pnetctl- controlling pdev from cima

LAST MOD:24oct08

todo
history.
pnetctl description
message protocol
pnetctl commands
conf keywords
pset keywords
Setup info needed for fits file

Other pdev info

Terminology:

dump:ffts are accumulated and stored as 1 dump. A dump can have 1,2 or 4 spectra/ffts depending on whether it is running in polA,B, or stokes mode.
fftlen: length of fft computed in spectrometer. 16 .. 8192 in powers of 2
group: spectrometer boxes run as a single experiment.

For alfa there can be two groups. One for the main and one for the piggyback observations.
A single group can have boards that are configured differently (but there are limits).
For single pixel observations you may need multiple groups for a single experiments if the configurations are now allowed in a single group.

rec: A rec is the unit that is output to disc. It can be one or more dumps.
scan:A continuous set of dumps that begin with a start command and finish when the requested time has elapsed, a stop is sent, or an abort occurs. A scan will start a new file and may span multiple files.

todo (top)

10oct08: let observer specify maximum recs(spectra) in an output buf. It will be limited by the max size of the shm buffer (about 20mb).

conf keyword: numblksio=n

0 --> default is put as many blocks from prun as fit in 20mb shm buffer
>0 The number of blocks from prun to put in one 20mb shm buffer for output. This will be clipped to make sure that it fits in the 20 mb buffer.

pnet --blksinbuf=n .. pass in to pnet as an argument.

setup default so undef --> 0
msg to psrv built in make_headers .. pack %hdrs... line 1884..

mbuf[0],ibuf[0] = blksize in bytes (data plus hdr at end) if tmDomain, just data
mbuf[4],ibuf[1] = number of blocks until done
mbuf[8],ibuf[2] = max file size in KB before switching to next file. 0--> no switch
mbuf[12],ibuf[3] = spid
mbuf[16],ibuf[4] = sprev
mbuf[20],ibuf[5] = bswap
mbuf[24],ibuf[6] = check sequence numbers each block
New entry goes here... put header to location 32, ibuf[8]
mbuf[28],ibuf[7] = start of header (6 words pdev header) ..

in psrv.phil... routine:psrv_write_file()

test m[28] if non zero than use this as blocks/iobuf.. still limit to 20mb buff.

bpOut.c .. is it receiving blocks/buf already..

02sep08:

test data taken 26aug08 for pshift overflow did not always store the updated values of rms a/d in the header. Log file shows that the sigstat message was being sent from pnetctl to bpOut via bpCmd... Need to figure out why it doesn't always work.. see x101/080826.

28aug08:

pnetctl does not return staterr done all the time. It is being set by the socket close to pnet. The idea is that if pnet calls anything but exit 0 then pnetctl should get an error on the close. Maybe i've got the $xxx wrong on the check..
A simple way to fix it would be to look for All spectrometers done.. string. If this doesn't come back before datataking is done (the pnet socket is closed) then we would return staterr done.

11jul08:

when computing dumps/row.

maybe want to have a variable that the user can pass in.

put it in the standard header in rsvd location 2. If 0 then do the default computation.

Time domain sampling. .pdev files: things that still need to be done.

When using timedomain, pnetctl should check that the pset command has filetype set to pdev.
Some of the pset keywords are used to setup things like filename.. This is probably not being done when using pdev files.
Some of the frequencies in the main header: lo1mix are getting loaded after we start when in fits mode. Pdev mode does not delay the writing of the header. So the values it has in the main pdev header are probably wrong...Maybe need to delay the writing of the pdev header too??
The main header needs to be reviewed to make sure that current data is getting written

02jul08: in ashift computation i'm assuming the default pshift value. If pshift keyword was used then the ashift value may be incorrect.
18jun08:

for integrations > .1 second, should probably downshift before accumulator so the number doesn't grow too big...Need to worry about 16 bits and ashift... probably only worry about 32 bits.

14jun08:

8 bit stokes. fits 8 bits is unsigned. pdev polA,polB, stokesI is unsigned but stokesU,V is signed.
For this to work i'd have to clip the unsigned values to 127 or play the bscale game.. Probably not worth allowing 8 bits full stokes.

10jun08:

pnetbld .. add packet length for timedomain as a keyword...
15jun08:done.. check -->FIXed but need to verify<-- In psrv.phil.c it is byteswapping the record header even in time domain mode. This is wrong since there is no record header for time domain mode. Need to check in psrv.phil.c for timedomain and skip the bswap. (mathews moon data has the last 8 bytes of each packet interpretted as rechdr and they are byte swapped.)

06jun08:

if an error occurs while taking data staterr done is returned instead of statok done.

If a single group is being used this works ok.
If multiple groups are used, then it will wait till all groups are stopped before returning the staterr done message. Probably should force the other groups to stop.

need usealfa keyword in conf setup to tell which boxes.conf to use.
need outformat keyword in pset to let use choose the kind of output format to use.

The default will be fits.
You can select pdev
Probably want to select psrfits at some point.

