Version 0.2 Next draft March 1996
Version 1.0 First official version May 1996
Version 2.0 After major software upgrade August 1996
2.2 Communicating with Industry Pack Modules
2.3 Communicating with Galil motion control cards
2.4 Communicating with BVME780 framegrabber cards
3.2 Gripper unit
3.3 XY carriage
3.4 Off axis probe
3.5 Fibre modules
3.6 Instrument optical systems
3.7 Westinghouse TV camera interface unit
4.2 VME crate electronics
4.3 Robot electronics
4.4 Westinghouse interface electronics
4.5 Servo amplifier electronics
5.2 Robot internal cabling
5.3 External cabling
6.2 Gripper lubrication
6.3 Prevention of surface rust
7.2 Dismantling parts of the gripper unit
7.3 Dimantling parts of the XY carriage
7.4 Dismantling the off axis probe
7.5 Dismantling the fibre module
8.2 Failure to pick up a fibre
8.3 Failure to put down a fibre
8.4 Other fibre movement errors
8.5 Extracting robot from limit switch
10.2 Performing grid tests
10.3 Repeatability tests
10.4 Tuning XY DC servo loops
10.5 Tuning Z, theta and probe DC servo loops
Figures
Appendix A Cable schedules
Appendix B Circuit diagrams
1. Autofib-2 Control system software manual
2. TP32V CPU board manual
3. BVME framegrabber manual
4. Galil DMC300 motion control board manual
5. Greenspring Industry Pack carrier board manual
6. Greenspring Digital IO Industry Pack board manual
7. Greenspring ADC Industry Pack board manual
8. Cable schedules (Appendix A of this manual)
9. Circuit and layout diagrams (Appendix B of this manual)
10. PVP power amplifier data sheet
General descriptions of the instrument may be found in the following references:
SPIE paper references here
Briefly Autofib-2 is a robotic fibre positioner for the Prime focus of the William Herschel Telescope. It consists of the robot and fibre module located in place of the Prime Focus Instrument Platform on the Prime Focus Turntable (PFTT) with ancillary electronics located on the top end ring and in the GHRIL control room. The optical fibres from Autofib-2 feed via connectors (large fibre module only?) on the top end ring to the WYFFOS spectrograph located in the GHRIL hut.
The TP32V cpu card is located in the left hand slot, this is necessary for the correct operation of the VME system and also due to the length of various ribbon cables. Six connections are made to the cpu board.
1. SCSI connector
2. RS232 connector
3. Parallel printer connector
4. Hard/floppy disk boot sequence switch
5. Ethernet transceiver
6. Floppy disk drive connector
Next two BVME780 frame grabbers/graphics cards are mounted side by side in the crate. One acts as a frame grabber for the robot gripper TV system allowing the software to grab images of the fibres to determine their position. This card has two external connections, a video input from the gripper TV camera and a processed video output in order to display what the frame grabber and software are doing with the images of the fibres. The second BVME780 is used as a graphics card for the engineering mimic display. This card has five video connections, the inspection TV camera inside Autofib-2 is connected to the video input and the output RGB and Sync are connected to a colour monitor.
The fourth card in the VME crate is a custom designed and manufactured board that simply provides an interconnect facility for all the other cards and the multipole connectors to the outside world. Although this card resides in the VME crate it is not a true VMEbus card and simply takes power from the backplane with no other connections to the VMEbus being made via the backplane.
After a couple of spaces to allow ribbon cables to access the custom circuit board the next four cards are all Galil DMC320 motion control cards. Each of these cards is capable of controlling two motor axes. from left to right the cards control the following motors and encoders:
1a. X axis motor and rotary encoder
1b. X axis linear encoder
2a. Y axis motor and rotary encoder
2b. Y axis linear encoder
3a. Z axis motor and rotary encoder
3b. Theta axis motor and rotary encoder
4a. Off axis probe motor and rotary encoder
4b. Unused
Each motor or encoder axis uses a 26way ribbon cable to connect to the interface board. This cable provides all of the motor command, encoder signals, limit and home position switch signals.
The last card in the VME crate to be connected directly into the backplane is a Greenspring Industry Pack carrier board. Industry Packs are a modular means of providing general purpose input and output. The carrier board in the crate supports two Industry Packs which are simply plugged into the carrier board. One of the two Industry Pack modules provides 24 channels of digital ttl input and output. The second Industry Pack provides 16 channels with Analogue to digital conversion. All the control and status of the instrument is provided by the digital I/O and a single channel of the ADC provides a means for measuring a voltage which in turn gives the temperature of the instrument.
The final module in the VME crate are a pair of disk drives connected directly to the power supply and the cpu board. The hard disk is a SCSI 40MB disk. The floppy disk drive is capable of formatting and read/writing 1MB 3 1/2" double sided double density disks, note that it is unable to use high density 2MB disks. In the blanking panel to the right of the disk drive unit is a key operated switch connected to a jumper on the cpu board to allow easy switching between booting from hard disk (normal operation) to floppy disk.
2.1 Short IO space memory map
The short IO memory space for the TP32V cpu board starts at 0x2000000. The various VME cards used in the Autofib micro computer all have their addresses set by inserting or removing jumpers. The current base address values for the cards are as follows:
Industry Pack module A (digital IO) 0x20000000
Industry Pack module B (ADC) 0x20000100
Galil DMC320 X axis motion control card 0x2000FE00
Galil DMC320 Y axis motion control card 0x2000FE80
Galil DMC320 Z and theta axis motion control card 0x2000FF00
Galil DMC320 Probe motion control card 0x2000FF80
Gripper TV frame grabber ?
Engineering mimic graphics card ?
2.2 Communicating with Industry Pack modules
Two Industry Pack modules are used within Autofib-2, a 24 line digital IO card and a 16 channel 12 bit ADC card. The two industry packs are mounted on a single 3U carrier card which is fitted with a 6U height front panel to match the rest of the VME crate. The base address of the digital IO card is 0x2000000 and the base address of the ADC card is 0x20000100. The two industry pack cards are connected to the rest of the Autofib-2 electronics by two 50-way ribbon cables that emerge from the front panel and loop back into the VME interconnect circuit board to the left of the industry pack carrier card.
Two simple and crude test programs are briefly described for accessing the two industry pack cards: digio.c and adctest.c.
2.2.1 Digital input and output card
The 24 channels of the digital IO card are addressed on 8 bit boundaries (16 and 32 bit boundaries are also possible but not used). At present only channels 1-8 are used for output and channels 9-16 are used for input, channels 17-24 remaining unused. The channels may be addressed as follows:
Channels 1-8 bits 0-7 at BASE +1 (alternate BASE +5)
Channels 9-16 bits 0-7 at BASE +0 (alternate BASE +4)
Channels 17-24 bits 0-7 at BASE +3 (alternate BASE +7)
By default all channels are used for output, for example to write 1's to channels 1-8:
outp(0x2000001,0xFF)
Confirmation of the status of output channels may be obtained by reading back the value from the same address, this also enables individual bits to be masked. For example:
inp(0x20000001)
will read in the last char written to channels 1-8.
To use a channel for status input, it is necessary to write a 1 to the required channel and read the status from the alternate address location. To set up channels 9-16 for status input write 1's to channels 9-16 at address BASE +0 and read the status values from BASE +4. This is the set-up used in the Autofib-2 control software:
outp(0x2000000,0xFF) write 1's to output channels 9-16
inp(0x20000004) read in status of channels 9-16
A short test program was written to test the hardware, ensure that no other software is running then from the OS/9 prompt type:
digio [return]
This program sets up channels 1-8 as control lines (output) and channels 9-16 as status lines (input) in exactly the same manner as the control software uses. Identical software functions are used to write characters to the memory addresses and bit masking as in the control software. On powering up the VME crate the default is to have 1's written to the output lines so the convention of 1 = off and 0 = on is used to avoid having equipment powering up if the VME crate is reset. The output and input channels have the following significance if the robot hardware is connected and powered up:
Channel Control line number 1 Instrument power on/off 2 Motor relays on/off 3 Gripper open/close 4 Instrument inspection lights on/off 5 Limit switch override control 6 Guide fibre back illumination on/off 7 Spec. fibre back illumination (not used) 8 not used
Channel Status meaning number 9 Instrument power status 10 Air pressure status 11 Motor relay status 12 Fibre module proximity switch A 13 Fibre module proximity switch B 14 Servo amplifier status 15 Spec. fibres back illumination status (n/u) 16 Guide fibre back illumination status
Channels 17-24 are currently unused but are connected to the VME interconnect board if a future use for them is found, they may be accessed by a 16-way header on the VME circuit board.
2.2.2 Analogue to Digital Converter (ADC) card
The ADC card is only used to measure a voltage which is generated by the instrument temperature sensor circuit. Only channel 1 of a possible 16 is used.
