hostmot2 − LinuxCNC HAL driver for the Mesa Electronics HostMot2 firmware.
See the config modparam section below for Mesa card configuration. Typically hostmot2 is loaded with no parameters unless debugging is required.
loadrt hostmot2 [debug_idrom=N] [debug_module_descriptors=N] [debug_pin_descriptors=N] [debug_modules=N]
debug_idrom [default: 0]
Developer/debug use only! Enable debug logging of the HostMot2 IDROM header.
debug_module_descriptors [default: 0]
Developer/debug use only! Enable debug logging of the HostMot2 Module Descriptors.
debug_pin_descriptors [default: 0]
Developer/debug use only! Enable debug logging of the HostMot2 Pin Descriptors.
debug_modules [default: 0]
Developer/debug use only! Enable debug logging of the HostMot2 Modules used.
use_serial_numbers [default: 0]
When creating HAL pins for smart-serial devices name the pins by the board serial number rather than which board and port they are connected to. With this option set to 1 pins will have names like hm2_8i20.1234.current rather than hm2_5i23.0.8i20.0.1.current. The identifier consists of the last 4 digits of the board serial number, which is normally on a sticker on the board. This will make configs less portable, but does mean that boards can be re-connected less carefully.
hostmot2 is a device driver that interfaces the Mesa HostMot2 firmware to the LinuxCNC HAL. This driver by itself does nothing, the boards that actually run the firmware require their own drivers before anything can happen. Currently drivers are available for the 5i20, 5i22, 5i23, 5i25, 3x20, 4i65, and 4i68 (all using the hm2_pci module) and the 7i43 (using the hm2_7i43 module).
The HostMot2 firmware provides modules such as encoders, PWM generators, step/dir generators, and general purpose I/O pins (GPIOs). These things are called "Modules". The firmware is configured, at firmware compile time, to provide zero or more instances of each of these Modules.
The HostMot2 firmware runs on an FPGA board. The board interfaces with the computer via PCI, PC-104/Plus, or EPP, and interfaces with motion control hardware such as servos and stepper motors via I/O pins on the board.
Each I/O pin can be configured, at board-driver load time, to serve one of two purposes: either as a particular I/O pin of a particular Module instance (encoder, pwmgen, stepgen etc), or as a general purpose digital I/O pin. By default all Module instances are enabled, and all the board’s pins are used by the Module instances.
The user can disable Module instances at board-driver load time, by specifying a hostmot2 config string modparam. Any pins which belong to Module instances that have been disabled automatically become GPIOs.
All IO pins have some HAL presence, whether they belong to an active module instance or are full GPIOs. GPIOs can be changed (at run-time) between inputs, normal outputs, and open drains, and have a flexible HAL interface. IO pins that belong to active Module instances are constrained by the requirements of the owning Module, and have a more limited interface in HAL. This is described in the General Purpose I/O section below.
All the board-driver modules (hm2_pci and hm2_7i43) accept a load-time modparam of type string array, named "config". This array has one config string for each board the driver should use. Each board’s config string is passed to and parsed by the hostmot2 driver when the board-driver registers the board.
The config string can contain spaces, so it is usually a good idea to wrap the whole thing in double-quotes (the " character).
The comma character (,) separates members of the config array from each other.
For example, if your control computer has one 5i20 and one 5i23 you might load the hm2_pci driver with a HAL command (in halcmd) something like this:
loadrt hm2_pci config="firmware=hm2/5i20/SVST8_4.BIT num_encoders=3 num_pwmgens=3 num_stepgens=3,firmware=hm2/5i23/SVSS8_8.BIT sserial_port_0=0000 num_encoders=4"
Note: this assumes that the hm2_pci driver detects the 5i20 first and the 5i23 second. If the detection order does not match the order of the config strings, the hostmot2 driver will refuse to load the firmware and the board-driver (hm2_pci or hm2_7i43) will fail to load. To the best of my knowledge, there is no way to predict the order in which PCI boards will be detected by the driver, but the detection order will be consistent as long as PCI boards are not moved around. Best to try loading it and see what the detection order is.
The valid entries in the format string are:
Load the firmware specified by F into the FPGA on this board. If no "firmware=F" string is specified, the FPGA will not be re-programmed but may continue to run a previously downloaded firmware.
The requested firmware F is fetched by udev. udev searches for the firmware in the system’s firmware search path, usually /lib/firmware. F typically has the form "hm2/<BoardType>/file.bit"; a typical value for F might be "hm2/5i20/SVST8_4.BIT". The hostmot2 firmware files are supplied by the hostmot2-firmware packages, available from linuxcnc.org and can normally be installed by entering the command "sudo apt-get install hostmot2-firmware-5i23" to install the support files for the 5i23 for example.
The 5i25 / 6i25 come pre-programmed with firmware and no "firmware=" string should be used with these cards. To change the firmware on a 5i25 or 6i25 the "mesaflash" utility should be used (available from Mesa). It is perfectly valid and reasonable to load these cards with no config string at all.
num_dplls [optional, default: -1]
The hm2dpll is a phase-locked loop timer module which may be used to trigger certain types of encoder. This parameter can be used to disable the hm2dpll by setting the number to 0. There is only ever one module of this type, with 4 timer channels, so the other valid numbers are -1 (enable all) and 1, both of which end up meaning the same thing.
num_encoders [optional, default: -1]
Only enable the first N encoders. If N is -1, all encoders are enabled. If N is 0, no encoders are enabled. If N is greater than the number of encoders available in the firmware, the board will fail to register.
