POL: Hardware and software: Difference between revisions

From DaqWiki
Jump to navigation Jump to search
Line 187: Line 187:
== Software ==
== Software ==


An example of the PPG pulses that get output is shown below.
=== How the Frontend Program Works ===
The Frontend program {{File|name=fepol_32bit.exe}}
controls the DAQ hardware in the VME crate and sends out the data in
the form of MIDAS-format data banks.  (See [[#To_build_frontend_pol_fevme.exe|POL frontend code]] for
location of code, and how to build the frontend).
At the beginning of run, the PPG program is compiled based on the user's settings (number of time bins etc), then loaded onto the [[#Hardware|PPG]]. An example of the PPG pulses that get output is shown below.


[[Image:Pol_ppg.png|800px]]
[[Image:Pol_ppg.png|800px]]
Line 195: Line 202:
In the new DAQ a cycle is defined as the scalers having seen N time bins. That means the PPG program is run multiple times within a cycle (e.g. if there are 100 time bins per program, and the cycle requires 1000 bins, the program will run 10 times). The first iteration is meaningless, as it is started by software rather than the RFQ (so the timing is not synced with any ToF of interest). The first time bin of each iteration is also meaningless, as the scaler has been counting for the entire time between when we sent a signal to the RFQ and when the RFQ sent a signal to us.
In the new DAQ a cycle is defined as the scalers having seen N time bins. That means the PPG program is run multiple times within a cycle (e.g. if there are 100 time bins per program, and the cycle requires 1000 bins, the program will run 10 times). The first iteration is meaningless, as it is started by software rather than the RFQ (so the timing is not synced with any ToF of interest). The first time bin of each iteration is also meaningless, as the scaler has been counting for the entire time between when we sent a signal to the RFQ and when the RFQ sent a signal to us.


=== How the Frontend Program Works ===
Once the scaler has seen all the bins we told it to expect, the data for this cycle can be completely read out and sent to midas banks.
 
The Frontend program {{File|name=fepol_32bit.exe}}
The frontend also keeps track of the buffer level in the MCS, and will read data during the cycle if the buffer is getting full. This means both that the buffer will never get full (causing data loss) and that the final end-of-cycle readout is faster (reducing deadtime).
controls the DAQ hardware in the VME crate and sends out the data in
 
the form of MIDAS-format data banks.  (See [[#To_build_frontend_pol_fevme.exe|POL frontend code]] for
Once the cycle is ended and data has been read from the scaler, the frontend program increments the DAC voltage.
location of code, and how to build the frontend).  
At the beginning of run, the hardware is initialized and set up.  The PPG program is downloaded into
the [[#Hardware|PPG]], with the input parameters [[vmic_pol.html#List_of_Input_Parameters|input parameters]]
e.g. ''number of bins per cycle'',
''DAC sleep time'',
etc. set to the initial user-defined values.  When the PPG is started, the PPG program will run for one '''PPG cycle''', then
stop. During the PPG cycle, the PPG outputs the desired number and sequence of pulses.  A '''SuperCycle''' is the number of PPG cycles
run with the DAC voltage constant. It may be one or more PPG cycles. After each SuperCycle is complete, the DAC voltage is stepped.
At the beginning of each SuperCycle, the PPG is started by software.
During a PPG cycle, outputs the user-selected ''number of bins per cycle'' pulses as the PPG ''MCS_Next'' output signal. This
signal is connected to the MultiChannel Scaler ([[#Hardware|MCS]]) ''LNE (Load Next Event)'' Input.  The MCS has a register that is pre-loaded
at beginning-of-run with the number of LNE (Load Next Event) pulses it must receive for one complete SuperCycle,  i.e. ''Number of
cycles per SuperCycle'' * ''Number
of bins per cycle''. These are both input
parameters.  