01jun08:

If you send a timed start for far in the future and then send a stop command before the timed start occurs, you have to wait for the timed start. pnet is looping on the 1 second tick syncs with prun.. so it is not looking at standardinput.

history: (top)

11nov08:

when closing files, zero fill till file length is a multiple of 2880 (fits blk size).
Files previous to this date are not a multiple of 2880 which is not according to the fits standard.

24oct08: Fixed bug in call to galToJ2000(raRd,decRd,Direction,outC1Rd,outC2Rd). This routine converts from ra,decJ2000 to l,b. I was passing in crval2,crval3 which were in degrees instead of Radians. Fixed the call on 24oct08. Prior to this the galactic coordinates glon,glat were wrong. You could fix them by recalling galToJ2000() (in c or idl) to redo the computation.
21oct08: fixed bug in alfaBmOffsets(). Prior to this date, alfa_ang was passed to this routine as degrees (same as the idl version), but the c version was expected radians. Prior to this date, if alfa_angle was non-zero, then the outer beam positions were wrong. You could fix it offline by taking beam0 position and recomputing the beam offsets using alfaAngle (although it might be a pain...). This mod was made in the checkpointed version rev879 and then installed online. The working version (that has the psrfits additions) has also been fixed)
10oct08: added conf keyword grprem to remove an already configured group
09oct08:

Allocating 1gb in shm. Need to have this much memory available or it will
bufpoolD.c When mmap fails on shm mmap i was putting the error return in the ptr. Later i would test for null. Turns out the error return is (void*)(-1) so on cleanup i would try to free up the mmap even though it hadn't been allocated. Fixed so ptr is null if not allocated.

07oct08:

pnetctl.. adjpwr, getpwr now works with group 0 and group1.

You need to have defined the group0 box to getpower on the group1 box. This should probably be fixed to allow group0 only.. The problem is that group0 boxes control the attenuators.

25aug08:

when a stop request was sent to pnetctl it was sending back stopok and then staterr done .

the stopok is correct.
The stat message definition (see below) says that a stop command will not return a statxx done.
The err part of staterr done ..is coming because the close of pnet is returning an error. This is probably because i haven't read all the data in the pipe at the close.
To make the program agree with the documenation, I've removed the statxxx done from the replies that are coming back from a stop request. After a stop you should now only get stopok

15jul08:

sigstat info rms,mean ,gainA/B now stored in fits bintable header at end
modified pnetctl,bpOut,aofitsLib.

12jun08:

pnetctl - now stores rms,mean,gain in %sigStat whenever it is measured. DALL will dump the current contents. Need to pass the info to bpOut via pset.

11jul08:

changed how the default dumps/output is computed

time domain then no change.
If winking cal:

try to make dumps/row a multiple of calon + caloff dumps
If this is too big and calon=caloff dumps then try to make is ta multiple of calon dumps
Still limit to 20 mb or 1 second.

If not winking cal and number of dumps > 15 try make it a multiple of 5 dumps.
Still need to add keyword to let user specify dumps/row.

Documentation conf keyowrd datatype: had tmda, tmdab.. should have been tmdma,tmdmab
New PSET keyword filetype=fits,pdev,psr .

Lets you specify what kind of output file you want to create. For now just fits and pdev is supported.
The state info is held in bpOut. When bpOut is started (pdevctall restart) it defaults to .fits . The state info is remember across pnetctl start/stops.
If filetype keyword is not sent, then it stays with the current value.

Time domain sampling now works from pnetctl.

You can use the conf keywords to setup the spectrometer.
You need to use the pset keywords to set the filetype to pdev. pnetctl does not check currently check fof .pdev files when i timedomain (it should).

30jun08:

updated sp1 header to include the following keywords that were not there:

SP1HDR_BLANKSEL,SP1HDR_BLANKPER, SP1HDR_OVFADC_THR , SP1HDR_OVFADC_DWELL , SP1HDR_CALSEL, SP1HDR_CALPHASE, SP1HDR_CALCTL, SP1HDR_CALON, SP1HDR_CALOFF

Changes were to sp1def.conf which is now in pdev/pnetbld/sp1def.conf, and libsrc/aofits/aofitsLib.c, which is called from bpOut.c. The idl routines were also updated. Note that the calctl info kis now in the fits header in 2 places... in the old keywords i used and the PHXXX keywords that i just added.
Also updated pnet.conf.templ to have default values for these new keywords.

pnetctl conf keyword blankatod added to blank fft's using the a/d overflow.

29jun08: added --blnkatod=n to pnetbld for atod overflow blanking.
26jun08: added pshift as keyword for pnetctl config...
19jun08: check polA B order. shut off some of the amps and they corresponded to the spectra in the fits files with no power.
16jun08:time domain sampling now correctly computes the duration of a packet.
15jun08:

there is no 8 byte hdr record for timedomain sampling.
fixed psrv.phil.c so it no longer swaps the last 8 bytes of the blk since it is data, not rec header
fixed psrv.phil.c so it computes how many blks in 1 second for timedomain. Was using the fft params which were not set..
pnet.conf.templ Changed default packetsize 0x10008 to 0x100000.. got rid of 8 byte rec hdr and increased to 1 mb.