ssi_chan_N [optional, default: ""]
Specifies how the bit stream from a Synchronous Serial Interface device will be interpreted. There should be an entry for each device connected. Only channels with a format specifier will be enabled. (as the software can not guess data rates and bit lengths)
biss_chan_N [optional, default: ""]
As for ssi_chan_N, but for BiSS devices
fanuc_chan_N [optional, default: ""]
Specifies how the bit stream from a Fanuc absolute encoder will be interpreted. There should be an entry for each device connected. Only channels with a format specifier will be enabled. (as the software can not guess data rates and bit lengths)
num_resolvers [optional, default: -1]
Only enable the first N resolvers. If N = -1 then all resolvers are enabled. This module does not work with generic resolvers (unlike the encoder module which works with any encoder). At the time of writing the Hostmot2 Resolver function only works with the Mesa 7i49 card.
num_pwmgens [optional, default: -1]
Only enable the first N pwmgens. If N is -1, all pwmgens are enabled. If N is 0, no pwmgens are enabled. If N is greater than the number of pwmgens available in the firmware, the board will fail to register.
num_3pwmgens [optional, default: -1]
Only enable the first N Three-phase pwmgens. If N is -1, all 3pwmgens are enabled. If N is 0, no pwmgens are enabled. If N is greater than the number of pwmgens available in the firmware, the board will fail to register.
num_stepgens [optional, default: -1]
Only enable the first N stepgens. If N is -1, all stepgens are enabled. If N is 0, no stepgens are enabled. If N is greater than the number of stepgens available in the firmware, the board will fail to register.
stepgen_width [optional, default: 2]
Used to mask extra, unwanted, stepgen pins. Stepper drives typically require only two pins (step and dir) but the Hostmot2 stepgen can drive up to 8 output pins for specialised applications (depending on firmware). This parameter applies to all stepgen instances. Unused, masked pins will be available as GPIO.
sserial_port_N (N = 0 ..
3) [optional, default: 00000000 for all
Up to 32 Smart Serial devices
can be connected to a Mesa Anything IO board depending on
the firmware used and the number of physical connections on
the board. These are arranged in 1-4 ports of 1 to 8
Some Smart Serial (SSLBP) cards offer more than one load-time configuration, for example all inputs, or all outputs, or offering additional analogue input on some digital pins.
To set the modes for port 0 use, for example sserial_port_0=0120xxxx
A ’0’in the string sets the corresponding port to mode 0, 1 to mode 1, and so on up to mode 9. An "x" in any position disables that channel and makes the corresponding FPGA pins available as GPIO.
The string can be up to 8 characters long, and if it defines more modes than there are channels on the port then the extras are ignored. Channel numbering is left to right so the example above would set sserial device 0.0 to mode 0, 0.2 to mode2 and disable channels 0.4 onwards.
The sserial driver will auto-detect connected devices, no further configuration should be needed. Unconnected channels will default to GPIO, but the pin values will vary semi-randomly during boot when card-detection runs, to it is best to actively disable any channel that is to be used for GPIO.
num_bspis [optional, default: -1]
Only enable the first N Buffered SPI drivers. If N is -1 then all the drivers are enabled. Each BSPI driver can address 16 devices.
num_leds [optional, default: -1]
Only enable the first N of the LEDs on the FPGA board. If N is -1, then HAL pins for all the LEDs will be created. If N=0 then no pins will be added.
If specified, this turns on a raw access mode, whereby a user can peek and poke the firmware from HAL. See Raw Mode below.
The hm2dpll module has pins like "hm2_<BoardType>.<BoardNum>.dpll" It is likely that the pin-count will decrease in the future and that some pins will become parameters. This module is a phase-locked loop that will synchronise itself with the thread in which the hostmot2 "read" function is installed and will trigger other functions that are allocated to it at a specified time before or after the "read" function runs. This can be applied to the three absolute encoder types, quadrature encoders and stepgen. In the case of the absolute encoders this allows the system to trigger a data transmission just prior to the time when the HAL driver reads the data. In the case of stepgens and quadrature encoders the timers can be used to reduce position sampling jitter. This is especially valuable with the ethernet-interfaced cards.
(float, in) hm2_<BoardType>.<BoardNum>.dpll.NN.timer-us
This pin sets the triggering offset of the associated timer. There are 4 timers numbered 01 to 04, represented by the NN digits in the pin name. The units are micro-seconds. Generally the value will be negative, so that some action is undertaken by the fpga prior to the execution of the main hostmot2 read.
For stepgen and quadrature encoders, the value needs to be more than the maximum variation between read times. -100 will suffice for most systems, and -50 will work on systems with good performance and latency.
For serial encoders, the value also needs to include the time it takes to transfer the absolute encoder position. For instance, if 50 bits must be read at 500kHz then subtract an additional 50/500kHz = 100uS to get a starting value of -200.
(float, in) hm2_<BoardType>.<BoardNum>.dpll.base-freq-khz
This pin sets the base frequency of the phase-locked loop. by default it will be set to the nominal frequency of the thread in which the PLL is running and wil not normally need to be changed.
(float, out) hm2_<BoardType>.<BoardNum>.dpll.phase-error-us
Indicates the phase eror of the DPLL. If the number cycles by a large amount it is likely that the PLL has failed to achieve lock and adjustments will need to be made.
(u32, in) hm2_<BoardType>.<BoardNum>.dpll.time-const
The filter time-constant for the PLL. The default value is a compromise between insensitivity to single-cycle variations and being resilient to changes to the Linux CLOCK_MONOTONIC timescale, which can instantly change by up to Â±500ppm from its nominal value, usually by timekeeping software like ntpd and ntpdate. Default 2000 (0x7d0)
(u32, in) hm2_<BoardType>.<BoardNum>.dpll.plimit
Sets the phase adjustment limit of the PLL. If the value is zero then the PLL will free-run at the base frequency independent of the servo thread rate. This is probably not what you want. Default 4194304 (0x400000) Units not known...
