** is used (only when using slsReceiver/ slsMultiReceiver), the IP of **rx_hostname** is picked up.
#.**udp_dstmac** : Mac address of UDP destination interface. MAC address of interface in receiver pc to list to packets from detector. Only required when using custom receiver, else slsReceiver/slsMultiReceiver picks it up from **udp_dstip**.
#.**udp_srcip** : IP address of UDP source interface. IP address of detector UDP interface to send packets from. Do not use for Eiger 1Gb interface (uses its hardware IP). For others, must be in the same subnet as **udp_dstip**.
#.**udp_srcmac** : MAC address of UDP source interface. MAC address of detector UDP interface to send packets from. Do not use for Eiger (uses hardware mac). For others, it is not necessary, but can help for switch and debugging to put unique values for each module.
To know more about detector receiver configuration, please check out :ref:`detector udp header and udp commands in the config file <detector udp header>`
When using a blocking acquire command (sls_detector_acquire or Detector::acquire), the control server is blocked until end of acquisition. However, stop server commands could be used in parallel.
Location
---------
@ -14,23 +23,78 @@ Arguments
..code-block::bash
Possible arguments are:
-v, --version : Software version
-p, --port <port> : TCP communication port with client.
-g, --nomodule : [Mythen3][Gotthard2] Generic or No Module mode.
One can start the on-board detector server automatically upon powering on the board.
When using a blocking acquire command (sls_detector_acquire or Detector::acquire), the control server is blocked until end of acquisition. However, stop server commands could be used in parallel.
**Location:** slsDetectorPackage/serverBin/ folder for every release.
.. note ::
For Mythen3, Gotthard2 and Eiger, you need to add scripts to automatically start detector server upon power on. See :ref:`Automatic start<Automatic start servers>` for more details.
.. note ::
Eiger requires a manual reboot. Or killall the servers and restart the new linked one. If you are in the process of updating firmware, then don't reboot yet.
6.1.1+ (no tftp required)
---------------------------------------
#. Program from console
#. Kill old server and copy new server
.. code-block:: bash
# Option 1: from detector console
# kill old server
ssh root@bebxxx
killall eigerDetectorServer
# the following command copies new server, creates a soft link to xxxDetectorServerxxx
# [Jungfrau][CTB][Moench] also deletes the old server binary and edits initttab to respawn server on reboot
# Then, the detector controller will reboot (except Eiger)
* 48 bytes of the SLS Detector Header (described in :ref:`the current detector header <detector udp header>`)
* 64 bytes of packet mask
..code-block::cpp
typedef struct {
uint64_t frameNumber;
uint32_t expLength;
uint32_t packetNumber;
uint64_t detSpec1;
uint64_t timestamp;
uint16_t modId;
uint16_t row;
uint16_t column;
uint16_t detSpec2;
uint32_t detSpec3;
uint16_t detSpec4;
uint8_t detType;
uint8_t version;
} sls_detector_header;
struct sls_receiver_header {
sls_detector_header detHeader; /**< is the detector header */
sls_bitset packetsMask; /**< is the packets caught bit mask */
};
.. note ::
| The packetNumber in the SLS Receiver Header will be modified to number of packets caught by receiver for that frame. For eg. Jungfrau will have 128 packets per frame. If it is less, then this is a partial frame due to missing packets.
| Furthermore, the bit mask will specify which packets have been received.
File format
--------------
* The file name format is [fpath]/[fname]_dx_fy_[findex].raw, where x is module index and y is file index. **fname** is file name prefix and by default "run". **fpath** is '/' by default.
Master file is in json format.
* Each acquisition will have an increasing acquisition index or findex (if file write enabled). This can be retrieved by using **findex** command.
The file name format is [fpath]/[fname]_dx_fy_[findex].raw, where x is module index and y is file index. **fname** is file name prefix and by default "run". **fpath** is '/' by default.
* Each acquisition can have multiple files (the file index number **y**), with **rx_framesperfile** being the maximum number of frames per file. The default varies for each detector type.
* Some file name examples:
Each acquisition will have an increasing acquisition index or findex (if file write enabled). This can be retrieved by using **findex** command.