14jun08:

pdev computes polA, polB, U,V, or stokesI. PolA, polB, and stokesI are unsigned while U,V are signed.
Fits only has signed formats for 16,32 bits. You can use bzero to scale them..

you would have to add different values to polA,polB,stokesI, than to U,V since the bzero is going to be applied to all of them (even though U,V are already signed) since they are in the same heap ptr field.
The other drawback is that when fits readers see bzero they automatically return floats. So you can't keep a large dataset in the more compressed format after reading it in

My solution:

Clip polA,polB, stokesI to the max signed integer for 16 bit and 32 bit data. Loses a bit of dynamic range but it works... 8 bit data is unsigned for fits so polA,polB, stokesI is ok but stokesU,V are wrong.

13jun08:

ASHIFT computation in pnetbld. Shift by 1 more for stokesU,V relative to pola,polB. This makes the rms of U,V the same as polA,B since jeff outputs 2*polA,2*polb, but only 1*U, 1*V. (note 13jun08 i switched to shift 1 less for u,v.. on 17jun08 i fixed it to shift 1 more.).

07jun08: added emaskacc, emaskash, maxfilesizemb to conf setup
06jun08:

if pnet aborts during datatking return: staterr done instead of statok done. One of the previous info messages will contain the error message.

02jun08:

additions to pset: backendmask,pcalctl, wcalon,wcaloff. These values are computed by pnetctl and sent to bpOut to put in the fits header. They are not sent by cima to pnetctl...
pnetctl logging if PSET command only log first 56 bytes.. rest is binary.

28may08:

bits selected for blanking and calinput were reversed from how the cables were plugged in .. Updated the software to match the hardware.
After taking data, realized that pnetbld also had the calselect bit in it.. Updated pnetbld..

23may08:

pdevif1 -> pdevif1hz .. now in hz so it is the same as the other frequencies specified.
datatype .. value "all" changed to "stokes" so it is more meaningful.

15may08: allow :: in conf and pset keywords. just ignore it. added prfeed keyword to pset. fixed emaskerr typo.
13may08: obsnm now allows embedded blanks...fixed emaskadc,emaskpfb .. had different keywords in pnetctl
12may08: req_vel .. I had it as char[8]. should be double.
09may08: object now char[16] in pset struct. added psetgrp in pset config. added req_sys,req_vel,req_vel_unit,caltype,obsmode,scantype.
08may08:added pset . projid, object moved to pset
02may08: emaskadc,emaskpfb, digfreqoffhz,pdevif1,pdevlohz are now implemented.
01may08: changed calena, calconf to single keyword: calctl. Added emaskadc, emaskpfb to conf keywords.
30apr08: moved projid, for conf to setp list. Added fileprefix as a conf keyword.
29apr08: Include reply messages in pnetctl command table, added asynchronous replies: info,done
27apr08: 4byte ascii commands sent to pnetctl are now upper case. command replies are lower case. This matches how jeff is talking to his programs.
27apr08: added getp command.
25apr08: start command STAR now accepts scan duration and starttime. scanduration no longer an keyword for CONF.

PNETCTL description (top)

pnetctl is the program that cima (the gui) uses to control the pdev software. It controls/talks to:

pnet: jeff's program to run the spectrometers. There can be multiple copies of pnet.
bpcmd:- phil's program sitting on each file servers that can configure bpInp,bpOut. There is one of these for each box used.
mix-ctl: These are the programs that are running on each spectrometer box that are used to control the mixer attenuators.

It's specs are:

It is a perl script located at /share/megs/phil/svn/pdev/datatk/pnetctl
It is started by xinetd automatically when connecting to its well known port (lets not put it on the web!!). Closing the socket makes it exit.
It uses tcp/ip.
Currently you can start multiple copies of pnetctl..It could theoretically work but we probably don't want to do this.
It will handle multiple groups: configuring, starting, stopping.
pnetctrl has a message passing prototocal.
There are a set of commands that pnetctl supports.
An example of using pnetctl.

The sequence of events to take data with pnetctl are:

send a conf command to pnetctl to configure a particular group. For alfa there can be 2 groups. This allows for the piggy backing of experiments. Each conf command will setup to run a copy of pnet. If multiple groups are used, you need to send multiple conf commands, and then multiple copies of pnet will be started when the start command is issued.
adjust the power levels
send pset command for each group. (note 2,3 can come in the reversed order if desired). This sends info to bpOut for the fits file.
send the start data taking command.
wait for data taking to end or send the stop command to terminate it prematurely.
go back to 1,2,3, or 4 or send exit command.

Message protocol: (top)

Message types:

pnetctl can handle 3 different kinds of message formats:

messageBinary

4 bytes message length (MLEN) (not including these 4 bytes) in binary (network byte order)
4 byte command in ascii. It does not have to be null terminated.
MLEN-4 bytes of ascii message

messageAscii:

4 bytes message length (MLEN) (not including these 4 bytes) in ascii. All 4 bytes must be ascii.
4 byte command in ascii. All 4 bytes should be ascii.
MLEN-4 bytes of ascii message

messageLine:

4 byte command in ascii.
message: byte number 5 till newline is the message.