Once armed, the MCS will count data in each bin
The next cycle is then started by re-arming the scaler and starting the PPG by a software signal.
until the LNE signal causes it to move to the next bin.  The
MCS  has an internal buffer in which it stores the data.
The PPG will stop after it has completed the cycle, i.e. it has
output the ''number of bins per
cycle'' pulses and the ''Start
TOF'' signal.  At this point, during the TOF (time of flight), the PPG signal ''counter gate'' may be
used to inhibit counting, otherwise the MCS will continue to count data in a single bin, since no LNE signal will be received until the next PPG cycle.
In any case, the data from the first cycle and the first bin will be discarded.    Once the TOF is
complete, the PPG will be restarted by the external signal ''End of TOF'', and will produce
another sequence of pulses.  The MCS will start counting real
data into its bins.  This process will be repeated until the
MCS has received the expected number of LNEs preloaded in the
register.  The MCS ignores any more LNE signals until it is
re-armed.    At this point, the SuperCycle is complete and
the data from the MCS can be read out. 
While the SuperCycle is in progress, the frontend program ''pol_fevme.cxx'' monitors the MCS
data buffer. It will read out the buffer to prevent overflow (the
pol experiment does not usually have enough data for this to be
needed).  This task is performed by the software equipment '' "VME".''  The frontend
program  also determines whether the  SuperCycle is ended
by polling on a register in the MCS that counts the number of LNE
that it has received. This task is performed by the software
equipment ''"POL_ACQ".  ''
Once the SuperCycle is ended,  the frontend program increments
the DAC voltage. The software equipment "''VME''"  reads out or flushes the last of the
data from the MCS buffer. This data is for the whole SuperCycle. If
enabled, it sends this raw data out as the  Event 11 data bank
"MCS0"  (debug mode).  The equipment "''HISTO''"  then decodes this
data into individual cycles. The MCS time bin data for all cycles
for the SuperCycle are added together to form the histograms.
At this point, the hardware is restarted for the next SuperCycle,
i.e the Scaler is re-armed, the external PPG start is enabled, and
the next SuperCycle is started by starting the PPG internally (by
software).  Meanwhile, the histogram data is sent out as Event
5 data banks.  Sometime later, the software equipment "''INFO''" runs, sending out various
statistical data as Event 3 data banks, and also updating the ODB.




Revision as of 11:30, 5 July 2022

Links

Hardware

The VME hardware for the POL DAQ system is located in a VME crate in a blue rack on the ISAC-1 experimental area floor underneath the BNQR platform.

The VME crate contains 3 modules:

Table 1: VME Hardware Modules used by Pol Experiment
Name Purpose VME Base Address
VMIC CPU running linux
SIS3820 MultiChannel Scaler (MCS) 0x38000000
VME-PPG32 TRIUMF Pulse Programmer (PPG) 0x00100000


Other DAQ hardware required is the Galil RIO DAC/ADC. The VMIC communicates with this module via a private Ethernet connection.

Table 2: Non-VME Hardware
Hardware Module Purpose
Galil Rio-47120 16bit DAC/ADC


A diagram of the DAQ Hardware setup for the Pol Experiment is shown in Figure 1 below.

Figure 1
Diagram of the DAQ Hardware setup for the Pol experiment

Figure 1:Diagram of the DAQ Hardware setup for the Pol Experiment


Table 3: VME Crate Contents for Pol Experiment
Module Name I/O Channel Number(s) Purpose
SIS3820 MCS input 21-24 data channels (lemo)) referred to in this document as Scaler Inputs 0,1,2,3
input 1 control channel (LNE**) connected to PPG Output 1 (MCS Next)
VME PPG-32 PPG output 1 MCS Next connected to MCS control input 1
output 2 Cntr Gate debug only
output 3 Start TOF to experiment to start the TOF
output 4 PPG Running debug only
output 5-8 Duplicate of outputs 1-4 respectively for debugging
input Ext. Trig External trigger starts PPG
** Load Next Event


VME Crate

Four MCS (Multi-Channel Scaler) channels are enabled, referred to in this document as Scaler Inputs 0,1,2,3. Input 1 is currently connected to a pulser to give a constant count.

Two PPG Outputs (1 and 3) are used in the experiment (more are programmed but are used for debugging purposes).

  • PPG output 1 (MCS Next) is connected to SIS3820 Input 1 (LNE). The PPG provides the "Load Next Event" pulse for the SIS3820, i.e. it advances the time bins.
  • PPG output 3 (DAQ SVC) produces a pulse at the end of the PPG cycle to send a signal to the RFQ and start the ToF.

Except for the first PPG cycle at the beginning of each SuperCycle, the PPG is started externally by an input signal from the experiment sent at the end of the ToF to the PPG External Trigger input.