Each acquisition can have multiple files (the file index number **y**), with **rx_framesperfile** being the maximum number of frames per file. The default varies for each detector type.
Some file name examples:
..code-block::bash
@ -108,17 +296,19 @@ File format
# second acquisition, first file
path-to-file/run_d0_f0_1.raw
* Each acquisition will create a master file that can be enabled/disabled using **fmaster**. This should have parameters relevant to the acquisition.
* SLS Receiver Header consist of SLS Detector Header + 64 bytes of bitmask, altogether 112 bytes. The packetNumber in the sls detector header part, will be updated to number of packets caught by receiver for that frame. Furthermore, the bit mask will specify which packets have been received.
Each acquisition will create a master file that can be enabled/disabled using **fmaster**. This should have parameters relevant to the acquisition.
**Binary file format**
*This is the default file format.
This is the default file format.
* Each data file will consist of frames, each consisting of slsReceiver Header followed by data for 1 frame.
* Master file is of ASCII format and will also include the format of the slsReceiver Header.
Each data file will consist of frames, each consisting of slsReceiver Header followed by data for 1 frame.
Master file is of ASCII format and will also include the format of the slsReceiver Header.
**HDF5 file formats**
@ -160,7 +350,7 @@ When handling callbacks, the control should be returned as soon as possible, to
@ -8,21 +8,24 @@ open an issue at our `github repo issues
Common
------
Missing Packets
^^^^^^^^^^^^^^^
Possible causes could be the following:
#.Receiver PC is not tuned for socket buffer size and input packet queue.
* Refer to :ref:`Increase rmem_default, rmem_max and max_backlog<Receiver PC Tuning>`
1.Total Failure of Packet Delivery
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
#.Wiring
*Faulty wiring or connecting cable to incorrect interface.
#.Data cable plugged into the wrong interface on board (Jungfrau)
*Please ensure that the data cable is plugged into the rightmost interface (default for single interface). The inner one is disabled for PCB v1.0 and must be selected via command for PCB v2.0.
#. Link up and speed
* Check to see if there is a blue LED on board to signal that the link is up. Check ethtool and find if Link Deteced:Yes and Speed is acceptable (>10k).
* Check ethtool and find if Link Deteced:Yes and Speed is acceptable (>10k).
* Check to see if the 10G link is up (blue or red LED on board, close to SFP+). If not:
* Check transeiver and fibers are compatible (all MMF 850nm or all SMF 1030nm)
* Check fiber
* Check fiber polarity (if short range, unplug the link anywhere, and look at the light/dark pattern: dark has to mate with light)
#. Detector is not acquiring (Not Eiger)
* Take an acquisition with many images and using the following steps instead of acquire:
..code-block::bash
sls_detector_put status start
@ -30,44 +33,79 @@ Possible causes could be the following:
# which means the detector is acquiring.
sls_detector_get framesl
.. note ::
If you are using multiple modules, the previous command can return -1 because each module will return different values. Then, check for a single module instead: sls_detector_get 0:framesl
# If you are using multiple modules, the previous command can return -1 because each module will return different values. Then, check for a single module instead: sls_detector_get 0:framesl
#. Data cable plugged into the wrong interface on board (Jungfrau)
* Please ensure that the data cable is plugged into the rightmost interface. The middle one is disabled for PCB v1.0 and must be selected via command for PCB v2.0.
#. Detector is not sending data
* Check the board to see if the green LED is blinking next to the data cable, which means that the detector is sending data.
#. Detector is not sending data (Except Eiger)
* Check the board to see if the green LED close to SFP is blinking (detector is sending data). If not, detector is not operated properly (period too short/long, no trigger in trigger mode) or misconfigured and needs reboot.
#.Firewall or security feature
*A firewall or some security feature could be blocking the reception of data.
#. Ethernet interface not configured properly
* Ensure that the interfaces used are configured properly with the right mask and ip. Eg. use ifconfig and route commands to verify.
#.Power supply
*Check if power supply has enough current.
* For Jungfrau, refer to :ref:`Jungfrau Power Supply Troubleshooting<Jungfrau Troubleshooting Power Supply>`.
#. Ethernet interface not configured for Jumbo frames (10Gb)
* Ensure that the interfaces used in receiver pc have MTU 9000 (jumbo frames) enabled.