The base address of the ADC card is 0x20000100, the card can only be accessed using 16 bit words (short int), at BASE +0 is a 16 bit control register and the selected channel reading may be accessed by reading a 16 bit word on BASE +10 noting that the most significant 4 bits are not used (it is only a 12 bit ADC). See the manual for details of how to program the control register but it is enough to know that the control register is used to select the channel, set up that gain and voltage range and select single line or differential voltages.
In the control software the control register is set to 0xFF30, for gain of unity, voltage range of -5V to 5V and reading channel 1. A simple test program using identical software to the main control system to access channel 1 in the same manner is provided, at the OS/9 prompt type:
adctest [return]
hitting return will give successive readings of the voltage on channel 1, after about 10 readings the program will return to the OS/9 prompt.
2.3 Communicating with Galil motion control cards
A simple test program is provided to allow communication with and testing of the DC servo loops used in Autofib-2. The source code for the program is in dmc300.c in the /h0/autofib/testprogs directory. Typing dmc300[return] at the OS/9 prompt will start the test program running. Note that all input to the program must be in upper case as the DMC320 control cards do not recognise lower case characters. The commands to the DMC320 motion control card are sent as an ASCII string terminated with a carriage return, then handshaking routine used for this and the return of status responses is defined in the DMC300 user manual. The C function used to communicate with the motion control card is closely derived from the example code supplied with the card. Each of the four motion control cards used in the crate is jumpered to appear at a different address and also to power up in a motor off state.
The X and Y axes use a rotary incremental encoder connected directly to the motor to provide feedback to the servo loop in order to provide high servoloop stability. A second linear encoder is used to provide high accuracy position information. A single two channel card is used for the X axis and a second two channel card for the Y axis, the linear encoder in both cases having its own motion controller although it does not have a motor to control.
The Z, theta and off axis probe motor control cards are used in a standard servo loop manner, that is a single incremental encoder is used to directly provide the feedback required to control the motor, the encoder, limit and home switches are connected directly to the motion control card via the 26-way Dtype connector. A single channel is used for each of the Z, theta and probe servo loops. With two motion control cards available each with two channels this leaves one spare channel unused.
The dmc300 test program allows communication with all five motor axes and two linear encoder axes via a single letter name (X,Y,Z,T,P,A,B), all other commands are straight from the standard command set of the motion control card.
A description of the handshaking routine may be found in the DMC320 manual but it should not be necessary to deal with the card at this level as software is provided to allow the writing of ASCII strings to the various Galil cards (dmc300.c)
See sections 10.4 and 10.5 for a description of tuning the servo loop parameters using the dmc300 test program.
2.4 Communicating with BVME780 framegrabber/graphics cards
The two BVME780 video cards are identical except for the VME address each it jumpered to appear at. These cards are communicated with via a file manager device driver rather than directly through VME addresses as for the motion control cards and IO cards.
Each device (/vid and /vid2) may be opened as a normal device (eg /t1 or /pr) and the device drivers installed on the system will handle commands and pass them onto the hardware.
Information about the C library functions available is kept in the RAVE software manual. Two demonstration programs are also available:
/h0/autofib/grabtest/grabtest.c***** to be checked
/h0/autofib/gfmdemo.c****** to be checked
The first of these programs was written to develop the centroiding code used in the fibre placement algorithm, the second program was supplied with the hardware and can be used to verify correct hardware operation as it tests both framegrabbing from a video camera and graphics display if a video CCD camera is connected to input channel 0 of the card being tested and an RGB+SYNC monitor is connected to the output.
The whole of the instrument has been assembled on an octagonal piece of aluminium alloy cast tool plate 19 mm thick (the baseplate). The reason for the shape is to avoid vignetting the telescope beam by keeping the instrument outline within the central obstruction of the telescope. The octagon is not regular in that there are four long sides fitted with thin cover plates and four short sides with structural casework which is used to support and attach the fibre module, see Figure 1.
The XY carriage subsystem and off axis probe are attached directly to the baseplate. The Y axis is raised off the level of the baseplate to clear the Prime focus corrector which protrudes through the central aperture of the baseplate to within 1mm of the underside of the Y axis carrying plate.
The various circuit boards, cable chains and bits cable trunking are arranged wherever possible within the remaining space resulting in a very cramped interior to the robot. A cover plate is fitted over the eight walls of the robot which a large central aperture to allow the robot access to the fieldplate of the fibre module.
3.2 Gripper unit
The gripper unit is a self contained unit which may be removed from the XY carriage subsystem. The gripper consists of two axes of movement, a pair of pneumatic gripper jaws and a robot vision system. A cross section of the gripper unit is shown in figure 2.
The gripper casework is used to mount the gripper unit onto the XY carriage using four M5(?) bolts. To enable the gripper to be replaced in exactly the same location two aids to alignment are provided. With the robot horizontal on the testing trolley the gripper should rest firmly on the gripper carrying plate to locate it in the Z direction. The Y location of the gripper is controlled by an index block fixed to the rear of the mounting plate of the gripper, the gripper unit should be moved in the Y direction so that this index block rests firmly against one of the gripper unit attachment points. Under no circumstances should it be necessary to remove this indexing block otherwise the offset between the gripper and mobile sky probe will require to be recalibrated which must be done on the telescope with a bright star and the new offsets put into the robot control system.
The casework for the gripper is used to carry the Z axis DC servo motor and leadscrew together with four hardened steel rods upon which the gripper body slides as it is driven in the Z direction. The TV vision system is also firmly mounted to the gripper casework and should not be removed under any circumstances as the centre of the CCD TV must remain aligned with the centre of rotation of the theta axis.
The main gripper body has four pairs of linear bushings, one at each corner to allow it to slide up and down the four steel rods in the gripper casework when driven by the Z motor. The leadnut for driving the gripper body from the Z axis leadscrew is connected to the gripper body by a spring and microswitch so that if the gripper meets an obstruction the microswitch is opened and the Z axis motor power is cut. On the side of the gripper casework is a metal flag which provides an obstruction for the Z axis home position optoswitch. If this flag is removed or knocked the definition of the fieldplate height will have to be checked and if necessary altered (see the PLATE variable in the software manual) or the gripper will descend to the wrong height and cause the jaws to miss the fibre button handle or jam the button in the gripper jaws.
Within the main body of the gripper is housed the theta bearing which supports the centrally mounted pneumatic gripper allowing it to rotate continuously through 360. The theta axis is driven by a DC servo motor which drives a worm and worm wheel system. The zero point to the theta axis is provided by a metal flag and an optoswitch. If this optoswitch is knocked the zero point of the theta axis will require recalibrating as the robot will fail to line up the gripper jaws on the handle of each of the fibre buttons when attempting to pick them up.
The pneumatic gripper consists of a brass annular piston which is driven towards the fieldplate to close the gripper jaws by compressed air and returned away from the fieldplate to open the gripper jaws by a return spring. Compressed air is fed to the piston from a chamber in the gripper body which runs around the whole of the theta axis and then through a small hole in the rotating part of the piston housing so that the gripper can operate at any theta position.
As already mentioned the gripper TV vision system is aligned so that the CCD TV is centred on the theta rotation axis with the pixels aligned along the X and Y directions and the image in focus when the fibre button is on the field plate, this means that the fibre image will go out of focus when the fibre button is lifted off the plate by the Z axis. Light from the fibre passes through a hole in the centre of the pneumatic gripper through a transfer lens and folded by a prism into a microscope objective to provide a magnified image at the CCD TV camera.
3.3 XY carriage
The XY carriage subsystem is mounted directly onto the main instrument base plate and provides a means of moving the instrument gripper unit and mobile sky viewing probe to any point in the focal plane or retracting them out of the field of view. The X and Y axes are identical in operation with the only physical differences being the length and diameter of the ballscrew used to drive the particular axis and the size of the motor used to power the axis.
Each of the X and Y axes is powered by a DC servo motor with a rotary encoder mounted directly onto the rear of the motor shaft to provide a means to close the servo loop with no compliance between the motor and encoder. The front of the motor shaft is connected to a precision ball screw which a pitch of 4mm. The ball nut is of the zero backlash type with the two halves pretensioned against each other. Note that there are two rubber O rings at the extreme ends of the ball screw to prevent the ball nut being wound off the end of the ballscrew in the event of the instrument being dismantled as this would result in the loss of a great number of ball bearings. The exterior of one end of the ball nut is screwed into a block which is part of the Y axis carrying plate (X axis) or a block which is part of the gripper unit carrying plate (Y axis). The Y axis carrying plate and the gripper carrying plate run on high precision linear bearing slides. The Y axis carrying plate is bolted and doweled into position on the X axis slide units and should not be removed. If either the X or Y axis requires a full dismantling, recalibration of the instrument (grid tests and temperature calibration) will be required.