Note: The minimum width of the external input "PPG start" signal is 10ns.


Galil Rio 47120 DAC/ADC (16 bits)

Galil Rio Documentation can be found on the galil website. Local copies can be obtained at the links below:

The Galil Rio used by Pol is 16 bit, meaning the LSB is .0003V when running in ±10V range. The user can choose to run with a ±5V, 0-10V, or 0-5V range instead to reduce the LSB.

The IP address of the Galil rio device is currently defined as 192.168.1.100 (note it uses a private link to lxpol).


The table below shows the channels in the Galil Rio that are currently used in the Pol Experiment and the typical ranges.


Galil-Rio Channels used by the POL experiment
Module Name I/O Channel Usual range Purpose Connected
Galil Rio-47120 16bit DAC Output 0 +/- 10V DAC Set Value DAC Readback Value
Output 1 +/-5V DVM
ADC Input 0 +/- 10V DAC Readback Value DAC Set Value
1 +/- 5V User DAC Readback HV Monitor
2 +/- 10V User Readback
3 +/- 10V User Readback

The POL Experiment is presently set up so that the DAC set value is output on Channel 0 of the DAC. This is connected directly to Channel 0 of the ADC to provide a direct readback of the voltage output by the DAC (for sanity-checking during analysis). It is also connected to the HV power supply to drive the output voltage.

Channel 1 of the ADC is reserved for a user-supplied readback. This is connected to the HV power supply output monitor.

The users can connect Channels 2 and 3 of the ADC as required for monitoring purposes. All the channels shown in the table above are read out into the data banks.

The ADC Average values are also sent out in one of the data banks.

DAC and ADC channel and range settings

The DAC and ADC channel numbers and range settings can now be changed by the users, rather than being hard-coded in to the frontend.

ADC Averaging program

A program has been downloaded into the Galil Rio to perform averaging of the first 4 ADC channels (0-3). A copy of this program can be found in Appendix 1. It takes parameters Filter Window (W) in volts and Filter Factor (F) (both arrays). Default values are F[*]=100 W[*]=0.1 volts. The Filter Window and Factor values for each channel have been stored in the ODB so they can be easily changed. They are located in the ODB at /Scanning/GalilRIOScan/RangeSettings/Filter window and /Scanning/GalilRIOScan/RangeSettings/Filter factor.

The program continually averages the ADC input values. If the difference between the average value and the latest readback is greater than W[i] for that channel i, the average is cleared and a new average calculated. This can be used to detect the DAC having stepped. The Filter Factor F is used to start a new average after a certain number of readings.

Software

How the Frontend Program Works

The Frontend program fepol_32bit.exe controls the DAQ hardware in the VME crate and sends out the data in the form of MIDAS-format data banks. (See POL frontend code for location of code, and how to build the frontend).

At the beginning of run, the PPG program is compiled based on the user's settings (number of time bins etc), then loaded onto the PPG. An example of the PPG pulses that get output is shown below.

Pol ppg.png

Each time a signal is sent to CH1, the scaler moves to the next time bin. At the end a signal is sent to the RFQ. A signal from the RFQ is used to start the PPG program again.

In the new DAQ a cycle is defined as the scalers having seen N time bins. That means the PPG program is run multiple times within a cycle (e.g. if there are 100 time bins per program, and the cycle requires 1000 bins, the program will run 10 times). The first iteration is meaningless, as it is started by software rather than the RFQ (so the timing is not synced with any ToF of interest). The first time bin of each iteration is also meaningless, as the scaler has been counting for the entire time between when we sent a signal to the RFQ and when the RFQ sent a signal to us.

Once the scaler has seen all the bins we told it to expect, the data for this cycle can be completely read out and sent to midas banks.

The frontend also keeps track of the buffer level in the MCS, and will read data during the cycle if the buffer is getting full. This means both that the buffer will never get full (causing data loss) and that the final end-of-cycle readout is faster (reducing deadtime).

Once the cycle is ended and data has been read from the scaler, the frontend program increments the DAC voltage.

The next cycle is then started by re-arming the scaler and starting the PPG by a software signal.