* Ensure that the interfaces (on NIC and the switch) used in receiver pc have MTU 9000 (jumbo frames) enabled.
#. Detector IP (Not Eiger)
* Ensure it is valid and does not end if 0 or 255. Also ensure that the detector ip is in the same subnet as rx_udpip and the masking in the interface configuration ensures this rule.
#.Tcpdump or wireshark
#.Check if 'rx_frames' counter in 'ifconfig' do not increment for interface.
* If no, check switch configuration if present. Port counters of switch can also help to identify problem.
* If yes, but receiver software does not see it:
* Check no firewall (eg. firewalld) is present or add rules
* Check that selinux is disabled ( or add rules)
#. Source UDP IP in config file (Not Eiger)
* Ensure it is valid and does not end if 0 or 255. Also ensure that the source ip 'udp_srcip' is in the same subnet as destination ip 'udp_dstip' and the masking in the interface configuration ensures this rule.
* Eg. If interface IP is 102.10.10.110 and mask is 255.255.255.0, udp_srcip has to be 102.10.10.xxx (same subnet)
* Use ifconfig and route commands to verify etheret interface configuration
#. Netstat and netcat
* Try with netstat to see if its really listening to the right interface. Or netcat to see if you get packets.
#. Wireshark or Tcpdump
* Use one of these to confirm that you receive packets (with the right filtering ie. source and destination ports, ip).
#. Check SFP modules
* Check if the SFP modules on both sides of the fiber are of same type.
2. Partial or Random Packet Loss (Performance)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.. note ::
The following suggestions are for convenience. Please do not follow blindly, research each parameter and adapt it to your system.
#. Receiver PC is not tuned for socket buffer size and input packet queue or other parameters.
* Refer to :ref:`Receiver PC Tuning<Receiver PC Tuning>`
#. Wiring
* Faulty wiring or connecting cable to incorrect interface.
#. Pinging the subnet (receiving only a few number of packets each time)
* If a switch is used between a receiver pc and detector instead of plugging the cables directly, one might have to ping any ip in the subnet of the Ethernet interface constantly so that it does not forget the ip during operation.
* Eg. if rx_udpip is 10.2.3.100, then ping constantly 10.2.3.xxx, where xxx is any ip other than 100.
* Using slsReceiver, you can use a command that does this for you:
.. code-block:: bash
# arping the interface in a separate thread every minute
sls_detector_put rx_arping 1
#. Only the slaves get no data
* Check trigger cabling and trigger configuration
* When you cannot stop Jungfrau slaves in sync mode, refer to :ref:`Cannot stop slaves<Jungfrau Troubleshooting Sync Slaves Cannot Stop>`.
.._Receiver PC Tuning:
Receiver PC Tuning Options
^^^^^^^^^^^^^^^^^^^^^^^^^^
.. note ::
| xth1 is example interface name in the following examples.
| These settings are lost at pc reboot.
#. Increase maximum receive socket buffer size and socket input packet queue.
* Temporarily (until shut down)
.. code-block:: bash
@ -111,6 +149,7 @@ Receiver PC Tuning Options
# check how many GB memory you can allocate, to avoid swapping otherwise
#. Modify ethtool settings.
* rx ring parameters
.. code-block:: bash
@ -127,6 +166,9 @@ Receiver PC Tuning Options
# check
ethtool -c xth1
# enable adaptive xoalescence parameters
ethtool -C xth1 adaptive-rx on
# set to max value in your pc settings
ethtool -C xth1 rx-usecs 100
@ -138,11 +180,43 @@ Receiver PC Tuning Options
# set to max value in your pc settings
ethtool -A xth1 rx on
.. note ::
* generic receiver offload (might not always work)
..code-block::bash
| xth1 is example interface name.
| These settings are lost at pc reboot.