Simple optoswitches with ttl output are used to provide an indexing position and end of travel limit detection (RS stock number 304-560) for each axis. The rotary encoder output (quadrature counts in line driven ttl format), limit switches and home index switch signals are fed to a line driver circuit for each axis which is mounted in close proximity to the corresponding motor encoder unit. All signals (now line driven) are set back to the main instrument circuit board.
By itself the servo system described above is enough to drive the robot at the high speeds required. However the rotary encoders used are unable to see the periodic errors associated with the ballscrews so an additional Heidenhain linear encoder system is used on each axis purely to report the actual position without errors due to the ballscrews. The linear encoders are fitted with interpolating electronics to give a resolution of 1m and the encoders themselves are certified accurate to within 3m over their whole travel.
Even this dual loop encoding cannot provide sufficient accuracy as the linear encoders are unable to see the errors associated with the straightness of the linear bearing tracks used, errors due to the X and Y axes not being aligned perfectly orthogonally and errors due to temperature variation. To remove these sources of error they are first measured by performing repeated measurements of an accurate grid of dots (grid tests) at different instrument attitudes and temperatures then tabulating the errors so that they can be removed in software.
The X and Y axes provide a means of moving the gripper carrying plate to a given point with an accuracy of a few microns. On the gripper carrying plate are mounted the gripper (see section 3.2) and the mobile sky viewing probe. This mobile probe is optically and mechanically identical to that used for the off axis probe but has the advantage that it can be driven to any point in the focal plane (unless the off axis probe has been driven in to access the edge of the focal plane). Like the off axis probe the image is relayed to the autoguider optics by a coherent fibre bundle which follows the route of the electronics cables through the cable trunking to the optics box.
3.4 Off axis probe
The off axis probe is mounted on the Y=0 axis and can move only in the X direction to get it completely out of the 1 degree field of view or to access the very edge of the field of view. Although it may be moved to any position in its range, only the two extreme positions are useful during astronomical use. The sky viewing probe is mounted on a linear bearing and driven via a ballscrew by a DC servo motor with integral rotary encoder unit. No linear encoder is provided as its absolute position is not important as long as it is repeatable. The probe has two limit switches of the same type as for the X and Y axes which are connected together with the motor-encoder unit to a line driver circuit identical to that used for each of the X and Y axes. After being line driven the signals are directed back to the main instrument circuit board. Note that the off axis probe has no home index switch, instead the probe is slowly driven into its reverse limit switch (moving the probe out of the field of view) to define its zero point as part of its initialisation procedure before being automatically moved out of the limit switch for normal operation.
The probe itself consists of a folding flat mirror to redirect the telescope light beam, a relay lens to relay the folded focal plane and provide a slight change of image scale at the surface of a coherent bundle which is used to transfer the image acquired by the off axis probe to the autoguider optics.
With the off axis probe driven in to access the edge of the focal plane it would be possible to drive the gripper unit into the space occupied by the off axis probe thus causing a collision. This is prevented when controlling the instrument using the normal control system from either the engineering interface or the normal observers interface by a software interlock. However if the instrument hardware is being controlled from software other than the normal control system (for example the DMC300 test utility) it should be noted that it is possible to drive these two mechanisms into each other at considerable speed thus causing serious permanent damage.
3.5 Fibre modules
Two fibre modules have been planned for use with the robot positioner, although currently only one containing large (153m or 2.6 arcsec) fibres has been constructed and commissioned. The following description therefore applies strictly to the 'large fibre module' although it is anticipated that the 'small fibre module' will be very similar in design although perhaps with no fibre connectors on the top end ring.
The large fibre module is stored and lifted to Prime focus access in a purpose made dust proof box. Inside the box are four attachment points which mimic the normal mounting points for the fibre module to allow the fibre module to be fixed firmly within the box for shipping and lifting purposes
The fibre module consists of a X shaped support frame which attaches to the robot on the four corner posts. The fibre module itself fits between this support frame and the robot positioner. The fieldplate is supported by four sturdy legs from the support frame. See figure 3.
The fibres are attached to the circular ptfe coated steel fieldplate by permanent NdFeB rare earth magnets. The fieldplate is surrounded by two concentric rings, the first is steel and provides a loading and parking position for the fibres so that the central field plate can be (carefully) removed for recoating or cleaning, the outer ring is aluminium and provides the fibre pivot points at its outer edge. Embedded in the outer aluminium ring are four reference fibres the position of which may be measured as part of the quality control procedures for the instrument. The steel parts of the fieldplate are coated in a black protective ptfe layer, the aluminium parts are black anodised. The flexible part of the fibre is encased in a lightproof 16 sided box with removable sides. The 15 blue fibre bundles (each with 9 fibres) emerge from the rear of the fibre module in the centre of a coiling drum. The guide fibres are bundled together and emerge in the same place as a single black conduit with a heavy brass adapter on the end.
The fibre bundles are coiled in the drum for storage and uncoiled to be routed across the telescope spider to the fibre connector panel when the instrument is to go on the telescope. The guide fibres are likewise routed across the telescope spider but the brass adapter is plugged into the Westinghouse TV interface box which is located alongside the PFIP controller.
3.6 Instrument optics
See also technical note 1 for a more detailed specification of the optical design of the various optical systems within Autofib-2.
3.6.1 Gripper TV system optics (or fibre viewing system FVS)
When light is fed into the optical fibres from the Westinghouse TV Interface box or the WYFFOS slit area it emerges from the fibre and is bent through 90 by the microprism. This light passes up between the gripper jaws and through the central part of the main gripper body before being reimaged by a transfer lens and folded by a right angles prism. This secondary image is magnified by a microscope objective lens and projected onto a video CCD TV camera (see figure 2). The optical axis of the FVS is aligned to be close to the rotation axis of the gripper theta axis and the CCD TV camera is oriented so that the pixels are aligned with the XY axes of the robot. If the FVS is removed or dismantled these systems will require realignment and recalibration (see technical noe 2 for a detailed list of instructions if this is required).
3.6.2 Sky viewing probe optics (SVS)
The two sky viewing probes (mobile and fixed) use identical optics to relay the image from the telescope focal plane to the input face of a coherent optical fibre bundle. A 45 front silvered mirror is used to fold the light from the Prime focus corrector perpendicular to the optical axis of the telescope, it then passes through a triplet relay lens which adds a slight image scale change to avoid fringing effects with the individual fibres in the coherent bundle. The mobile probe has a 3m long coherent fibre bundle 5mm square at input and output faces. The fixed probe has a 1m long coherent fibre bundle also 5mm square at input and output.
3.6.3 Autoguider head optics
The output faces of the two probe coherent bundles have to be viewed by the autoguider's Peltier cooled CCD camera.
The fixed probe is reimaged 1:1 by a pair of achromats directly onto the CCD. The mobile probe image follows a more convoluted path to enable it to be reimaged immediately alongside the fixed probe image although the physical size of the coherent bundle prevents this being done directly. The light from the mobile probe coherent bundle is relayed to an intermediate image by a triplet lens and it is this image which is positioned alongside the output face of the fixed probe coherent bundle using a knife edge mirror (see figure 6).
3.6.4 Focusing the instrument optics
It is important to ensure that the various optical systems are aligned and focused to be parfocal as the telescope is focused onto the guide and spectroscopic fibres by using the image seen by the mobile and fixed autoguider probes.
The first stage is to set the fieldplate to be precisely parallel, this was achieved for the large fibre module by introducing shims under the mounting points for the fieldplate. This was checked at different orientations and after removal and replacement of the fibre module. The depth of the fieldplate when measured from the mounting face of the robot was set to 201.47mm (without the 2mm stainless steel spacing pad in place). Note that when the second fibre module is commissioned it will be necessary to ensure that its fieldplate is precisely the same distance away from the robot mounting face.
The next stage is to focus the gripper fibre viewing system onto the fibre which must be illuminated from the far end using near IR light (as used in the guidefibre and spectroscopic fibre back illumination). If white light is used then the focus will be wrong when the robot is used on the telescope.
Now with the robot on the testing trolley and using the attached piece of optical bench an artificial star must be projected into the interior of the robot using the projection lenses provided. care must be taken to make sure that the lens mount does not foul the robot as it is moved. By examining the reflected image using a beamsplitter and camera arrangement the artificial star must be aimed at and focused exactly on one of the fibres which for convenience should be positioned at the centre of the fieldplate (0,0).
Now move the mobile probe in to intercept the 'starlight' (viewsky 0 0) and focus the probe optics (the dust cover must be carefully removed) so that a sharp image falls on the input face of the coherent optical fibre bundle.
The above process must now be repeated for the fixed probe. Noting that when the probe is driven into the edge of the field it the centre of the probe appears at a robot position of approximately (-102000,0).