(u32, out) hm2_<BoardType>.<BoardNum>.dpll.ddsize
Used internally by the driver, likely to disappear.
(u32, in) hm2_<BoardType>.<BoardNum>.dpll.prescale
Prescale factor for the rate generator. Default 1.
Encoders have names like ""hm2_<BoardType>.<BoardNum>.encoder.<Instance>".". "Instance" is a two-digit number that corresponds to the HostMot2 encoder instance number. There are "num_encoders" instances, starting with 00.
So, for example, the HAL pin that has the current position of the second encoder of the first 5i20 board is: hm2_5i20.0.encoder.01.position (this assumes that the firmware in that board is configured so that this HAL object is available)
Each encoder uses three or four input IO pins, depending on how the firmware was compiled. Three-pin encoders use A, B, and Index (sometimes also known as Z). Four-pin encoders use A, B, Index, and Index-mask.
The hm2 encoder representation is similar to the one described by the Canonical Device Interface (in the HAL General Reference document), and to the software encoder component. Each encoder instance has the following pins and parameters:
(s32 out) count
Number of encoder counts since the previous reset.
(float out) position
Encoder position in position units (count / scale).
(float out) velocity
Estimated encoder velocity in position units per second.
(bit in) reset
When this pin is TRUE, the count and position pins are set to 0. (The value of the velocity pin is not affected by this.) The driver does not reset this pin to FALSE after resetting the count to 0, that is the user’s job.
(bit in/out) index-enable
When this pin is set to True, the count (and therefore also position) are reset to zero on the next Index (Phase-Z) pulse. At the same time, index-enable is reset to zero to indicate that the pulse has occurred.
(s32 out) rawcounts
Total number of encoder counts since the start, not adjusted for index or reset.
(float r/w) scale
Converts from ’count’ units to ’position’ units.
(bit r/w) index-invert
If set to True, the rising edge of the Index input pin triggers the Index event (if index-enable is True). If set to False, the falling edge triggers.
(bit r/w) index-mask
If set to True, the Index input pin only has an effect if the Index-Mask input pin is True (or False, depending on the index-mask-invert pin below).
(bit r/w) index-mask-invert
If set to True, Index-Mask must be False for Index to have an effect. If set to False, the Index-Mask pin must be True.
(bit r/w) counter-mode
Set to False (the default) for Quadrature. Set to True for Step/Dir (in which case Step is on the A pin and Dir is on the B pin).
(bit r/w) filter
If set to True (the default), the quadrature counter needs 15 clocks to register a change on any of the three input lines (any pulse shorter than this is rejected as noise). If set to False, the quadrature counter needs only 3 clocks to register a change. The encoder sample clock runs at 33 MHz on the PCI AnyIO cards and 50 MHz on the 7i43.
(float r/w) vel-timeout
When the encoder is moving slower than one pulse for each time that the driver reads the count from the FPGA (in the hm2_read() function), the velocity is harder to estimate. The driver can wait several iterations for the next pulse to arrive, all the while reporting the upper bound of the encoder velocity, which can be accurately guessed. This parameter specifies how long to wait for the next pulse, before reporting the encoder stopped. This parameter is in seconds.
(s32 r/w) hm2_XiXX.N.encoder.MM.timer-num (default: -1)
Sets the hm2dpll timer instance to be used to latch encoder counts. A setting of -1 does not latch encoder counts. A setting of 0 latches at the same time as the main hostmot2 write. A setting of 1..4 uses a time offset from the main hostmot2 write according to the dpll’s timer-us setting.
Typically, timer-us should be a negative number with a magnitude larger than the largest latency (e.g., -100 for a system with mediocre latency, -50 for a system with good latency).
If no DPLL module is present in the FPGA firmware, or if the encoder module does not support DPLL, then this pin is not created.
When available, this feature should typically be enabled. Doing so generally reduces following errors.
(Not to be confused with the Smart Serial Interface)
One pin is created for each SSI instance regardless of data format: (bit, in) hm2_XiXX.NN.ssi.MM.data-incomplete This pin will be set "true" if the module was still transferring data when the value was read. When this problem exists there will also be a limited number of error messages printed to the UI. This pin should be used to monitor whether the problem has been addressed by config changes. Solutions to the problem dpend on whether the encoder read is being triggered by the hm2dpll phase-locked-loop timer (described above) or by the trigger-encoders function (described below).
The names of
the pins created by the SSI module will depend entirely on
the format string for each channel specified in the loadrt
command line. A typical format string might be
This would interpret the LSB of the bit-stream as a bit-type pin named "error" and the next 24 bits as a Gray-coded encoder counter. The encoder-related HAL pins would all begin with "position".
There should be no spaces in the format string, as this is used as a delimiter by the low-level code.
The format consists of a string of alphanumeric characters that will form the HAL pin names, followed by a % symbol, a bit-count and a data type. All bits in the packet must be defined, even if they are not used. There is a limit of 64 bits in total.
format characters and the pins they create are:
p: (Pad). Does not create any pins, used to ignore sections of
the bit stream that are not required.
hm2_XiXX.N.ssi.MM.<name>. If any bits in the
designated field width are non-zero then the HAL pin will be
(bit, out) hm2_XiXX.N.ssi.MM.<name>-not. An inverted version of the above, the HAL pin will be "true" if all bits in the field are zero.
(float, out) hm2_XiXX.N.ssi.MM.<name>. The value of the bits interpeted as an unsigned integer then scaled such that the pin value will equal the scalemax parameter value when all bits are high. (for example if the field is 8 bits wide and the scalmax parameter was 20 then a value of 255 would return 20, and 0 would return 0.