Reply messages:

For Binary,or Ascii mode .. 4 bytes message len (MLEN)
4 byte ascii command that was sent but converted to lower case
4 byte status in ascii:

"ok ": ok
"err ": trouble with request
"die ": program is dieing.. reply msg will have why..

ascii reply message len=(mlen-8 bytes).

The all ascii format are for debugging or maybe driving it from a scripting language. Cima will use the messageBinary format.
The I/O in pnetctl is done via select.

PNETCTL COMMANDS: (top)

Message commands are in ascii. They span 4 bytes in the message. If the command is less than 4 bytes then it should be padded with spaces. The message follows the 4th byte of the message command. The commands for pnetctl are:

PNETCTL COMANDS (18apr08)
command	params	Description
ADJP	rmsVal	adjust power levels
adjp	replyMsg	sent back after power adjusted and new values measured. takes about 6 secs for 7 boxes. Each brd info separated by a newline. First line includes the adjpok . Example reply: time for 14 boards: 5.38 seconds adjpok b0s0g0 rms:A 30.1 28.1 B 29.6 29.5 mn:A 0.3 0.1 B 0.7 0.3 gn: 11 11 b0s1g0 rms:A 30.8 29.5 B 28.4 30.3 mn:A -0.0 -0.0 B -0.1 -0.2 gn: 2 2 b1s0g0 rms:A 31.1 29.8 B 29.2 29.2 mn:A -0.0 0.5 B -0.1 0.3 gn: 7 7 b1s1g0 rms:A 32.1 29.7 B 31.2 31.0 mn:A 0.6 -0.1 B 0.1 -0.3 gn: 2 2 b2s0g0 rms:A 28.9 28.9 B 30.4 29.1 mn:A 0.2 0.2 B -0.2 0.7 gn: 10 11 b2s1g0 rms:A 27.7 28.3 B 32.3 30.1 mn:A -0.1 -0.7 B -0.4 0.1 gn: 1 3 b3s0g0 rms:A 26.3 29.4 B 30.1 27.6 mn:A 0.6 -0.1 B 0.6 0.2 gn: 6 7 b3s1g0 rms:A 30.0 29.3 B 28.2 29.2 mn:A -0.4 0.6 B 0.4 -0.1 gn: 1 1 b4s0g0 rms:A 28.5 27.4 B 28.6 28.9 mn:A 0.7 -0.0 B 0.4 0.8 gn: 10 11 b4s1g0 rms:A 32.5 29.4 B 30.1 28.9 mn:A 0.1 0.2 B 0.3 1.2 gn: 2 3 b5s0g0 rms:A 29.2 29.8 B 28.1 30.0 mn:A 0.6 0.7 B 0.2 1.2 gn: 11 13 b5s1g0 rms:A 28.9 28.5 B 28.1 28.4 mn:A 0.4 0.8 B -0.2 -0.3 gn: 0 2 b6s0g0 rms:A 32.4 30.1 B 30.8 32.3 mn:A 0.4 0.5 B 0.4 0.3 gn: 9 12 b6s1g0 rms:A 31.8 29.6 B 30.7 30.3 mn:A 0.1 0.3 B 0.7 -0.1 gn: 0 3
CONF	see conf	Configure a group
conf	replyMsg	Sent back when configuration is complete: example reply: confok
DALL		dump current program info back to caller