The star image is now focused onto the input ends of the coherent bundles. The next stage is to focus the autoguider optics onto the output face of the coherent bundles. Firstly provide low level illumination to the input face of the coherent bundles and ask the autoguider to grab images continuously. Now focus the fixed probe coherent bundle onto the autoguider using the relay lenses. Once this is aligned do not touch the fixed probe coherent bundle or the pair of achromat relay lenses. To focus the mobile probe coherent bundle move the coherent bundle and the triplet relay lens. Check the overall focus using the artificial star down both probes.
3.6.5 Calibration of offsets between FVS and autoguider probes
The mobile probe is located approximately -55000,75000 microns away from the gripper axis such that the current mobile probe co-ordinate is given by
mobile probe X = gripper X + X offset (where X offset is ~ -55000)
mobile probe Y = gripper Y + Y offset (where Y offset is ~ 75000)
The exact value of the offset is kept as two global variables in the control software (see software manual) and must be calibrated if the gripper indexing block is removed or the gripper unit is dismantled in a major way.
The simplest way to calibrate the offset between the gripper and the mobile probe is to perform it on sky using a bright star (say MV = 10-12). Position a guide fibre at the centre of the fieldplate (0,0) and point the telescope at the star, acquire the star in the guide fibre and let the telescope track. Now use the mobile probe to look at the object which is going down the fibre (af2_viewobject n at the ICL prompt where n is the fibre number being used). Start the autoguider grabbing images continuously in the mobile probe with a short integration time. Centre the image in the mobile probe by moving the robot but not the telescope and check the robot gripper position. The gripper position will then be equal in size but opposite in sign to the offset between the gripper and probe. For example if the above values were correct then to place the mobile probe at 0,0 the gripper would be moved to 55000,-75000.
3.7 Westinghouse TV camera interface unit
The Westinghouse TV camera interface unit is used to interface the guide fibres from Autofib-2 to a standard WHT Westinghouse intensified camera. The unit is mounted on the telescope top end ring adjacent to the PFIP local controller. The camera must be mounted onto the baseplate using the locating keyway before the baseplate is attached to the telescope.
Located on the baseplate is an aluminium box containing the relay optics and back illumination hardware and a die cast box containing some control electronics. Connections to the Westinghouse TV unit consist of the guide fibre bundle from Autofib-2 at prime focus, a 240V AC mains cable from the rear of the PFIP controller and a signal cable from the Autofib-2 controller at GHRIL via the Autofib-2 electrical connector panel on the top end ring. The 240V AC supply must be switched on using a switch on the front of the PFIP controller panel before the guide fibre back illumination can be used and in practice it is left on while Autofib-2 is in use on the telescope.
The guide fibres are reimaged onto the photocathode of the Westinghouse TV camera by a pair of Canon 50mm lenses. On a suitable signal from the Autofib-2 controller a solenoid is switched on which rotates an arm down through 45 blocking the guide fibres from view by the Westinghouse TV. The arm closes a microswitch thus turning on 10 near IR LED's which illuminate the central fibre in each of the guide fibres. The position of the arm in the ON position (and thus the alignment of the back illumination) is governed by a nylon screw and locknut. If the system is dismantled for any reason the back illumination should be checked to ensure that the central fibre of each guide fibre bundle is the one being illuminated.
The electronics for Autofib are contained in one of four places. A simple interconnect card in the VME rack provides the interface between the VMEbus cards and the outside world. A circuit board controls the back illumination of the guide fibres in the Westinghouse camera interface box and the servo amplifier crate takes the motor command voltages from the motion control cards and converts them to motor power. But the majority of the electronics reside within the robot itself.
4.2 VME crate electronics
The electronics in the VME rack consists of a single custom circuit board with picks up supply voltages from the VMEbus backplane and provides an interconnect and line driving/receiving facility for all of the other cards in the rack which communicate with the robot hardware. This circuit contains seven 26way connectors for the seven axes of the Galil motion control cards, two 50 way connectors for the digital IO and ADC cards, a single connector for the motor command voltage signals to the servo amplifier, two 40 way connectors for line driven control and status signals to the robot and a connector for the Westinghouse camera interface unit. Note that currently none of these connectors is labelled and since many are identical great care should be taken when removing any connectors.
The BVME780 cards are connected directly to the rear of the VME crate using miniature coaxial cables. The frame grabber camera channel 0 being connected to the BNC input from the gripper TV camera and the green channel output being connected to the 'monitor' BNC connector to allow viewing of the gripper TV processed signal. The graphics card has input channel 0 connected to the BNC connector fed by the video from the inspection TV within Autofib-2. The RGB and SYNC outputs are connected directly to similar BNC connectors on the rear of the VME crate.
A series of hardware interlocks are implemented in hardware by a series of digital AND gates and status on six limit switches is provided by red LED's on the front panel of the interconnect circuit board. Basically the signal to enable the motor relays is dependent on the inputs from the limit switches on the X and Y axes and the motors on/off signal. If the robot moves into any one of the XY limit switches the motor power is turned off and the brakes engaged. To allow the robot to be driven out of a limit switch an override signal is provided which bypasses the effect of the limit switches.
4.3 Robot electronics
The following sub sections should be read in conjunction with the corresponding circuit diagrams for which they provide a limited description with particularly relevant points noted.
4.3.1 Power supply control circuit board
The mains 240V AC input to the robot positioner is fed directly into this circuit board with a switched output 240V AC being fed to the switched mode power supply used to generate low voltages for the rest of the instrument. The 240V AC mains is used to power a small transformer and rectifier which generates 12V DC, this is then used to provide a stable 5V DC supply for use powering ttl integrated circuits. A limited part of the main control circuit board is fed with a permanent 12V DC supply to power the few integrated circuits necessary to switch on the main instrument power.
This circuit board and the limited part of the main circuit board provided with the permanent 12V DC supply remain live at all times even with the instrument power switched off. When the signal is sent to switch on instrument power it is routed directly from the main instrument circuit board to the power supply control board where it is used to flip a relay to switch the 240V AC output to the main switched mode power supply on. Once the switched mode power supply is on all of the instrument electronics become live.
4.3.2 Local voltage supplies
The three voltages required locally at the robot are generated by a switched mode power supply available from RS and Farnell. The 240V AC input for this unit is controlled by a relay in the power supply control circuit board, this allows control of the voltage supplies within the instrument (+5V, +12V, +24V).
4.3.3 Main instrument circuit board
The so called main instrument circuit board links the various bits of electronics within the robot to the outside world via the two 41-way signals cables and the motor power cable. It therefore has a large number of connectors which are labelled on the lid of the box containing the circuit board with the location and number of the cable.
This circuit board deals mainly with the rerouting of control and status signals and control of the 3 large relays for switching the motor power and inspection light power. However some aspects of this circuit board deserve a little description. The air pressure status is set to change from 0V to 5V at about 30-35psi using a preset resistor, this is below normal operating pressure but high enough to signal a failure of the air supply without triggering due to variation in the air pressure. The temperature sensor circuit is connected to the main circuit board by soldering the cables directly without a connector to reduce noise. The signal from the temperature sensor is amplified and sent down a twisted pair signal cable as a voltage difference to the VME interface card where a matching circuit decodes it and feeds a temperature dependent voltage to the ADC card for reading by the control software. The transmitting and receiving parts of this circuit have been matched and calibrated to give a good representation of the instrument temperature over the -5 to +25C range. After setting the gain and offset of the matching transmitter and receiver to cover the temperature range with a voltage of 5V the ADC reading must be checked at two widely different temperatures to enable the conversion from voltage to temperature to be made. The current conversion factor was made by reading the ADC with the instrument at -3.2C (2224 or 0x8ad) and 24.2C (822 or 0x366) giving a conversion of:
temperature = (-0.0195 *ADC reading) + 40.168
When the instrument power is turned off from ICL or the engineering interface most of the status signals float to a random level and cannot be relied upon. However a small part of the main circuit board is fed by a 12V supply from the power supply control board even when the instrument power is switched off to enable correct determination of the instrument power status and to enable the power to be turned back on again safely. Note that all encoders, limit switches and other lights and power are removed. The control software ensures that sensible status responses are made of the instrument power is turned off.
4.3.4 Line driver circuit boards
Three identical circuit boards exist in the robot to convert ttl signals to line driven signals suitable to be transmitted over 20 metres from the instrument at Prime focus to the VME rack located in the GHRIL control room. The three circuits are used for the rotary encoder, limit and home switches for the X and Y axes and the rotary encoder, limit and home switches and motor power for the off axis probe.