Stepping the DAC

The Galil Rio DAC used by the POL experiment is 16 bit, meaning the LSB is 0.0003 V for a DAC range of +/- 10Volts. The frontend code pol_fevme.cxx checks that the Galil Rio is responding when the code is first run (in subroutine frontend_init) by setting the DAC to a value of 2.5 Volts, then checking that the difference between the set value and readback is less than 0.05V.

When the DAC is stepped during a run (unless the check on the DAC step is disabled), a more sophisticated test is performed (in subroutine incr_DAC) to verify that the DAC has changed, using the following algorithm: 

set_diff = fabs(dac_val - prev_dac_val); // difference between the present and previous DAC Set values
read_diff= fabs(dac_read - prev_dac_read); // difference between the present and previous DAC Readback values
fdiff = fabs (set_diff - read_diff); // difference between the two

// Check that DAC has changed
if ( fdiff > fabs(ps.input.dac.jitter__v_) * 5 ) // jitter is usually set to the LSB of the DAC
return (error);


Running the experiment (Real Data)

To change the input parameters, see Running the Pol Experiment.

  • When running with real data, the following input parameters should be set as indicated:
  • the first cycle of each SuperCycle should be discarded. The first TOF is started at the end of the PPG Cycle, so the data from every first cycle will not be valid.
  • the first bin of every cycle should be discarded. If the data inhibit signal is not connected to the Scaler, this bin will have been gathering data (noise) during the time the PPG is stopped between each cycle. Otherwise it will read zero.
  • External PPG start must be enabled

For maximum speed

  • set DAC sleep time to zero. The Galil RIO DAC changes very quickly. It is stepped during the data readout (by software), and the first cycle of each supercycle is invalid. This gives it plenty of time to change.
  • set TOF signal width to zero. The TOF start only needs an edge, and the PPG should stop as soon as possible.
Note: the minimal delay of

this PPG is 30ns (3 clock cycles), so setting pulse widths and delays to zero actually gives 30ns.


The Galil RIO DAC responds so quickly that the "DAC sleep time" parameter can be set to zero, setting the delay to the minimum of 30ns (3 PPG clock cycles). The "Start TOF" signal width can also be set to zero (i.e. 30ns). After sending the "Start TOF" signal, the PPG stops running. It will now respond to the next "PPG start" signal at the end of the TOF. External "PPG Start" signals that occur when the PPG is still running are ignored, hence the "Start TOF" signal width should be kept as short as possible.

The parameters in the information box should be set as indicated:

  • Alarm system enabled TRUE
  • Disable DAC Step checks FALSE
  • DAC jitter (V) 0.005 (0.0003) DAC LSB (see above)
  • DAC Minimum increment (V) should be set to > DAC jitter

When running with very small DAC Increments or for testing, Disable DAC Steps check can be turned off. This bypasses the DAC step check algorithm and the check on the minimum step size.


Testing the DAQ

For testing the DAQ, the TOF mechanism can be bypassed.

The PPG is triggered internally (i.e. by a VME command) on each cycle (i.e. not just the initial cycle in each SuperCycle). In order to do this, uncheck the "Use pulsed beam" box on the main status page.

When running in test mode, i.e. no TOF, the PPG is run with internal start. In this case, the End TOF signal does nothing. The PPG is restarted internally when the PPG stops at the end of each cycle, until the SuperCycle is completed.

Timing tests of the software can be done more easily in this mode, where the uncertainty of the TOF is removed temporarily.


Environment variables

The name of the MIDAS experiment is "pol" and it is currently run from isdaq01. Relevant Environment variables are set up as follows: 

MIDASSYS=/home/pol/packages/midas
MIDAS_EXPTAB=/home/pol/online/pol/exptab
MIDAS_EXPT_NAME=pol
HOSTNAME=isdaq01.triumf.ca
DAQ_HOST=isdaq01
VMIC_HOST=lxpol

The experimental clients are run on the host machine isdaq01 except for the frontend code fepol_32bit.exe which is run on the VMIC lxpol in POL's VME crate. lxpol has access to the pol user's home directory via NFS.