(float, out) hm2_XiXX.N.ssi.MM.<name>. The value of the bits interpreted as a 2s complement signed number then scaled similarly to the unsigned variant, except symmetrical around zero.
hm2_XiXX.N.ssi.MM.<name>-NN. The value of each
individual bit in the data field. NN starts at 00 up to the
number of bits in the field.
(bit, out) hm2_XiXX.N.ssi.MM.<name>-NN-not. An inverted version of the individual bit values.
hm2_XiXX.N.ssi.MM.<name>.count. The lower 32 bits of
the total encoder counts. This value is reset both by the
...reset and the ...index- enable pins.
(s32, out) hm2_XiXX.N.ssi.MM.<name>.rawcounts. The lower 32 bits of the total encoder counts. The pin is not affected by reset and index.
(float, out) hm2_XiXX.N.ssi.MM.<name>.position. The encoder position in machine units. This is calculated from the full 64-bit buffers so will show a true value even after the counts pins have wrapped. It is zeroed by reset and index enable.
(bit, IO) hm2_XiXX.N.ssi.MM.<name>.index-enable. When this pin is set "true" the module will wait until the raw encoder counts next passes through an integer multiple of the number of counts specified by counts-per-rev parameter and then it will zero the counts and position pins, and set the index-enable pin back to "false" as a signal to the system that "index" has been passed. this pin is used for spindle-synchronised motion and index-homing.
(bit, in) (bit, out) hm2_XiXX.N.ssi.MM.<name>.reset. When this pin is set high the counts and position pins are zeroed.
h: (Split encoder, high-order bits)
Some encoders (Including Fanuc) place the encoder part-turn counts and full-turn counts in separate, non-contiguous fields. This tag defines the high-order bits of such an encoder module. There can be only one h and one l tag per channel, the behaviour with multiple such channels will be undefined.
l: (Split encoder, low-order bits)
Low order bits (see "h")
g: (Gray-code). This is a
modifier that indicates that the
format string is gray-code encoded. This is only valid for encoders (e, h l) and unsigned (u) data types.
Two parameters is universally created for all SSI instances
(float r/w) hm2_XiXX.N.ssi.MM.frequency-khz
This parameter sets the SSI clock frequency. The units are kHz, so 500 will give a clock frequency of 500,000 Hz.
(s32 r/w) hm2_XiXX.N.ssi.MM.timer-num
This parameter allocates the SSI module to a specific hm2dpll timer instance. This pin is only of use in firmwares which contain a hm2dpll function and will default to 1 in cases where there is such a function, and 0 if there is not. The pin can be used to disable reads of the encoder, by setting to a nonexistent timer number, or to 0.
Other parameters depend on the data types specified in the config string.
p: (Pad) No Parameters.
b: (Boolean) No Parameters.
hm2_XiXX.N.ssi.MM.<name>.scalemax. The scaling factor
(float, r/w) hm2_XiXX.N.ssi.MM.<name>.scalemax. The scaling factor for the channel.
f: (bitField): No parameters.
hm2_XiXX.N.ssi.MM.<name>.scale: (float, r.w) The
encoder scale in counts per machine unit.
(u32, r/w) hm2_XiXX.N.ssi.MM.<name>.counts-per-rev (u32, r/w) Used to emulate the index behaviour of an incemental+index encoder. This would normally be set to the actual counts per rev of the encoder, but can be any whole number of revs. Integer divisors or multimpilers of the true PPR might be useful for index-homing. Non-integer factors might be appropriate where there is a synchronous drive ratio between the encoder and the spindle or ballscrew.
BiSS is a bidirectional variant of SSI. Currently only a single direction is supported by LinuxCNC (encoder to PC).
One pin is created for each BiSS instance regardless of data format:
(bit, in) hm2_XiXX.NN.biss.MM.data-incomplete This pin will be set "true" if the module was still transferring data when the value was read. When this problem exists there will also be a limited number of error messages printed to the UI. This pin should be used to monitor whether the problem has been addressed by config changes. Solutions to the problem dpend on whether the encoder read is being triggered by the hm2dpll phase-locked-loop timer (described above) or by the trigger-encoders function (described below)
The names of the pins created by the BiSS module will depend entirely on the format string for each channel specified in the loadrt command line and follow closely the format defined above for SSI. Currently data packets of up to 96 bits are supported by the LinuxCNC driver, although the Mesa Hostmot2 module can handle 512 bit packets. It should be possible to extend the number of packets supported by the driver if there is a requirement to do so.
The pins and
format specifier for this module are identical to the SSI
module described above, except that at least one
pre-configured format is provided. A modparam of
fanuc_chan_N=AA64 (case sensitive) will configure the
channel for a Fanuc Aa64 encoder. The pins created are:
hm2_XiXX.N.fanuc.MM.batt indicates battery state
hm2_XiXX.N.fanuc.MM.batt-not inverted version of above
hm2_XiXX.N.fanuc.MM.comm The 0-1023 absolute output for motor commutation
hm2_XXiX.N.fanuc.MM.crc The CRC checksum. Currently HAL has no way to use this
hm2_XiXX.N.fanuc.MM.encoder.count Encoder counts
hm2_XiXX.N.fanuc.MM.encoder.index-enable Simulated index. Set by counts-per-rev parameter
hm2_XiXX.N.fanuc.MM.encoder.position Counts scaled by the ...scale paramter
hm2_XiXX.N.fanuc.MM.encoder.rawcounts Raw counts, unaffected by reset or index
hm2_XiXX.N.fanuc.MM.encoder.reset If high/true then counts and position = 0
hm2_XiXX.N.fanuc.MM.valid Indicates that the absolute position is valid
hm2_XiXX.N.fanuc.MM.valid-not Inverted version
Resolvers have names like hm2_<BoardType>.<BoardNum>.resolver.<Instance>. <Instance is a 2-digit number, which for the 7i49 board will be between 00 and 05. This function only works with the Mesa Resolver interface boards (of which the 7i49 is the only example at the time of writing). This board uses an SPI interface to the FPGA card, and will only work with the correct firmware. The pins allocated will be listed in the dmesg output, but are unlikely to be usefully probed with HAL tools.