For the X and Y axes the line driver circuit is located close to the corresponding motor encoder unit. The limit switches and home index switch are connected to the circuit using three 3-way mini din plugs and the line driven output from the rotary encoder is connected by an 8-way mini din plug. Parts of the circuit necessary for the off axis probe are redundant in this instance and the circuit not populated. The limit and home index switches are line driven and all output signals are routed back to the main instrument circuit board via a 25-way Dtype connector which also provides the 5V supply for the circuit board.
For the off axis probe the line driver circuit is fully populated but instead of connecting just a rotary encoder into the circuit, the motor encoder unit is connected using a 6-way mini din plug. Limit switches are connected as before but there is no home index switch to be connected. This time both encoder phases as well as the limit switch signals are line driven before being routed back to the main instrument circuit board. In this instance the motor power and circuit board 5V supply are both fed via the 25-way Dtype connector.
4.3.5 Temperature sensor circuit board
This circuit board is mounted so that the temperature sensor is held in contact with the baseplate of the instrument. This is the only circuit board that has no connector to reduce the effects of noise. The four leads to the main instrument circuit board (where the amplifier is located) are soldered directly to pins at both ends.
4.3.6 Gripper circuit boards
Two circuit boards are used on the gripper unit connected by a flexible ribbon cable to allow for the movement of the Z axis. The theta circuit board is attached to the side of the main gripper body and moves with the gripper in the Z direction. This circuit board is used mainly to connect various sensors to one point to allow all the signals to be routed via a ribbon cable to the second gripper PCB.
The theta PCB contains four Molex type connectors and one IDC 20-way header connector. An optoswitch acting as a home index switch for the Z axis is mounted directly on the circuit board with a similar switch for the theta axis connected by a 3-way Molex connector. End of travel limit switches for the Z axis (normally closed micro switch units) are connected using two 2-way Molex connectors. If the gripper is driven too far in the Z direction the switches are opened and the Z motor power which flows through both switches is cut preventing further movement in that direction. Reversing out of the limit switches is facilitated by two diodes across the limit switches allowing the reverse current to flow even when the switches are open. The theta motor encoder unit is connected to the theta PCB by the final 6-way Molex connector. All signals, supply voltages and motor power are carried to or from the theta PCB by the 20-way ribbon cable to the second of the gripper circuit boards.
The Z axis PCB is mounted on the main gripper casework and moved neither in theta or Z. The PCB contains three connections, a 25-way Dtype connector for a cable routing signals back to and power from the main instrument circuit board, a 20-way IDC connector for the ribbon cable to the gripper theta PCB and a 6-way Molex connector for the motor power and encoder signals to the Z axis motor unit.
All signals for both Z and theta axes are line driven by two surface mounted line driver chips (used due to the limited space available) before routing them back to the main instrument circuit board. The Z motor power is routed via the theta PCB even though the Z motor connector is located on the Z PCB as the end of travel limit switches may only be accessed in this manner.
4.3.7 Opto switch circuit boards
Numerous near IR transmitter-detector optoswitches are used for home index and limit switches throughout Autofib, these are all mounted on identical PCB's which the necessary resistor and capacitor. All these circuits have the same cable and connector attached (except for the length of the cable) which is available from RS (RS 304-560).
4.4 Westinghouse interface electronics
The electronics for the Westinghouse camera interface unit are contained within a black painted die cast aluminium box mounted on the lid of the Westinghouse interface unit. The connections to the outside world are a 240V mains connector and a 4pin signals cable. The signals cable consists of a line driven ttl control signal and a line driven ttl status signal. The control signal is converted back into ttl and used to switch a relay. The relay in turn switches on a solenoid which swings an arm in front of the guide fibres and closes a micro switch. Once the micro switch has closed, 10 near IR LED's are switched on. The status signal reports whether the LED's have successfully been turned on.
4.5 Servo amplifier electronics
The five servo amplifiers required to power the five dc servo motors used in Autofib are housed in a 3U 19inch rack. 240V AC mains is fed into the rear of the amplifier feeding three transformers and rectifiers to provide rough 24V(?) DC voltages for the servo amplifier modules. One transformer provides power for the X axis, one provides power for the Y axis and the third transformer provides power for the Z, theta and off axis probe motors. The reason for this split is that the X and Y axes require the most power and the Z, theta and probe motors are never driven together. The 24V supply for the Y motor is also used to power a relay to close a normally open switch which acts as a status switch to indicate that the power to the power amplifier is switched on.
The power amplifiers themselves are of two types, high current types are used for the X and Y axes and are in theory interchangeable (if the trim potentiometers are set correctly). The Z, theta and probe motors are powered by three identical lower current amplifiers. The trim of these three amplifiers is not so critical but if swapping for a spare amplifier it is good practice to check that the potentiometers on the spare match those of the defective unit.
The control of the power amplifiers comes from the motor command voltages from the Galil motor control cards. These command voltages are fed via five miniature coax cables from a 10 way connector on the back of the rack to the 32way connector on the back of the servo amplifier. The motor power output is fed from the 32way connector on each servo amplifier to a high current 12(?)way military connector on the back of the rack.
Swapping one power amplifier for a spare (one of each type is kept as a spare) involves removing the power amplifier from the rack, swapping over the front panel and checking visually that all 10 or so trim potentiometers are set identically to the redundant unit before inserting it back into the rack.
The cabling of each transformer, amplifier set is identical, for reference see figure 4 and the PVP power amplifier manual.
There are three connections to the servo amplifier, a standard 240V AC IEC mains cable with 10A fuse and two military metal shelled connectors for command voltages into the servo amplifier and motor power out of the servo amplifier. The command voltages for each axis are carried inside miniature coaxial cable with the outer sheath grounded for noise immunity. The connector for carrying the motor power is of a high current type rather than a normal signal carrying type. The two connectors are or different types and patterns and so cannot be confused.
Also contained in the servo amplifier crate is a small 24V relay. This is connected to the Y axis 24V supply to close a switch when the power to the amplifier is switched on so that the control system can detect if the technician has switched on the power amplifiers or if a fuse has blown. The relay switch is connected to the main control circuit in the robot by two wires running inside the motor power cable where a line driven ttl signal is generated to indicate the state of the servo amplifier.
The cabling of the VME crate includes connections between the individual VME cards within the crate and cables from the cards to the connectors on the rear of the crate. Connections within the crate are covered in other sections (4.2). All the status and control signals to the robot go from the custom interface card to the back of the VME crate via two 40 way twisted pair ribbon cables where they are attached to 41-way military connectors with different connector shell orientation and colour coding (red and blue). Motor command signals go from the interface board to a 10 way military connector located on the rear of the VME crate via a ribbon cable. The Westinghouse TV interface unit signals go from the custom interface circuit to the 4 way military connector on the rear of the VME crate via a ribbon cable. All video signals use miniature coax cables between the Cohnex connectors on the BVME780 card and the BNC connectors on the rear of the VME crate.
5.2 Robot internal cabling
All cables within the body of the robot positioner are connectorised with the exception of the temperature sensor which is soldered directly to the main instrument circuit board and the flying leads on the various optoswitch sensors and magnetic brakes which are soldered directly to the component at one end and connectorised at the free end. Within the instrument extensive use is made of DIN and mini DIN connectors, Dtype connectors and plastic shelled QM multipole connectors, all available from RS.
Wherever possible the cable identity is labelled at its socket on the box containing the circuit board it is plugged into. All internal cables are labelled which an identity number which can be checked in the cable schedules (Appendix A).
5.3 External cabling
The external cables are very limited in number and are detailed in Appendix A. Here is a list of the cables and their purpose.
Cable number Purpose 46 Motor command signals from VME crate to servo amp. 50 Motor power from servo amp. to top end ring panel 51 Signals from VME to top end ring connector panel 52 Signals from VME to top end ring connector panel 53 Signals from VME to top end ring connector panel 54 Motor power from top end ring connector panel to robot 55 Signals from top end ring connector panel to robot 56 Signals from top end ring connector panel to robot 57 Signals from top end ring connector panel to Westinghouse TV interface unit no number Servo amplifier mains cable no number Robot mains cable no number VME crate mains cable no number Westinghouse TV unit mains cable
In addition there are two 75 coaxial cables between the robot via the top end ring connector panel to the VME crate, one carrying the gripper video signal and one the inspection TV signal, these cables are colour coded red and blue for ease in identification. It is important to use 75 cables as the normal 50 cables will not work correctly with the frame grabber due to the length.
At the rear of the VME crate there will remain a number of unused BNC connectors for various display purposes. It will be most useful to have the engineering mimic display connected to the RGB and SYNC connectors and a black and white monitor connected to the 'monitor' BNC connector to display the processed gripper video.
6.1 Instrument cleanliness
It is absolutely essential that the internals of the robot positioner and the outside of the fibre module are kept spotlessly clean and free from dust, grease and metal swarf. The latter has a habit of finding its way to the base of the fibre buttons and sticking to the powerful permanent magnets. This is so far the cause of the only reliability problems associated with the instrument.