# check
ethtool -k xth1
# enable generic receiver offload
ethtool -K xth1 gro
#. Disable power saving in CPU frequency scaling and set system to performance
* Check current policy (default might be powersave or schedutil)
.. code-block:: bash
# check current active governor and range of cpu freq policy
cpupower frequency-info --policy
# list all available governors for this kernel
cpupower frequency-info --governors
* Temporarily (until shut down)
..code-block::bash
# set to performance
sudo cpupower frequency-set -g performance
# or
cpufreq-info
for i in ‘seq 0 7‘;do cpufreq-set -c $i -g performance;done
* Permanently
.. code-block:: bash
# edit /etc/sysconfig/cpupower to preference
# enable or disable permanently
sudo systemctl enable cpupower
#. Give user speicific user scheduling privileges.
..code-block::bash
@ -155,18 +229,29 @@ Receiver PC Tuning Options
This is also set if slsReceiver is run as root user.
#. Some more advanced options:
.. warning ::
#. Disable power saving in CPU frequency
.. code-block:: bash
Please do not try if you do not understand
# or similar command depending on your distribution
cpupower frequency-info
cpupower frequency-set -g performance
#. reduce the number of queue per NIC to the number of expected streams: ethtool -L xth0 combined 2
#. assign each queue to its stream: ethtool -U xth0 flow-type tcp4 dst-port 50004 action 1
#. assign to each queue (IRQ) one CPU on the right socket: echo "3"> /proc/irq/47/smp_affinity_list #change the numbers looking at /proc/interrupts
#. disable irqbalance service
#. Be sure that the switch knows the receiver mac address. Most switches reset the mac lists every few minutes, and since the receiver only receives, there is not a periodic refresh of the mac list. In this case, one can set a fixed mac list in the switch, or setup some kind of script arping or pinging out from that interface (will be available in 7.0.0).
#. assign the receiver numa node (also with -m) to the socket where the NIC is attached. To know it, cat /sys/class/net/ethxxx/device/numa_node
#. ensure file system performance can handle sustained high data rate:
* One can use dd:
# or
cpufreq-info
for i in ‘seq 0 7‘; do cpufreq-set -c $i -g performance; done
#. Starting receiver as root to have scheduling privileges.
#. For 10g, enable flow control
.. code-block:: bash
sls_detector_put flowcontrol10g 1
#. Increase slsReceiver fifo depth between listening and processing threads.
#. Increase slsReceiver ring buffer depth
This can be tuned depending on the number of receivers (modules) and memory available.
..code-block::bash
sls_detector_get rx_fifodepth
# sets number of frames in fifo to 5000
sls_detector_put rx_fifodepth 5000
# sugggested not to use more than half memory of CPU socket in case of NUMA systems) for this
sls_detector_get rx_fifodepth
# sets number of frames in fifo to 1024 ~1GB per receiver. Default is 2500
sls_detector_put rx_fifodepth 1024
#. Increase number of frames per file
This can reduce time taken to open and close files.
#. Increase number of frames per file to reduce time taken to open and close files.
..code-block::bash
sls_detector_get rx_framesperfile
@ -193,10 +285,31 @@ slsReceiver Tuning
# writes all frames into a single file
sls_detector_put rx_framesperfile 0
#. Disable file write
This can ensure it is not the file system performance hampering high date rate.
..code-block::bash
sls_detector_put fwrite 0
Shared memory error
^^^^^^^^^^^^^^^^^^^
|For errors due to access or size, delete shared memory files nd try again.
For errors due to access or size, use any of the following suggestions.
#. Delete shared memory files and try again
#. Use environment variable to use a different shared memory ending in jfxx
.. code-block:: bash
# shared memory ending in jfxx
export SLSDETNAME=jfxx
#. USe a different multi shared memory ID
..code-block::bash
sls_detector_put 2-config xxxx.config
# or
sls_detector_put 2-hostname bchipxxx
To list all shared memory files of sls detector package.
.. code-block:: bash
@ -247,6 +360,19 @@ Possible causes could be the following:
* For Jungfrau, refer to :ref:`Jungfrau Power Supply Troubleshooting<Jungfrau Troubleshooting Power Supply>`.
Cannot ping module (Nios)
^^^^^^^^^^^^^^^^^^^^^^^^^
If you executed "reboot" command on the board, you cannot ping it anymore unless you power cycle. To reboot the controller, please use the software command ("rebootcontroller"), which talks to the microcontroller.
Gotthard2
---------
Cannot get data without a module attached
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
You cannot get data without a module attached as a specific pin is floating. Attach module to get data.