(float, out) angle
This pin indicates the angular position of the resolver. It is a number between 0 and 1 for each electrical rotation.
(float, out) position
Calculated from the number of complete and partial revolutions since startup, reset, or index-reset multiplied by the scale parameter.
(float, out) velocity
Calculated from the rotational velocity and the velocity-scale parameter. The default scale is electrical rotations per second.
(s32, out) count
This pins outputs a simulated encoder count at 2^24 counts per rev (16777216 counts).
(s32, out) rawcounts
This is identical to the counts pin, except it is not reset by the ’index’ or ’reset’ pins. This is the pin which would be linked to the bldc HAL component if the resolver was being used to commutate a motor.
(bit, in) reset
Resets the position and counts pins to zero immediately.
(bit, in/out) index-enable
When this pin is set high the position and counts pins will be reset the next time the resolver passes through the zero position. At the same time the pin is driven low to indicate to connected modules that the index has been seen, and that the counters have been reset.
(bit, out) error
Indicates an error in the particular channel. If this value is "true" then the reported position and velocity are invalid.
(float, read/write) scale
The position scale, in machine units per resolver electrical revolution.
(float, read/write) velocity-scale
The conversion factor between resolver rotation speed and machine velocity. A value of 1 will typically give motor speed in rps, a value of 0.01666667 will give (approximate) RPM.
(u32, read/write) index-divisor (default 1)
The resolver component emulates an index at a fixed point in the sin/cos cycle. Some resolvers have multiple cycles per rev (often related to the number of pole-pairs on the attached motor). LinuxCNC requires an index once per revolution for proper threading etc. This parameter should be set to the number of cycles per rev of the resolver. CAUTION: Which pseudo-index is used will not necessarily be consistent between LinuxCNC runs. Do not expect to re-start a thread after restarting LinuxCNC. It is not appropriate to use this parameter for index-homing of axis drives.
(float, read/write) excitation-khz
This pin sets the excitation
frequency for the resolver. This pin is module-level rather
than instance-level as all resolvers share the same
Valid values are 10 (~10kHz), 5 (~5kHz) and 2.5 (~2.5kHz). The actual frequency depends on the FPGA frequency, and they correspond to CLOCK_LOW/5000, CLOCK_LOW/10000 and CLOCK_LOW/20000 respectively. The parameter will be set to the closest available of the three frequencies.
A value of -1 (the default) indicates that the current setting should be retained.
pwmgens have names like "hm2_<BoardType>.<BoardNum>.pwmgen.<Instance>". "Instance" is a two-digit number that corresponds to the HostMot2 pwmgen instance number. There are ’num_pwmgens’ instances, starting with 00.
So, for example, the HAL pin that enables output from the fourth pwmgen of the first 7i43 board is: hm2_7i43.0.pwmgen.03.enable (this assumes that the firmware in that board is configured so that this HAL object is available)
In HM2, each pwmgen uses three output IO pins: Not-Enable, Out0, and Out1.
The function of the Out0 and Out1 IO pins varies with output-type parameter (see below).
The hm2 pwmgen representation is similar to the software pwmgen component. Each pwmgen instance has the following pins and parameters:
(bit input) enable
If true, the pwmgen will set its Not-Enable pin false and output its pulses. If ’enable’ is false, pwmgen will set its Not-Enable pin true and not output any signals.
(float input) value
The current pwmgen command value, in arbitrary units.
(float rw) scale
Scaling factor to convert ’value’ from arbitrary units to duty cycle: dc = value / scale. Duty cycle has an effective range of -1.0 to +1.0 inclusive, anything outside that range gets clipped. The default scale is 1.0.
(s32 rw) output-type
This emulates the output_type load-time argument to the software pwmgen component. This parameter may be changed at runtime, but most of the time you probably want to set it at startup and then leave it alone. Accepted values are 1 (PWM on Out0 and Direction on Out1), 2 (Up on Out0 and Down on Out1), 3 (PDM mode, PDM on Out0 and Dir on Out1), and 4 (Direction on Out0 and PWM on Out1, "for locked antiphase").
In addition to the per-instance HAL Parameters listed above, there are a couple of HAL Parameters that affect all the pwmgen instances:
(u32 rw) pwm_frequency
This specifies the PWM frequency, in Hz, of all the pwmgen instances running in the PWM modes (modes 1 and 2). This is the frequency of the variable-duty-cycle wave. Its effective range is from 1 Hz up to 193 kHz. Note that the max frequency is determined by the ClockHigh frequency of the Anything IO board; the 5i20 and 7i43 both have a 100 MHz clock, resulting in a 193 kHz max PWM frequency. Other boards may have different clocks, resulting in different max PWM frequencies. If the user attempts to set the frequency too high, it will be clipped to the max supported frequency of the board. Frequencies below about 5 Hz are not terribly accurate, but above 5 Hz they’re pretty close. The default pwm_frequency is 20,000 Hz (20 kHz).
(u32 rw) pdm_frequency
This specifies the PDM frequency, in Hz, of all the pwmgen instances running in PDM mode (mode 3). This is the "pulse slot frequency"; the frequency at which the pdm generator in the AnyIO board chooses whether to emit a pulse or a space. Each pulse (and space) in the PDM pulse train has a duration of 1/pdm_frequency seconds. For example, setting the pdm_frequency to 2e6 (2 MHz) and the duty cycle to 50% results in a 1 MHz square wave, identical to a 1 MHz PWM signal with 50% duty cycle. The effective range of this parameter is from about 1525 Hz up to just under 100 MHz. Note that the max frequency is determined by the ClockHigh frequency of the Anything IO board; the 5i20 and 7i43 both have a 100 MHz clock, resulting in a 100 Mhz max PDM frequency. Other boards may have different clocks, resulting in different max PDM frequencies. If the user attempts to set the frequency too high, it will be clipped to the max supported frequency of the board. The default pdm_frequency is 20,000 Hz (20 kHz).