6.2 Gripper Lubrication
The gripper pneumatic piston may need regreasing at infrequent intervals. The period for doing this is unknown but regular observing of the speed and smoothness of the gripper jaw opening and closing should be enough to warn of a potential problem. See section on gripper dismantling for a guide on how to remove and dismantle the gripper. A small quantity of high quality vacuum grease it the best lubricant for the piston. On no account should WD40 be used as this dissolves away existing grease, migrates to the gripper optics and although it gives an immediate improvement it does not last (it has been tried!)
6.3 Prevention of surface rust.
After shipping from the UK a slight build up of surface rust on the linear bearings, ballscrews and hardened steel roller guide was found and cleaned off. To prevent further build up of rust these items should be wiped down with a lint free cloth liberally sprayed with WD40. No excess lubricant should be allowed to remain. The gripper should be moved during this process to access the parts of the bearings and ballscrews hidden by the gripper unit in its normal parked position. This procedure should be performed every time the instrument is tested in the test focal station before going on the telescope.
It would be impossible to give instructions on the dismantling of the complete robot to its constituent components so a brief note on particular problems and a few warnings are given for the main subassemblies.
Extensive use is made of metric Allen bolts and care should be taken to use the correct size particularly for the gripper and various parts of optics as these make use of extremely small grub screws for which the set of Allen keys (RS 181-626 or similar with 0.9,1.27mm Allen keys) is essential.
7.2 Dismantling parts of the gripper unit
7.3 Dismantling parts of the XY carriage
The main points about the XY carriage to note are the fact that the Y axis carrying plate is doweled to the X axis linear bearing slides and will be exceedingly difficult to remove.
If any of the home or limit switches are removed or knocked out of alignment then the software offsets for these items will need to be checked and revised. Similarly if either of the linear bearings is removed from its supporting plate then the orthogonality correction and straightness correction will become invalid when the instrument is reassembled and will require a full recalibration using multiple grid tests.
If the magnetic brake assemblies are removed for any reason it is important to ensure that they are reassembled with the correct spacing between the end plate and magnet. To do this bolt the end plate to the ballscrew and the magnet onto the baseplate with the power off and a 150m shim between the two (check this dimension). Once both halves are tightened down remove the shim and check for correct operation particularly brake release and clamping when power is applied and removed.
7.4 Dismantling the off axis probe
To remove the off axis probe as a complete sub unit from the robot baseplate will require the 2mm spacer to be removed from the mounting face of the baseplate and the robot to be mounted vertically on its handling frame and trolley (not the testing trolley) as the fixing bolts are accessed from the rear of the baseplate.
As for other items the location of the limit switches is important and should be carefully checked if the probe unit is dismantled. Upon reassembly the tension on the drive belt should not be too high to avoid unnecessary wear in the motor bearings.
7.5 Dismantling the fibre module
After removing the fibre module from its protective box it should be placed fieldplate upwards on four stools or chairs, one under each mounting point. This will support the fibre module horizontally at a convenient height. This will provide easy access to clean the fieldplate and fibre buttons (particularly the button handle and magnetic base).
To access the internal parts of the fibre module the cover may be removed by undoing all of the button head Allen bolts from the upper rim and the 10 or so bolts from the inner edge of the cover after which the cover may be carefully removed.
Each group of 9 fibres is held in place at the pivot ring by a plastic retaining strip which clips over the pivot wall.
The fieldplate may be removed for cleaning or recoating by removing four countersunk bolts and using a magnetic clamp from the optics lab carefully lifting the fieldplate out of its support ring. Note that all the fibres must be in the loadring or parked for this operation to be undertaken.
To remove a set of 9 fibres remove the cover and fieldplate as detailed above. Carefully remove the plastic strip which keeps the fibre tubes in the correct pivot slots and group the 9 fibre together by placing the magnetic buttons on both sides of a thin piece of steel approximately 10mm by 10mm by 1mm thick. Now carefully feed the bundle of fibres from the outside of the pivot circle back under the fieldplate (taking care not to tangle with the other fibres) and back up through the central hole where the fieldplate has been removed. Remove the retaining nut from the bottom of the fibre module and carefully thread the bundle of fibres out from the fibre module.
Most error conditions will be self explanatory with no further actions necessary before proceeding other than to enter the correct number or type of parameters next time. However some commands may fail with an error likely to leave the instrument in an unsafe state or prevent any further action with the instrument. This section deals with these types of errors and the steps required to extract the robot from the error and give an understanding as to the cause of the problem. Please read the whole section before trying to correct any problem as there may be more to the problem than is at first apparent.
In all cases when dealing with errors occurring when moving fibres the back illumination should be turned on. If a fibre remains in the gripper jaws then it will appear as an out of focus doughnut on the gripper TV monitor, if the fibre is safely on the fieldplate it will appear in focus on the gripper TV monitor.
If you try a recover command after failing to move a fibre and it returns with a FIBREINJAWS error then a fibre is still held in the gripper jaws.
If you try to move a fibre and it returns with a LOSTFIBRE error then a previously lost fibre has not yet been recovered properly. This should be sorted out before proceeding.
In general an error while configuring the fibres for the next field will terminate with two errors, a CON_SETUPERR to indicate that an error occurred while configuring the fibres and a more specific ROB error indicating the exact problem. Also a report should be made of the fibre number where the problem has occurred. Entering the whichfibre command from the engineering interface will also return the number of the fibre where the problems occurred.
8.2 Failure to pick up a fibre
Any one of four error conditions may occur when picking up a fibre, in general these are not too serious as the fibre is left safely on the field plate and the error is simple to recover. However if this happens frequently there is obviously something wrong and an attempt should be made to rectify the problem.
The immediate action should be to identify and understand what has happened and recover the lost fibre so that observing can continue.
NOSECONDFIBRE The robot thinks that it is already holding a fibre and is therefore refusing to pickup a second fibre. It must be determined whether the robot is really holding another fibre. If the robot is not holding a fibre issue a resetcurrfib command followed by a recover command at the engineering interface. This test should not fail if a previous error condition has been handled correctly, it is most likely to occur after a put-down error (see next section) has been recovered incorrectly.
NOKNOWNFIBRE The robot has been driven to a x,y position to pickup a fibre but does not know of any fibre within 500m of this position and so aborts the pickup. Issue a recover command to proceed. This check is really a hangover from the early days of testing the software and has not occurred in recent use.
NOFIBREVISIBLE The robot does a quick test to check if it can see the fibre it is attempting to pick up. This can fail for one of two reasons: one, the fibre is not visible because it is in the wrong place or the back illumination has failed, or two, the centroiding routine used cannot cope with overilluminated fibres and occasionally rejects a fibre even though it is visible if the fibre is particularly bright. This latter case can occur if the back illumination has only just been turned on and has not settled down to its normal level of illumination.
MOVETOPLATEFAIL After moving the Z axis towards the fieldplate the robot checks that the Z movement was successful and that it has moved to the correct height. This test might fail if the Z is misbehaving and the Z servo loop times out or trips out due to an excessive position error or if the Z axis meets an obstruction and stops moving before reaching the correct position. To proceed issue a recover command and determine which of the above problems has occurred.
8.3 Failure to put down a fibre
This failure of the robot to place a fibre on the fieldplate successfully is potentially the most serious error condition. Fibre placement errors are of three types and it is important to understand and be able to identify the difference between them.
NOFIBTOPLACE The robot does not think that it has a fibre in its jaws to place on the fieldplate. This error condition is a hangover from early testing days and should not occur as a pickup error should already have caused the robot to abort the fibre movement. If this error is seen check that no fibre is in the gripper jaws and issue a recover command.
ROBBADCENTROID The robot failed to centroid the fibre during the iterative placement of the fibre on the fieldplate. The good news about this error is that the fibre is on the fieldplate, it has probably jumped out of the gripper TV field of view or the backillumination has failed. To proceed check that no fibre is held in the gripper jaws and issue a recover command noting that it will do a spiral search around the pickup position before performing a spiral search around the put-down position (which is where the fibre is almost certainly located).
LOWERTOPLATEFAIL The robot checks the position of the Z axis as it lowers the fibre to the fieldplate to check for obstructions or a failing Z axis. The main point about this error is that the Z axis is raised away from the fieldplate retaining the fibre in the gripper jaws. Check that the fibre is visible as an out of focus doughnut in the gripper jaws by turning on the back illumination.
It is important that this error is handled by someone experienced with the use of the engineering interface to the instrument. The first stage is to get the fibre placed on the fieldplate safely, this can be done by use of engineering commands or as a last resort by manual intervention which will cost approximately 60 minutes of lost time.