Building the software

Updating POL software

The main repository uses submodules for some dependencies, so checking out updates requires: 

git pull
git submodule update

To compile the real software, run the following on isdaq01 from the repository root: 

# For the first install
mkdir build
cd build
cmake ..
make install

# For updates
cd build
make install

This will:

  • Cross-compile the frontend to run on the 32-bit VMIC machine
  • Compile and/or cross-compile various debug executables
  • Compile and/or cross-compile various automated tests

To compile a dummy version of the software that does not talk to the real hardware or EPICS (and which can be run on a laptop): 

mkdir build
cd build
cmake .. -DDUMMY_MODE=1
make install

Updating midas

# Get updates
cd $MIDASSYS
git pull
git submodule update

# Compile 64-bit version
cd build
make install

# Cross-compile 32-bit version
cd ..
make linux32

# Change status page from midas default to a symlink
if ! [ -L $MIDASSYS/resources/status.html ]; then
  mv $MIDASSYS/resources/status.html $MIDASSYS/resources/status.html.orig
  ln -s ~/packages/pol/pol_settings.html $MIDASSYS/resources/status.html
fi

Then re-compile the POL software.

Scripts

start-all

start-all is an alias to the script "/home/pol/online/pol/bin/start-daq-tasks". This script will start the various clients for the experiment, including the Midas web browser client mhttpd on port 8088. The experiment is controlled through the Midas web browser (see Main Status page).

start-all will also start the frontend program running using the screen manager utility screen, which multiplexes a physical terminal between several processes. This has the advantage that deleting the frontend window (by mistake) will not kill the frontend program. Furthermore, the frontend window can be shown on any terminal, not just the terminal where start-all was originally run.

NOTE
If this is not successful, or the frontend window immediately

disappears, there may be hardware issues. See POL#Troubleshooting">troubleshooting.

start-all can be run at any time during the experiment, and will restart any of the clients if they have died. Clients can also be restarted using the

mhttpd programs page.


show-windows

A script show-windows run on any terminal (logged in as pol to isdaq01) will display the frontend window. An alias show-fe has been set up to run show-windows.

Appendix 1: Averaging Program

This program by Donald Arseneau is downloaded in the Galil Rio module and performs averaging. It can also be found at

~pol/online/rio/POL.dmc. 

REM Galil RIO used for ADC and DAC and digital IO by POL
REM Version 1.01 20-Jun-2014
REM
REM variables:
REM AV[] filtered voltage readings
REM F[]  Filter number or factor (Set like F[0]=30)
REM W[]  Filtering window in volts.
REM
REM Query a filtered reading with "MG AV[i]"
REM Query all four with "MG AV[0],AV[1],AV[2],AV[3]"

REM ------------------------------
REM Automatic initialization.
REM ------------------------------
#AUTO
REM ------------------------------
REM Device configuration:
REM Digital outputs logic-normal
REM Analog inputs 0-3 range +-10V
REM Analog outputs 0-3 range +-10V
REM ------------------------------
IQ65535
AQ0,2;AQ1,2;AQ2,2;AQ3,2
DQ0,4;DQ1,4;DQ2,4;DQ3,4
REM ------------------------------
REM Set up filtering for readbacks.
REM DFloop runs in thread 3
REM Default filter factor 50,
REM Default filtering window 0.1V
REM ------------------------------
DM F[8];DM W[8];DM AV[8]
F[0]=50;F[1]=50
F[2]=50;F[3]=50
W[0]=0.1;W[1]=0.1
W[2]=0.1;W[3]=0.1
I=0
#IDF
AV[I]=@AN[I];W[I]=1.0;F[I]=100;I=I+1
JP#IDF,I<8
XQ#DFloop,3
EN
REM ------------------------------
REM The digital filter for analog inputs 0-4
REM AV[] The averaged voltages
REM F[]  The filter number
REM W[]  The window (volts)
REM Query filtered readings with "MG AV[i]"
REM ------------------------------
#DFloop
I=0
#DFch
IF(F[I]<1);F[I]=1;ENDIF
VI=@AN[I];FI=F[I]
IF(@ABS[(VI-AV[I])]>W[I]);FI=1;ENDIF
AV[I]=AV[I]*(FI-1)/FI+(VI/FI)
I=I+1
JP#DFch,I<4
JP#DFloop
EN
REM 56789 123456789 123456789 123456789
Retrieved from "https://daq00.triumf.ca/DaqWiki/index.php?title=POL:_Hardware_and_software&oldid=6444"