Three-Phase PWM generators (3pwmgens) are intended for controlling the high-side and low-side gates in a 3-phase motor driver. The function is included to support the Mesa motor controller daughter-cards but can be used to control an IGBT or similar driver directly. 3pwmgens have names like "hm2_<BoardType>.<BoardNum>.3pwmgen.<Instance>" where <Instance> is a 2-digit number. There will be num_3pwmgens instances, starting at 00. Each instance allocates 7 output and one input pins on the Mesa card connectors. Outputs are: PWM A, PWM B, PWM C, /PWM A, /PWM B, /PWM C, Enable. The first three pins are the high side drivers, the second three are their complementary low-side drivers. The enable bit is intended to control the servo amplifier. The input bit is a fault bit, typically wired to over-current detection. When set the PWM generator is disabled. The three phase duty-cycles are individually controllable from -Scale to +Scale. Note that 0 corresponds to a 50% duty cycle and this is the inialization value.
A-value, B-value, C-value: The PWM command value for each
phase, limited to +/- "scale". Defaults to zero
which is 50% duty cycle on high-side and low-sidepins (but
see the "deadtime" parameter)
(bit input) enable
When high the PWM is enabled as long as the fault bit is not set by the external fault input pin. When low the PWM is disabled, with both high- side and low-side drivers low. This is not the same as 0 output (50% duty cycle on both sets of pins) or negative full scale (where the low side drivers are "on" 100% of the time)
(bit output) fault
Indicates the status of the fault bit. This output latches high once set by the physical fault pin until the "enable" pin is set to high.
(u32 rw) deadtime
Sets the dead-time between the high-side driver turning off and the low-side driver turning on and vice-versa. Deadtime is subtracted from on time and added to off time symmetrically. For example with 20 kHz PWM (50 uSec period), 50% duty cycle and zero dead time, the PWM and NPWM outputs would be square waves (NPWM being inverted from PWM) with high times of 25 uS. With the same settings but 1 uS of deadtime, the PWM and NPWM outputs would both have high times of 23 uS (25 - (2X 1 uS), 1 uS per edge). The value is specified in nS and defaults to a rather conservative 5000nS. Setting this parameter to too low a value could be both expensive and dangerous as if both gates are open at the same time there is effectively a short circuit accross the supply.
(float rw) scale
Sets the half-scale of the specified 3-phase PWM generator. PWM values from -scale to +scale are valid. Default is +/- 1.0
(bit rw) fault-invert
Sets the polarity of the fault input pin. A value of 1 means that a fault is triggered with the pin high, and 0 means that a fault it triggered when the pin is pulled low. Default 0, fault = low so that the PWM works with the fault pin unconnected.
(u32 rw) sample-time
Sets the time during the cycle when an ADC pulse is generated. 0 = start of PWM cycle and 1 = end. Not currently useful to LinuxCNC. Default 0.5.
In addition the per-instance parameters above there is the following parameter that affects all instances
(u32 rw) frequency
Sets the master PWM frequency. Maximum is approx 48kHz, minimum is 1kHz. Defaults to 20kHz.
stepgens have names like "hm2_<BoardType>.<BoardNum>.stepgen.<Instance>". "Instance" is a two-digit number that corresponds to the HostMot2 stepgen instance number. There are ’num_stepgens’ instances, starting with 00.
So, for example, the HAL pin that has the current position feedback from the first stepgen of the second 5i22 board is: hm2_5i22.1.stepgen.00.position-fb (this assumes that the firmware in that board is configured so that this HAL object is available)
Each stepgen uses between 2 and 6 IO pins. The signals on these pins depends on the step_type parameter (described below).
The stepgen representation is modeled on the stepgen software component. Each stepgen instance has the following pins and parameters:
(float input) position-cmd
Target position of stepper motion, in arbitrary position units. This pin is only used when the stepgen is in position control mode (control-type=0).
(float input) velocity-cmd
Target velocity of stepper motion, in arbitrary position units per second. This pin is only used when the stepgen is in velocity control mode (control-type=1).
(s32 output) counts
Feedback position in counts (number of steps).
(float output) position-fb
Feedback position in arbitrary position units. This is similar to "counts/position_scale", but has finer than step resolution.
(float output) velocity-fb
Feedback velocity in arbitrary position units per second.
(bit input) enable
This pin enables the step generator instance. When True, the stepgen instance works as expected. When False, no steps are generated and velocity-fb goes immediately to 0. If the stepgen is moving when enable goes false it stops immediately, without obeying the maxaccel limit.
(bit input) control-type
Switches between position control mode (0) and velocity control mode (1). Defaults to position control (0).
(float r/w) position-scale
Converts from counts to position units. position = counts / position_scale
(float r/w) maxvel
Maximum speed, in position units per second. If set to 0, the driver will always use the maximum possible velocity based on the current step timings and position-scale. The max velocity will change if the step timings or position-scale changes. Defaults to 0.
(float r/w) maxaccel
Maximum acceleration, in position units per second per second. Defaults to 1.0. If set to 0, the driver will not limit its acceleration at all - this requires that the position-cmd or velocity-cmd pin is driven in a way that does not exceed the machine’s capabilities. This is probably what you want if you’re going to be using the LinuxCNC trajectory planner to jog or run G-code.
(u32 r/w) steplen
Duration of the step signal, in nanoseconds.