For the former method we can assume that the current robot position is the put-down position and this is a legal position for the fibre as the anticollision checking will have already been performed. We require to carefully place the fibre on the fieldplate using low level engineering commands. Get an OS/9 session open on an Xterm and start the engineering interface. The command sequence to lower the fibre to the fieldplate is as follows.
local[return]
extended[return] (prompt for password)
zmove 10000 [return]
gripper open[return]
zmove 0[return]
restricted[return]
remote[return]
This assumes that all the commands work correctly, the zmove command values are not too important as the engineering interface will set these rather large values to the maximum value allowed for the current plate height setting. Before opening the gripper jaw check the Z axis height to make sure it has reached the correct value. if not repeat the command, if you still cannot get the Z axis to descend to the plate issue the zmove 0 command and seek help as it would appear that the robot is unable to lower the fibre to the fieldplate.
By now the fibre should be safely on the fieldplate. Issue a resetcurrfib command followed by a recover command before proceeding.
If this method encounters problems or manual intervention is required proceed as follows. bring the telescope down to AP3 and replace two of the bolts holding on the fibre module with a length of M10 studding. Remove the Autoguider controller and cabling and the cable ties for the first part of the fibre cables. remove the second pair of fibre module retaining bolts and carefully slide back the fibre module just enough to reach in with a hand. Open the gripper jaws (this can be done by rebooting the micro or switching off the instrument power or preferably from the engineering interface) and the fibre will be released. Place the fibre in the load ring before replacing the fibre module and autoguider controller and recabling the prime focus area.
To recover the fibre, wake up the robot and from the engineering interface enter the 'ringfibre n' command (if you failed on fibre n) followed by a 'viewfibre n' and a 'loadfibre n 2' commands. Now enter a resetcurrfib command and a resetlostflag command to reset the flags that prevent another fibre being moved until the lost fibre was recovered.
8.4 Other fibre movement errors
GRIPTOOLOW One error which may be caused by the same mechanical problems as cause the MOVETOPLATEFAIL and LOWERTOPLATEFAIL errors can occur when the robot is lifting a fibre away from the fieldplate. The robot checks the current Z axis position before moving the fibre to its new location, if the Z axis has tripped out due to servo problems the Z axis may not have reached a point far enough away from the plate to be allowed to move to its new location and the robot will return a GRIPTOOLOW error to indicate that it has been asked to move the robot with the gripper to close to the fieldplate. It is important to note that this error will result in a fibre remaining in the gripper jaws and recovery should follow the procedure outlined in the previous section for the LOWERTOPLATEFAIL error except that the fibre is at its pickup position.
8.5 Extracting robot from limit switch
The most likely explanations for the robot ending up in a limit switch are being jolted during the installation at prime focus, failure of a power amplifier thus removing motor power. Much less likely are a brake failure or a motor runaway. If the robot is in one of the XY or probe limit switches a red LED will illuminate on the micro computer indicating which axis is at fault.
This error condition will probably be noted either when powering up the motors in that the motor power fails to turn on when the robot is in a limit switch. This also means that the motor power will automatically be turned off if the robot runs into a limit switch.
Using the EXTRACT command from the engineering interface temporarily overrides the motor power fail-safe in a controlled manner while the robot is safely driven out of an X or Y limit switch. To use this command the instrument power must be switched on. After a successful extraction of the robot from the limit switch the robot will not have the motor power switched on and so must be woken up in the normal manner. If this recovery fails the robot will return a ROBSTUCK error, in this case repeat the EXTRACT command before assuming that something more serious is wrong. One problem that has occurred in the past is that the status relay in the servo amplifier has failed so that the micro computer was not aware that the servo amplifier is not switched on. In this case the micro computer attempted to switch on the motor power and release the brakes and the robot rolled into a limit switch and hardware stop. Of course the extract command did not work as there was no motor power but as soon as the servo amplifier was powered up the extract command worked first time. Note that this situation would not have occurred if the status relay had not failed.
To extract the robot from a probe limit switch it is simply necessary to reinitialise it, the INIT command from the engineering interface or simply the AF2_WAKE command from ICL.
Manual extraction from a limit switch is also possible, after removing the fibre module and all electrical connections to the robot (mains, motor power and signals) a 24V supply may be connected using the cable provided to the socket adjacent to the fibre module proximity switches. This will override the X axis brake allowing the X axis to be wound out of its limit switches by hand. The probe may be would out of its limit switches by hand with no other intervention. The Y axis has no override for the brake but can usually be wound out far enough by hand to extract it from a limit switch.
For storage a handling trolley is provided. This handling trolley accepts the robot positioner with its lifting frame attached in an almost upright manner ready for lifting by crane. The internal components of the robot are protected by a yellow cover which bolts on in place of the fibre module. When the robot positioner is lowered onto the handling frame it should be clamped in place by two levers at the top of the handling frame.
The fibre module is stored in a dust proof container which has attachment points to allow it to be lifted up to the prime focus access point ready for installation on the telescope. Under no circumstances should the bare fibre module be craned up from the instrument store to prime focus access.
For testing Autofib in the test focal station it may be necessary to move Autofib onto its testing trolley which allows the instrument to be moved in most of the directions it would experience on the telescope. The robot positioner should be lifted from the handling trolley on its lifting frame and the handling trolley moved out of the way. The testing trolley should be swivelled so that the circular mounting plate is vertical and the trolley moved adjacent to the robot positioner . The lifting frame may be lowered carefully to the same height as the handling trolley. After removing the yellow protective cover the robot may be centred on the circular mounting plate and bolted in place with four M10 Allen head bolts. Replacement of Autofib on the handling trolley is a reversal of this procedure.
When lifting Autofib up to Prime focus access it should be lifted using the lifting frame and the handling frame should be left on the ground floor as the lifting frame was not designed to handle the additional weight of the handling frame during lifting.
When testing and setting up the instrument in the test focal station or the electronics lab it may be necessary to run the instrument independently from the observing system.
The VME crate and instrument should be connected using the 10m long set of cables normally used for connecting the instrument to the Autofib-2 top end ring connector panel. This does mean that the full cable run is not being used for testing but this should not affect the operation of the instrument. No connection to the NIU 25way D type connector is necessary or advised if the instrument is not to be tested from the observing system. This avoids any possibility of problems due to two sets of commands to the instrument. The engineering terminal should be connected to its 25way D type connector on the rear of the VME crate.
The instrument power (240V) must be switched on before the instrument control software is started. After booting the micro computer log in as username Autofib and start the control system (startall[return]). Interactive control of the instrument is provided by an engineering interface which may be started by typing eng[return].
In normal use a software interlock will prevent the motors being energised and the X and Y axis brakes being unlocked unless the proximity switches detect the presence of one of three possible fibre modules in place on the robot. This is to prevent the possibility of the robot being powered up when the protective yellow cover is in place or when no fibre module is in place and somebody may be working in the vicinity of the gripper. Remember that the robot will not stop if somebody places their hand in the path of the robot! If it is essential that the robot be operated without the fibre module in place then one of the proximity switches may be fooled by placing a M8 nut and bolt onto the thread of the orange tipped proximity switches visible adjacent to one of the fibre module attachment points on the robot.
Alternatively very low level control of the instrument may be achieved by booting the micro computer and logging in as username Autofib then controlling various instrument functions using several utility programs to switch TTL digital I/O lines, move motors, enquire status etc. It should be noted that this is a very risky procedure with the strong possibility of damage to the instrument if the user is not fully of the dangers of this instrument. However this may be useful when trying to identify software or mechanical problems or when testing the DC servo loops.
10.2 Performing gridtests
Grid tests are performed to check and calibrate the non-orthogonality of the X and Y axes and the straightness of the X and Y axes. This procedure also gives a good measure of the effectiveness of the thermal compensation model used to allow for varying instrument temperature. The robot is used to measure a grid of dots on a photographic glass plate and the results are compared with those obtained when measuring the same glass plate in COSMOS. The difference between the two sets of measurements represent the instrument flexure, straightness and thermal expansion.
To prepare Autofib for performing grid tests the instrument should be mounted on its testing frame to allow the instrument to be moved to different gravity vectors. See section 9.0 on the different handling frames for this procedure.
The eight screw clips should be removed from the support frame before attempting to mount the glass plate in the frame. The test grid glass plate should be carefully mounted on its support frame in the correct orientation (check rotation and back to front!). The four corners of the glass plate are marked with a letter corresponding to the letter stamped in the aluminium support frame. The eight clips may now be replaced making sure that the glass plate is secure but also that the clips have not been over tightened.
When the glass plate is secure on the support frame the entire assembly may be attached to the robot positioner in place of the fibre module. The orientation of the support frame relative the robot is absolutely crucial for the results to make any sense. The two 'legs' must be located over the Y axis motor (check this). See figure 5 which shows the correct orientation of the glass plate and the robot as seen from above when the robot is mounted on the testing trolley and turned to a horizontal position.