Gotthard
----------
@ -294,6 +420,24 @@ Cannot get multi module data
#. Check :ref:`Common Multi Module Troubleshooting<common troubleshooting multi module data>`
#. Power Supply
* Jungfrau needs a ~4A per module for a short time at startup. If not, it reboots misconfigured.
* Comment out this line in the config file: powerchip 1
* Powering on the chip increases the power consumption by a considerable amount. If commenting out this line aids in getting data (strange data due to powered off chip), then it could be the power supply current limit. Fix it (possibly to 8A current limit) and uncomment the powerchip line back in config file.
#. If cabling is accessible, ensure termination board and flatband cable between the masters and the slaves are connnected properly. Then try to stop.
#. If cabling is inaccessible, unsync first so that the slaves can get the stop directly from the client using the command. Then, don't use sync mode until the cabling is fixed.
..code-block::bash
# unsync, slaves command will fail as it is still in waiting state
sls_detector_put sync 0
# stop should now be successful as master does not determine the stop anymore
These command are recommended to be placed before "rx_hostname" if it is used.
Source Port
-----------
Hardcoded in detector server, starting at 32410.
udp_srcip - Source IP
---------------------
IP address of detector UDP interface to send packets from. Do not use for Eiger 1Gb interface (uses its hardware IP). For others, must be in the same subnet as **udp_dstip**.
udp_srcmac - Source MAC
-----------------------
MAC address of detector UDP interface to send packets from. Do not use for Eiger (uses hardware mac). For others, it is not necessary, but can help for switch and debugging to put unique values for each module.
udp_dstport - Desintation Port
-------------------------------
Port in receiver pc to listen to packets from the detector.
udp_dstip - Destination IP
--------------------------
IP address of interface in receiver pc to listen to packets from detector. If **auto** is used (only when using slsReceiver/ slsMultiReceiver), the IP of **rx_hostname** is picked up.
udp_dstmac - Destination MAC
----------------------------
MAC address of interface in receiver pc to list to packets from detector. Only required when using custom receiver, else slsReceiver/slsMultiReceiver picks it up from **udp_dstip**.
Since there is no difference in the format of the UDP header from the detector
from the previous version (v2.0), the version number stays the same.
Only the struture member names have changed in sls_detector_defs.h
Description
------------
***Detector specific field** descriptions are found :ref:`here<detector specific fields>`.
***frameNumber**: framenumber to which the current packet belongs to.
***expLength**: measured exposure time of the frame in tenths of microsecond. It is instead the sub frame number for Eiger.
***packetNumber**: packet number of the frame to which the current data belongs to.
***timestamp**: time measured at the start of frame exposure since the start of the current measurement. It is expressed in tenths of microsecond.
***modId**: module ID picked up from det_id_[detector type].txt on the detector cpu.
***row**: row position of the module in the detector system. It is calculated by the order of the module in hostname command, as well as the detsize command.
***column**: column position of the module in the detector system. It is calculated by the order of the module in hostname command, as well as the detsize command.
***detType**: detector type from enum of detectorType in the package.
***version**: current version of the detector header (0x2).
NR==FNR {if (match($0,"Rev")) {l[0]="0x"$(NF);} else if (match($0,"Date")) {l[1]="0x"$4; gsub("-","",l[1]);} else if (match($0,"URL")) {l[2]="\""$(NF)"\"";} else if (match($0,"Author")) {l[3]="\""$(NF)"\"";} else if (match($0,"UUID")) {l[4]="\""$(NF)"\"";} else if (match($0,"Branch")) {l[5]="\""$(NF)"\"";};next;}
{if (match($2,"REV")) {$0=$1" "$2" "l[0];} else if (match($2,"DATE")) {$0=$1" "$2" "l[1];} else if (match($2,"URL")) {$0=$1" "$2" "l[2];} else if (match($2,"AUTH")) {$0=$1" "$2" "l[3];} else if (match($2,"UUID")) {$0=$1" "$2" "l[4];} else if (match($2,"BRANCH")) {$0=$1" "$2" "l[5];}}1'$fin$ftmp > $fout
File diff suppressed because it is too large
Load Diff
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