(u32 r/w) stepspace
Minimum interval between step signals, in nanoseconds.
(u32 r/w) dirsetup
Minimum duration of stable Direction signal before a step begins, in nanoseconds.
(u32 r/w) dirhold
Minimum duration of stable Direction signal after a step ends, in nanoseconds.
(u32 r/w) step_type
Output format, like the step_type modparam to the software stegen(9) component. 0 = Step/Dir, 1 = Up/Down, 2 = Quadrature, 3+ = table-lookup mode. In this mode the step_type parameter determines how long the step sequence is. Additionally the stepgen_width parameter in the loadrt config string must be set to suit the number of pins per stepgen required. Any stepgen pins above this number will be available for GPIO. This mask defaults to 2. The maximum length is 16. Note that Table mode is not enabled in all firmwares but if you see GPIO pins between the stepgen instances in the dmesg/log hardware pin list then the option may be available.
In Quadrature mode (step_type=2), the stepgen outputs one complete Gray cycle (00 â 01 â 11 â 10 â 00) for each "step" it takes. In table mode up to 6 IO pins are individually controlled in an arbitrary sequence up to 16 phases long.
(u32 r/w) table-data-N
There are 4 table-data-N parameters, table-data-0 to table-data-3. These each contain 4 bytes corresponding to 4 stages in the step sequence. For example table-data-0 = 0x00000001 would set stepgen pin 0 (always called "Step" in the dmesg output) on the first phase of the step sequence, and table-data-4 = 0x20000000 would set stepgen pin 6 ("Table5Pin" in the dmesg output) on the 16th stage of the step sequence.
(s32 r/w) hm2_XiXX.N.stepgen.MM.timer-num (default: -1)
Sets the hm2dpll timer instance to be used to latch stepgen counts. A setting of -1 does not latch encoder counts. A setting of 0 latches at the same time as the main hostmot2 write. A setting of 1..4 uses a time offset from the main hostmot2 write according to the dpll’s timer-us setting.
Typically, timer-us should be a negative number with a magnitude larger than the largest latency (e.g., -100 for a system with mediocre latency, -50 for a system with good latency).
If no DPLL module is present in the FPGA firmware, or if the stepgen module does not support DPLL, then this pin is not created.
When available, this feature should typically be enabled. Doing so generally reduces following errors.
The Smart Serial Interface allows up to 32 different devices such as the Mesa 8i20 2.2kW 3-phase drive or 7i64 48-way IO cards to be connected to a single FPGA card. The driver auto-detects the connected hardware port, channel and device type. Devices can be connected in any order to any active channel of an active port. (see the config modparam definition above).
For full details of the smart-serial devices see man sserial.
The BSPI (Buffered SPI) driver is unusual in that it does not create any HAL pins. Instead the driver exports a set of functions that can be used by a sub -driver for the attached hardware. Typically these would be written in the "comp" pre-processing language: see http://linuxcnc.org/docs/html/hal_comp.html or man comp for further details. See man mesa_7i65 and the source of mesa_7i65.comp for details of a typical sub-driver. See man hm2_bspi_setup_chan, man hm2_bspi_write_chan, man hm2_tram_add_bspi_frame, man hm2_allocate_bspi_tram, man hm2_bspi_set_read_funtion and man hm2_bspi_set_write_function for the exported functions.
The names of the available channels are printed to standard output during the driver loading process and take the form hm2_<board name>.<board index>.bspi.<index> For example hm2_5i23.0.bspi.0
The UART driver also does not create any HAL pins, instead it declares two simple read/write functions and a setup function to be utilised by user-written code. Typically this would be written in the "comp" pre-processing language: see http://linuxcnc.org/docs/html/hal_comp.html or man comp for further details. See man mesa_uart and the source of mesa_uart.comp for details of a typical sub-driver. See man hm2_uart_setup_chan, man hm2_uart_send, man hm2_uart_read and man hm2_uart_setup.
The names of the available uart channels are printed to standard output during the driver loading process and take the form hm2_<board name>.<board index>uart.<index> For example hm2_5i23.0.uart.0
I/O pins on the board which are not used by a module instance are exported to HAL as "full" GPIO pins. Full GPIO pins can be configured at run-time to be inputs, outputs, or open drains, and have a HAL interface that exposes this flexibility. IO pins that are owned by an active module instance are constrained by the requirements of the owning module, and have a restricted HAL interface.
GPIOs have names like "hm2_<BoardType>.<BoardNum>.gpio.<IONum>". IONum is a three-digit number. The mapping from IONum to connector and pin-on-that-connector is written to the syslog when the driver loads, and it’s documented in Mesa’s manual for the Anything I/O boards.
So, for example, the HAL pin that has the current inverted input value read from GPIO 012 of the second 7i43 board is: hm2_7i43.1.gpio.012.in-not (this assumes that the firmware in that board is configured so that this HAL object is available)
The HAL parameter that controls whether the last GPIO of the first 5i22 is an input or an output is: hm2_5i22.0.gpio.095.is_output (this assumes that the firmware in that board is configured so that this HAL object is available)
The hm2 GPIO
representation is modeled after the Digital Inputs and
Digital Outputs described in the Canonical Device Interface
(part of the HAL General Reference document). Each GPIO can
have the following HAL Pins:
(bit out) in & in_not
State (normal and inverted) of the hardware input pin. Both full GPIO pins and IO pins used as inputs by active module instances have these pins.
(bit in) out
Value to be written (possibly inverted) to the hardware output pin. Only full GPIO pins have this pin.
Each GPIO can
have the following Parameters:
(bit r/w) is_output
If set to 0, the GPIO is an input. The IO pin is put in a high-impedance state (weakly pulled high), to be driven by other devices. The logic value on the IO pin is available in the "in" and "in_not" HAL pins. Writes to the "out" HAL pin have no effect. If this parameter is set to 1, the GPIO is an output; its behavior then depends on the "is_opendrain" parameter. Only full GPIO pins have this parameter.