With the glass plate in place, the robot can be controlled from the normal engineering interface to locate the spot nearest the centre (0,0) of the robot. If you cant see the spot with the robot gripper TV system make sure that the lights for the glass plate are turned on and try looking at the direct video output from the robot rather than the frame grabber processed image. Under no circumstances should the Z axis be driven near the glass plate! Once the centre spot has been located the robot should be driven an integer number of 10000 micron steps in one direction and then back in the reverse direction to check the squareness of the glass place with respect to the robot co-ordinate system. (For example if the centre spot is located at 500,500 the robot could be driven in the X direction to view the spot located at 100500,500 and -99500,500 for a total spacing of 200mm). To square the glass plate the 4 mounting bolts should be slackened and the entire support frame rotated on the slack bolt holes to improve the alignment before retightening the bolts. It may occasionally be necessary to omit one of the bolts to obtain the best alignment, this should not cause problems provided that the remaining three bolts are tightened fully. The reason for this alignment procedure is to ensure that when the robot drives to the expected position of the spots the image of the spot falls completely within the TV field of view so that the centroiding algorithm works correctly. (NB See section 10.1 about running the robot without a fibre module in place).
Once the glass plate is in the correct position put the robot to sleep and shut down the micro computer software and to be sure that nothing is left running reboot the micro computer. Log in as Autofib on the micro computer and change to the directory you want the output data file to be placed in then type 'grid' at the OS/9 command prompt. This starts a program running that controls the robot hardware in a predefined manner for measuring the glass plate, the program takes approximately 30(?) minutes to run. The user will be prompted for a number of parameters the program requires:
10.3 Repeatability tests
Repeatability tests involve using the robot to repeatedly measure the apparent position of a specified fibre a large number of times moving to a different random position within the travel of the robot between each measurement. This procedure is performed using a single command from the engineering interface.
At the OS/9 prompt type eng[return] to enter the engineering interface and wake up the instrument in the normal manner. Arrange for the fibres to be back illuminated in a nice even manner. Choose a fibre as near to the centre of the field plate as possible and view this fibre with the gripper TV system to check its illumination (viewfibre n[return]).
Two types of repeatability test may be performed, static centroids where the robot remains stationary between measurements as a test of the centroiding routine and random movements between the repeated measurements. The commands for these two modes are:
STATIC fibn n
RANDOM fibn n
where fibn is the number of the fibre chosen and n is the number of measurements required. The maximum number of measurements that can currently be requested is 5000, additional measurements may be made by repeating the command providing that the robot is not reinitialised between the commands.
The measurement data is written to a file random.dat in the /h0/autofib/logfiles directory. The data is appended to an existing file or if no file exists a new file is created. It is recommended that the previous random.dat file is renamed to avoid confusion between different data sets whenever a repeatability test is performed. The measurement number and the x,y position of the fibre are written as an ASCII string, one line per measurement in the data file.
10.4 Tuning DC motor servo loops for the X and Y axes
The procedure for tuning the DC servo motors is critical as poor servo performance will slow the robot down considerably, reduce its accuracy and reliability. The tuning for the X and Y axes is the most critical, tuning the Z, theta and probe axes is much easier. Tuning the servo loops may be required is the robot is dismantled for any reason, if the servo loops seem to be unresponsive, sticky or are drawing excessive current, if the robot is becoming noisy and resonating during motor movement or if during the life of the instrument a decision is made to try and improve the speed of the instrument. Retuning due to problems should not be treated lightly as it is almost certainly indicative of a serious problem elsewhere and a thorough check of the instrument should be made.
The servo motors should be tuned with the fibre module removed from the robot and the robot mounted on the testing trolley to allow the instrument to be rotated through various attitudes.
The tuning process consists of two parts, software parameters for the servo loop and hardware settings for the servo loop. Primary among these is the feedback gain of the servo loop which has a hardware setting in the power amplifier and a software setting in the control software for the Galil motion control card. Also important is the software integrator which automatically increases the software gain to nudge the servo loop the last few microns in small moves.
The hardware settings for the servo loops are made by adjusting trim potentiometers behind the front panel of each servo amplifier (accessible through a small hole). Software settings may be changed by telling the Galil motion control card what the new settings are by issuing a command. These settings are automatically communicated to the Galil card on start-up of the normal control software but must be set manually when controlling the motor using the utility program (DMC300). The normal control software reads in the current software servo loop parameters from the data file (/h0/autofib/datafiles/af2params.dat) when it starts running so having decided on and tested new values for the servo loop it is a simple matter to change the values used by editing the data file. As a good procedure the old value should remain in the file as a comment to allow a history of the servo parameters to be kept.
Before starting to test the servo loops it is important to ensure that the robot is functioning correctly by performing the normal start-up checks using the engineering interface to the normal control software. It is also important to ensure that the user is familiar with the command set used by the Galil motion control cards (see Galil manual).
Firstly the normal control software should be shut down as communications from two sources may cause problems with one or other of the processes trying to control the Galil card (Note that even when the normal control software is idle with motors switched off, it is regularly polling the Galil cards for the status of each motor axis). To start testing the servo loops run the DMC300 test program (type dmc300 [return] at the OS/9 prompt) remembering that all commands must be entered in upper case. Select one axis for testing by entering the name of the axis as prompted.
The DMC300 cards have all been set up so that the motors are powered off when the VME power is switched on. Therefore the first commands to the DMC300 axis should be to set up rough servo loop parameters, set the current position and turn the motor on. The values for the servo gain and acceleration should be obtained from the current af2params.dat data file as a starting point. Enter
GN n [return] set servo loop gain
AC accn [return] set current acceleration
SP 2000 [return] set current maximum speed very low
DH [return] define current position as zero
SV [return] start motor servoing
The motor should start servoing about its current position. The next step is to find the index mark for the axis in question to define its zero point correctly.
FE [return] enter a find edge command
BG [return] begin moving
The motor should turn, moving the current axis until it finds the home index point and then stop immediately. Define the new robot position as the zero point of the encoder and set normal servo parameters by issuing the following commands (the current operational servo parameters are normally a good starting point)
DH 0 [return]
SP speed [return]
AC accn [return]
The motor may now be put into a repeat cycle to allow the effect of changing the instrument servo parameters and instrument attitude to be checked. The motor movements should be allowed to finish before issuing further commands.
PR -100000 [return] move the robot relative to current position by -100000 steps
BG [return] start moving
PR 200000 [return]
WT 2000 [return] delay 2 seconds
RR [return] repeat in reverse direction
BG [return]
With the robot cycling between position -100000 and 100000 the instrument may be moved to a different attitude or the servo parameters changed. The speed, acceleration, gain and integrator values may be changed while the motor is moving, also the current position may be checked at any time but it is particularly useful to check the position at the pause in each cycle to ensure that the position achieved is that requested.
The aim of the tuning procedure is to arrive at a set of parameters for the servo loop which will work at any instrument attitude and allow smooth, fast motor movement without shaking the instrument to pieces, juddering, hunting for final position or resonating about the final position.
Further details on the command set for the Galil cards may be found in the Galil DMC320 manual.
10.5 Additional notes for tuning Z, theta and probe servo motors
Tuning the Z, theta and off axis probe servo motors follows much the same procedure as for the X and Y axes with a few exceptions. These axes do not require the critical tuning of the X and Y axes and if tuning is required it is probably indicative of problems elsewhere.
To define the zero point of the off axis probe the motor should be switched on and driven into the reverse limit switch (PR -100000), then it should be driven out of the limit switch by 500 encoder counts (PR 500) before defining the zero point (DH 0). The travel of the probe is then between 0 and approximately 70000, so the commands for cycling the motor (with the motor starting at zero) are:
PR 70000 [return]
WT 2000 [return]
RR [return]
BG [return]
The Z axis has a home index switch and so the normal method for defining its zero position works but its travel range is between 0 and ~5000. The command sequence to cycle the Z motor is therefore (starting with the motor at zero):
PR 5000 [return]
WT 2000 [return]
RR [return]
BG [return]
The theta axis also has a normal home index switch but may be allowed to travel in either direction without limit. As a guide the full circle travel is equivalent to 21600 encoder counts, to cycle the motor between -5400 and 5400 (half a turn) the following command sequence is used after initialising the motor in the normal manner:
PR -5400 [return]
BG [return]
PR 10800 [return]
WT 2000 [return]
RR [return]
BG [return]
Figures
Figure 1 Instrument structure and housing
Figure 2 Gripper unit cross section
Figure 3 Fibre module plan and cross section
Figure 4 Servo amplifier connections
Figure 5 Grid plate orientation for gridtests
Figure 6 Autoguider relay optics
Appendix A Cable Schedules
Appendix B Circuit diagrams