dall	replyMsg	Debuging info sent back. each line separated by a newline. This data can change as more debug info added. Example reply: dallok tkDataCnt:0 grpList:g0 grp:g0 stat:0 tmDump:0.1 nDumps:0 boxesList:b0 b1 b2 b3 b4 b5 b6 Fnm:pnet.conf :/dump/pdev/files/pnetg0.conf Fnm:boxes.conf:/dump/pdev/files/boxesg0.conf box:b0 pdevH:pdev-110 psrvH:pdevs1 bandNms:b0s0g0 b0s1g0 suNms: box:b1 pdevH:pdev-109 psrvH:pdevs2 bandNms:b1s0g0 b1s1g0 suNms: box:b2 pdevH:pdev-107 psrvH:pdevs3 bandNms:b2s0g0 b2s1g0 suNms: box:b3 pdevH:pdev-118 psrvH:pdevs4 bandNms:b3s0g0 b3s1g0 suNms: box:b4 pdevH:pdev-116 psrvH:pdevs5 bandNms:b4s0g0 b4s1g0 suNms: box:b5 pdevH:pdev-104 psrvH:pdevs6 bandNms:b5s0g0 b5s1g0 suNms: box:b6 pdevH:pdev-113 psrvH:pdevs7 bandNms:b6s0g0 b6s1g0 suNms: pnetbld cmd:/usr/local/pdev/bin/pnetbld --pipe --fbtempl=/usr/local/pdev/etc/boxes.conf.templ --fptempl=/usr/local/pdev/etc/pnet.conf.templ --fcal=/usr/local/pdev/etc/cal.conf --box="g0 0 1 2 3 4 5 6" --spctype='AB ' --adcclk=172.032000 --tmdump=.1 --fpnet=/dump/pdev/files/pnetg0.conf --fboxes=/dump/pdev/files/boxesg0.conf ComInfo struct: ioName:masR ioType:msgB ioDir:R ioName:masW ioType:msgB ioDir:W ioName:pdevs1 ioType:msgB ioDir:RW ioName:pdevs2 ioType:msgB ioDir:RW ioName:pdevs3 ioType:msgB ioDir:RW ioName:pdevs4 ioType:msgB ioDir:RW ioName:pdevs5 ioType:msgB ioDir:RW ioName:pdevs6 ioType:msgB ioDir:RW ioName:pdevs7 ioType:msgB ioDir:RW
GETP	rms gain	query the current power levels. by default it returns the rms,mean, and gain leves. The params rms or gain will limit the value returned (it will speed things up bit).
getp	replyMsg	sent back after power/gain measured. The format of the message is the same as adjp except that it starts with getpok . All fields are always returned. If yu only asked for gain, the the rms, mean fields will have 0's. Examle reply to GETP: time for 14 boards:4.02 seconds getpok b0s0g0 rms:A 57.8 53.5 B 59.4 59.4 mn:A 0.4 0.1 B 0.7 0.5 gn: 11 11 b0s1g0 rms:A 57.6 55.4 B 55.9 59.7 mn:A -0.1 0.0 B -0.2 -0.3 gn: 2 2 b1s0g0 rms:A 58.5 56.0 B 58.0 58.2 mn:A -0.1 0.5 B -0.3 0.4 gn: 7 7 ... Example reply to GETPrms: time for 14 boards:3.98 seconds getpok b0s0g0 rms:A 53.9 49.9 B 57.8 57.8 mn:A 0.4 0.2 B 0.7 0.5 gn: 0 0 b0s1g0 rms:A 53.0 50.9 B 53.5 57.1 mn:A -0.1 -0.0 B -0.1 -0.3 gn: 0 0 b1s0g0 rms:A 53.8 51.4 B 55.8 56.1 mn:A -0.2 0.7 B -0.3 0.5 gn: 0 0 ... Example reply to GETPgain: time for14 boards=.95 seconds getpok b0s0g0 rms:A 0.0 0.0 B 0.0 0.0 mn:A 0.0 0.0 B 0.0 0.0 gn: 11 11 b0s1g0 rms:A 0.0 0.0 B 0.0 0.0 mn:A 0.0 0.0 B 0.0 0.0 gn: 2 2 b1s0g0 rms:A 0.0 0.0 B 0.0 0.0 mn:A 0.0 0.0 B 0.0 0.0 gn: 7 7 ...
NOP		nop command.
nop	replyMsg	reply from nop command Example reply: nop ok
PSET		Send set parameter block. This should be done after a successful conf command There should be 1 PSET command for each group that is being configured.
pset	replyMsg	Reply message from PSET command Example reply: psetok
STAR	scanDuration startTm	Start datataking. It will start all groups that have been configured. MParams: ScanDuration: how long to take data (in seconds). It will be rounded to a multiple of dumps. startTm: Time when the observation should start. If not supplied then the observation starts as soon as possible (on a 1 second tick). The time format is: hhmmss (ast). Program will cross ast midnite correctly. If it is 235958 and you want to start in 5 seconds you should enter 000003. unix time. seconds from 1970.
star	replyMsg	the reply for start datataking. It will be sent after all groups have been instructed to start. It will not wait for the actual start of datataking since the start time could be far in the future. An asynchronous stat message will be sent when each group starts. Example reply: starok
STOP		Stop datataking before the requested scanDuration has completed.
stop	replyMsg	Reply for STOP. It is sent after all of the connections to pnet (all groups) have been closed. Example reply: stopok
		ASYNCHRONOUS REPLIES Asynchonous replies use the standard reply format of:\|msglen\|cmd\|status\|msg. The only difference is that they are generated by pnetctl and sent to the master (cima) (they are not the reply from some command from the master).
info	asyncMsg	Whenever pnetctl receives an unsolicited msg, it will pass it back as an info mesage. An example would be the pnet messages that are output after datataking is started. The msg format is the standard reply format:msgLen\|cmd\|status\|msg . For messages from pnet the source pnetg0 or pnetg1 is prepended to the message. Example info messages after datataking started: infook pnetg0:Connected to all infook pnetg0:Connected to all psrv infook pnetg0:First dump sequence number is 01000 infook pnetg0:10.00 integrations per second infook pnetg0:Estimated dump time 100.0 s infook pnetg0:Estimated total dump size 0.92 GB infook pnetg0:Spectrometer box bandwidth estimates: infook pnetg0: pdev-104 1.31 MB/s infook pnetg0: pdev-107 1.31 MB/s infook pnetg0: pdev-109 1.31 MB/s infook pnetg0: pdev-110 1.31 MB/s infook pnetg0: pdev-113 1.31 MB/s infook pnetg0: pdev-116 1.31 MB/s infook pnetg0: pdev-118 1.31 MB/s infook pnetg0:Fileserver bandwidth and dump size estimates: infook pnetg0: pdevs1 1.31 MB/s, 0.13 GB infook pnetg0: pdevs2 1.31 MB/s, 0.13 GB infook pnetg0: pdevs3 1.31 MB/s, 0.13 GB infook pnetg0: pdevs4 1.31 MB/s, 0.13 GB infook pnetg0: pdevs5 1.31 MB/s, 0.13 GB infook pnetg0: pdevs6 1.31 MB/s, 0.13 GB infook pnetg0: pdevs7 1.31 MB/s, 0.13 GB infook pnetg0:Observation begins at 1209496349.01 Tue Apr 29 15:12:29 2008 infook pnetg0:All spectrometers running... infook pnetg0:0.00 MB/s 0.00 MB [0:0]/1000 blocks (0.0%)
stat	asyncMsg	stat messages are sent when the state of pnetctl changes. The current list of stat messages are: started starttime (unix time). Sent after datataking started. One messages is sent for each group that starts. statok started unixtime pnetg0 done sent when a scan has completed. The done message is sent once after all groups have completed. Note: A stat done is not sent after a stop request. statok kdone