(bit r/w) is_opendrain
This parameter only has an effect if the "is_output" parameter is true. If this parameter is false, the GPIO behaves as a normal output pin: the IO pin on the connector is driven to the value specified by the "out" HAL pin (possibly inverted), and the value of the "in" and "in_not" HAL pins is undefined. If this parameter is true, the GPIO behaves as an open-drain pin. Writing 0 to the "out" HAL pin drives the IO pin low, writing 1 to the "out" HAL pin puts the IO pin in a high-impedance state. In this high-impedance state the IO pin floats (weakly pulled high), and other devices can drive the value; the resulting value on the IO pin is available on the "in" and "in_not" pins. Only full GPIO pins and IO pins used as outputs by active module instances have this parameter.
(bit r/w) invert_output
This parameter only has an effect if the "is_output" parameter is true. If this parameter is true, the output value of the GPIO will be the inverse of the value on the "out" HAL pin. Only full GPIO pins and IO pins used as outputs by active module instances have this parameter.
Creates HAL pins for the LEDs on the FPGA board.
(bit in) CR<NN>
The pins are numbered from CR01 upwards with the name corresponding to the PCB silkscreen. Setting the bit to "true" or 1 lights the led.
The HostMot2 firmware may include a watchdog Module; if it does, the hostmot2 driver will use it. The HAL representation of the watchdog is named "hm2_<BoardType>.<BoardNum>.watchdog".
The watchdog starts out asleep and inactive. Once you access the board the first time by running the hm2 write() HAL function (see below), the watchdog wakes up. From them on it must be petted periodically or it will bite. Pet the watchdog by running the hm2 write() HAL function.
When the watchdog bites, all the board’s I/O pins are disconnected from their Module instances and become high-impedance inputs (pulled high), and all communication with the board stops. The state of the HostMot2 firwmare modules is not disturbed (except for the configuration of the IO Pins). Encoder instances keep counting quadrature pulses, and pwm- and step-generators keep generating signals (which are *not* relayed to the motors, because the IO Pins have become inputs).
Resetting the watchdog (by clearing the has_bit pin, see below) resumes communication and resets the I/O pins to the configuration chosen at load-time.
If the firmware includes a watchdog, the following HAL objects will be exported:
(bit in/out) has_bit
True if the watchdog has bit, False if the watchdog has not bit. If the watchdog has bit and the has_bit bit is True, the user can reset it to False to resume operation.
(u32 read/write) timeout_ns
Watchdog timeout, in nanoseconds. This is initialized to 5,000,000 (5 milliseconds) at module load time. If more than this amount of time passes between calls to the hm2 write() function, the watchdog will bite.
If the "enable_raw" config keyword is specified, some extra debugging pins are made available in HAL. The raw mode HAL pin names begin with "hm2_<BoardType>.<BoardNum>.raw".
With Raw mode enabled, a user may peek and poke the firmware from HAL, and may dump the internal state of the hostmot2 driver to the syslog.
(u32 in) read_address
The bottom 16 bits of this is used as the address to read from.
(u32 out) read_data
Each time the hm2_read() function is called, this pin is updated with the value at .read_address.
(u32 in) write_address
The bottom 16 bits of this is used as the address to write to.
(u32 in) write_data
This is the value to write to .write_address.
(bit in) write_strobe
Each time the hm2_write() function is called, this pin is examined. If it is True, then value in .write_data is written to the address in .write_address, and .write_strobe is set back to False.
(bit in/out) dump_state
This pin is normally False. If it gets set to True the hostmot2 driver will write its representation of the board’s internal state to the syslog, and set the pin back to False.
See man setsserial for the current way to set smart-serial eeprom parameters.
On boards with long turn around time for reads (at the time of writing, this applies only to ethernet boards), this function sends a read request. When multiple boards are used, this can reduce the servo thread execution time. In this case, the appropriate thread order would be
addf hm2_7i80.0.read-request addf hm2_7i80.1.read-request addf hm2_7i80.0.read addf hm2_7i80.1.read
which causes the read request to be sent to board 1 before waiting for the response to the read request to arrive from board 0.
This reads the encoder counters, stepgen feedbacks, and GPIO input pins from the FPGA.
This updates the PWM duty cycles, stepgen rates, and GPIO outputs on the FPGA. Any changes to configuration pins such as stepgen timing, GPIO inversions, etc, are also effected by this function.
Read the GPIO input pins. Note that the effect of this function is a subset of the effect of the .read() function described above. Normally only .read() is used. The only reason to call this function is if you want to do GPIO things in a faster-than-servo thread. (This function is not available on the 7i43 due to limitations of the EPP bus.)
Write the GPIO control registers and output pins. Note that the effect of this function is a subset of the effect of the .write() function described above. Normally only .write() is used. The only reason to call this function is if you want to do GPIO things in a faster-than-servo thread. (This function is not available on the 7i43 due to limitations of the EPP bus.)
This function will only appear if the firmware contains a BiSS, Fanuc or SSI encoder module and if the firmare does not contain a hm2dpll module (qv) or if the modparam contains num_dplls=0. This function should be inserted first in the thread so that the encoder data is ready when the main hm2_XiXX.NN.read function runs. An error message will be printed if the encoder read is not finished in time. It may be possible to avoid this by increasing the data rate. If the problem persists and if "stale" data is acceptable then the function may be placed later in the thread, allowing a full servo cycle for the data to be transferred from the devices. If available it is better to use the synchronous hm2dpll triggering function.
Mesa’s documentation for the Anything I/O boards, at <http://www.mesanet.com>