PNETCTL conf KEYWORDS: (top)

These are the values defined in the configure command to pnetctl.
The conf command configures a set of spectrometer boxes. Each conf command will map to a single group. The format is:
confkey1=value1 value2;a ..:key2=value1 value2;b..:key3=value1 value2;a ...

A single field is keyword=val1 val2 val3
Each conf command will configure a single group. This is specified as the first value in the boxes= command. If multiple groups are needed you have to send multiple conf commands (and pset commands).
: separate fields
spaces separate values within a field.
A ;x following a value assigns this value to configuration x. normally x will be a,b ...The boxes command then assigns different boxes to a configuration.
If a configuration a,b,c does not have a value assigned to it, it will take the value of the standard configuration (the one with no ;x).
keywords/values are not case sensitive. (i need to check this) . Exceptions to this are any filenames that may be used.

KeywordCodes:

SVA single value allowed for all groups
SVG single value within 1 group
MV multiple values within group (but may have restrictions).
IMP implemented in pnetctl,pnetbld..

Keywords used in conf command to pnetctl
pnetctl conf keyword	CODE	pdevParms	Notes
adcclkhz	SVA IMP	adcclk	clock value for pdev spectrometer in Hz. 170000384 is good for 1 sec. 172032000 is good for 1 millisecond. No code to control synths yet
bits	Mv IMP	fmtwid,hr_lpf	Determine bits to use for output. Values are: 4,8,16 if time domain sampling 8,16,32 if spectra
blankatod	MV IMP	OVFACD_THRESH	Spectra Blanking using a/d overflow. The user specifies the number of a/d overflows need for blanking to occur If the threshold is met then the fft computed will not be included in the sum. If the keyword is omitted, no adc blanking will occur. example:blankatod=0..65535 A value of 0 or > 8192 disables blanking. blankatd=1 will blank any spectra with at least 1 adc overflow
boxes	SVG IMP		There are 14 spectrometer boxes split into two sets: group0 0 thru 6 group1 0 thru 6 Each box has a low band (a) and a high band (b). If you place a ;x after a box then this box will use configuration x eg 0;a is config a. You can assign a single band of 1 box to a config with : 0a;c .. box 0, band a, config c. The first value of the boxes command specifies the group that this conf command refers to. For alfa there can be 2 groups of 7 boxes each. For single pixel there can be more than two groups. Each conf command will configure a single group. If multiple groups are needed then multiple conf commands (and pset commands) must be sent. example: boxes=g0 0 1 2 3 .. group 0 boxes 0,1,2,3 boxes=g1 0 1 3 4 5 6 .. group 1 boxes 0 .. 6 boxes=g0 0 1 2;a 3a;b 3b;c group 0. boxes 0,1 are the standard config, box 2 is config a, box 3a is config b, and box 3b is config c
bwmhz	MV IMP	hr_dec,hr_mode, lpf_cN	Set the bandwidth (MHz) to use. Values are limited by: adcclkhz defines the clock rate. Progran rounds to the closest bandwidth that is a submultiple of the adclk rate. The submultiples of the clock rate can be 1 to 1024.
calctl	SVN IMP	cal_ctl calon cal_off calphase	Control the cal. Values are: disable - cal is off, cal input is ignored c0: output cal value of 0, look at input value c1: output cal value of 1 , look at input value w Non Noff fract: winking cal. arguments are: Non number of dumps for cal to be on Noff number of dumps for cal to be off fract value 0..1 where cal should transition in the last dump of each state. Winking cal starts with the calOff then calOn sections. If you want to drive the cal from a separate source (interim correlator, wapps, etc..), set the cal multiplexor in the control room to select that device and then set calctl to c0 or c1 (since the output value is then ignored).
datatype	MV IMP	fmttype hr_lpf dcnt	Type of data to record. Values are: AB : record spectra polA and polb I : record spectra: sum of polA and polb STOKES: record spectra: polA/2, polB/2, stokesU,V TMDMA: record time domain data polA only. TMDMAB: record time domain data polA and polB For time domain data the bits recorded are determined by the bits keyword.
digfreqoffhz	MV IMP		Digital mixer offset in Hz. The pdevlohz keyword maps the IF to the center of the spectra where it is then sampled at the adcclk frequency. A digital mixer will then offset digfreqoffhz from this center. It actually rotates the band this amount. digffreqoffhz becomes the new center of the band. The direction (+/-) is at the IF frequency and not the RF frequency (beware of band flip from hi side LO'S). This should only be used if you are reducing the bandwith from the maximum.
dumptmsec	SVG IMP	fcnt	number of seconds to accumulate ffts for 1 dump. Values are: Must have at least 4 ffts in an accumulation May not have more that 65535 ffts in 1 accumlulation The time is nfftsfftlen/(adcclkhz). this is the wall time.The actual integration time can be less if blanking is enabled. The wall time will be 1 fftlenadcclkhz seconds longer if stokes spectra with 32 bits (DCNT is set to 1 in the spectrometer for this config). The dump time must be the same for all configurations within a group.
emaskacc	MV IMP	err3,4_thresh	limit the number of 40 bit accumulate overflow errors in a dump before reporting a message. The values 0..n are for errors counts 2^n. Since the err3,4_thresh is a 4 bit register, setting errmaskacc to 16 will completely block messages.
emaskadc	MV IMP	err0_thresh	limit the number of adc overflow errors in a dump before reporting a message. The values 0..n are for errors counts 2^n. Since the err0_thresh is a 4 bit register, setting errmaskadc to 16 will completely block messages.
emaskash	MV IMP	err5,6_thresh	limit the number of 40 bit accumulator upshift overflow errors in a dump before reporting a message. The values 0..n are for errors counts 2^n. Since the err5,6_thresh is a 4 bit register, setting errmaskash to 16 will completely block messages.
emaskpfb	MV IMP	err1_thresh	limit the number of polyphase fitlerband overflow errors in a dump before reporting a message. The values 0..n are for errors counts 2^n. Since the err1_thresh is a 4 bit register, setting errmaskpfb to 16 will completely block messages.
fftlen	MV IMP	len	Length of fft to compute. Values are: 16 .. 8192 in powers of 2.
fileprefix	SVG IMP	name	file prefix to use for disc files. This can include any subdirectories under the default /dump/pdev/ file directory. fileprefix=a2334 filename is then /dump/pdev/a2334.20080430.b0s0g0.000001.fits fileprefix=phil/sineTest filename is: /dump/pdev/phil/sineTest.20080430.b0s0g0.00001.fits
grprem	SVG IMP		remove an already configured group using the group name specified in the boxes keyword. The grprem keyword allows you to remove an already defined group. It only works on already defined groups. If grprem= keyword is present in the conf request then it will be the only keyword acted on in the request. This means that to remove a group you should call conf with just the grprem keyword: Examples: CONFgrprem=g0
maxfilesizemb	SVA IMP	-	maximum size of an output file before switching to another file (during a single scan). This is in units of 2^20. so 2048 is a 2Gb file.
pdevif1hz	SVG IMP		center of the IF frequency (hz) used by pdev. typically it is 250000000.
pdevlohz	SVA IMP		The LOs to use for the pdev spectrometer. The two values specified are shared by all spectrometer boxes (we only have two synths that are split 14 ways). There is 1 LO for the LOWif band and 1 value for the HIGH if band. (you need to always specify 2 values. example: 175000000 325000000 The LO value maps the IF to the center of the spectra. The digfreqoffhz then offsets from this value The low,high values are limited by the hardware filters and IQ phase generator.
pshift	SVA IMP	PSHIFT	How often to down shift in the butterfly stages of the fft. 0 to 13 bits. If a bit is present then downshift after the butterfly stage. If a bit is 0, then don't downshift. Bit 0 is the last butterfly stage. Bit N is the 1st butterfly stage where fftlen=2^(N+1) If not supplied then pshift will computed automatically. The keyword has been added for debugging.
-		DumpStart dumpStop	channels within spectra to dump. hardcoded to entire fftlen
-		DCNT	computed by pnetctl. 0 unless 32 bit stokes where it is set to 1.
Not yet implemented
blankenable		BLANKSEL	0: no blankg, 1: blanking enabled.
blankPeriod		BLANKPER	Number of usecs to blank starting at the rising edge of the blanking signal. Only valid if blankenable is 1. If blankPeriod is 0, then the blanking is determined by the length of the blanking signal.
		lo1mix	this is upstairs synth1.. either pass it in or we can grab it from scramnet.

pset command keywords. (top)


pnetctl pset keyword	CODE	FORMAT in Fits file	Description
frontend	SVA	char[8]	receiver name. names are:"327","430","610","lbw","sbn","sbw","sbh","cb","cbh","xb","alfa","ch"
object	SVG	char[16]	source name
obsnm	SVG	char[24]	observer name
projid	SVG	char[8]	project id. eg "a2314"
topoafreqhz	MV	double	topocentric freq center pixel of lo band. units-hz
topobfreqhz	MV	double	topocentric freq center pixel of high band. units=hz
restafreqhz	MV	double	rest freq center pixel of low band. units=hz
restbfreqhz	MV	double	rest freq center pixel of high band. units=hz.
patternid	SVG	int*4	pattern id. Ydddnnnnn where Y is last digit of the year, ddd is the daynumber. This increments on every new pattern.
scanid	SVG	int*4	scannumber Ydddnnnnn nnnnn increments each scan . Y is the last digit of the year.
req_sys	SVG	char[8]	Requested velocity coordinate system. values are: "TOPOCENT","GEOCENTR","BARYCENT","LSRK","UNKNOWN"
req_vel	SVG	double
req_vel_unit	SVG	char[8]	unit of req_vel. "m/s" or "Z"
req_vel_type	SVG	char[8]	values are:"VOPT-F2W","ZOPT-F2W","ZRAD-F2V", or "UNKNOWN"
caltype	SVG	char[8]	type of cal used."lcorcal","hcorcal","lcal","hcal","lxcal","hxcal","l90cal,"h90cal
obsmode	SVG	char[8]	Observing pattern used. Some names are: "ON","ONOFF","DPS","RADECMAP","DECRAMAP","BASKET","FIXEDAZ"
scantype	SVG	char[8]	Each observing pattern is made up of one or more scans. scantype names the type of scan within a pattern. Some names are: "ON","OFF","UP","DOWN". Patterns that have a single scantype are usually labeled with "ON".
prfeed	SVA	int*4	alfa beam that is being used to point the telescope. Values 0..6.
psetgrp	SVG	- not in structure	specifies which group this pset config belongs to. If no psetgrp is given then it defaults to g0. Example: psetgrp=g0:
filetype	MV	not in structure	The type of output file to generate. An integer value: 0 don't change the current filetype (default if not provided) 1 generate fits files (aofits).. 2 generate .pdev files (jeff mocks format). This is needed for time domain data 3 psr .. psrfits format (not yet implemented)

NOTE:

The pset structure sent from pnetctl to bpOut.c includes the above data plus a few other keyword/values. See svn/pdev/includes/pnetctlPset.h

Setup info needed for the fits file: (top)

Last modifed 30apr08
STATUS KEYS:

P : PENDING
D : done

FIELD	bpCmd Location	Src	Description
From pnet open command
baseFilename	FILEI fname[]	filepre conf	Program will increment 5 digit index number if more than 1 file is needed per scan.
dumpsToWrite	FILEI blkstoWrite	start command	How many dumps to output for this scan.
maxBytesFile	FILEI maxBytesFile	conf	When output file gets this large, switch to new file. 0 --> stay with one file
nextSeqNumber	FILEI nxtSeqNum[]	pnet	Next sequence number for start of scan.
From sp1 header sent via pnet
dataType	FILEI[]		B byte, I short,J int,E float,D double. This is the datatype to be stored in fits header.
dumpsPerRec	FILEI[]		Number of dumps per output record.
exposure	FITSINIT		integration time in seconds. From FCNT,adckHz in sp1 hdr.
polPerDump	FILEI[]		from sp1 fmttype. Can be 1,2, or 4.

From pnetctl PSET via bpCmd
frontend	char[8]		receiver name. alfa is : "alfa" .. not necessarily null terminated.
object	char[16]		source name. does not have to be null terminated.
obsNm	char[24]		observers name or initials.
projid	char[8]		like a1723 .
topoFreq	double[2]		frequency center of band. crval1 Hz. Low, hi band
restfreq	double[2]		rest frequency center of band Hz. Low, hi band.
pattern_id	int*4		pattern Id Ydddnnnnn this increments on every new pattern.
scan_id	int*4		scannumber Ydddnnnnn nnnnn increments each scan . Y is the last digit of the year.
req_sys	char[8]		"TOPOCENT","GEOCENTR","BARYCENT","LSRK","UNKNOWN"
req_vel	char[8]
req_vel_unit	char[8]		unit of req_vel. "m/s" or "Z"
req_vel_type	char[8]		values are:"VOPT-F2W","ZOPT-F2W","ZRAD-F2V", or "UNKNOWN"
caltype	char[8]		type of cal used."lcorcal","hcorcal","lcal","hcal","lxcal","hxcal","l90cal,"h90cal
obsmode	char[8]		Observing pattern used. Some names are: "ON","ONOFF","DPS","RADECMAP","DECRAMAP","BASKET","FIXEDAZ"
scantype	char[8]		Each observing pattern is made up of one or more scans. scantype names the type of scan within a pattern. Some names are: "ON","OFF","UP","DOWN". Patterns that have a single scantype are usually labeled with "ON".
backendmask	int[8]		one entry for each box. low band is lower 4 bits, upper band is upper 4 bits. 4bit coding: 1- polA,B or stokes I only 3 - pola, polB 15- stokes
pcalctl	int		0 caloff always, 1 cal on always, 2 winking cal
wcalon	int		dumps for winking cal on
wcaloff	int		dumps for winking cal off
wcalphase	double		0 to 1. where the cal transition should occur in the last dump of calon or caloff.

<- page up

home_~phil