Release Note
MPI Library Version 04.04.20

Description:

Add support for future AKD1G NDSQ model type.

The sqNode library will automatically use the correct resource mapping when the NDSQ model types are detected. With prior MPI versions, the "default" resource assignment will be used (defaults may lack a few special features, see "What's New" on the SynqNet support site).

Description:

Add MPI support for two new S200 parameters FbType and FilterType used by S200 LogicVer 4 (Control Logic versions 4.0G and later).

These new parameter types are not required for customers using the standard 3.2M feature set. However these new parameters can be used to select features that were formerly custom S200 logic versions. See S200 documentation for details.

Problem:

This problem cause is in incorrect actual velocity calculator of the filter.

Cause:

In a Gantry configuration, the Velocity Input to the filter was derived from the main axis' velocity (velPtr). This resulted in an unexpected velocity loop response.

Fix/Solution:

The Gantry configuration combined the actual velocity of both axes to obtain the actual velocity. The customer can use the same configuration of the sample code for Gantry configuration. This problem was fixed in firmware version 922A3.

In the latest firmware, the actual velocity will made from filterConfig.Axis[n].Ptr Therefore customer cannot use velPtr input from another encoder. This is important for feedback design for Gantry axis.

Problem:

MechaWare Library DIVIDE block does not work in Simlink Simulation.

Cause:

The cause is a wrong mask parameter -- it should be '*/' but it was '//'.

Fix/Solution:

This problem is fixed in 04.04.19.

Note workaround for customers with older MechaWare: Customers can fix this issue by replacing '//' to '*/' in the DIVIDE block of the library file, mwblocks.mdl.

Problem:

When loading BIN files, MDL2MW failed to find the BIN file.

Cause:

Case-insensitive searches were not performed when searching for BIN files.

Fix/Solution:

Case-insensitive file search is performed correctly. MDL2MW finds BIN files correctly.

Problem:

VM3 does not display the following variables: Coeffset HoldCondition. EStop Modify point Triggered Motion Modify point These are all variables allocated in extended buffer area.

Cause:

MPI wasn't coded to display them.

Fix/Solution:

Code was added to display the variables.

Problem:

Triggered Motion modify does not work.

Cause:

There was a problem handling the status, MFWStatusTRIGGERED_MODIFY when processing limits.

Fix/Solution:

The status was properly handled. Sample code 'usrlim8.c' added to example folder as countermeasure to verify the behavior.

Problem:

When the number of objects to be saved in flash is manually increased, and an attempt to save the parameters to flash memory, the increased set of objects are not saved.

Cause:

In the flash code, MPI only saves the number of objects that was originally assigned.

Fix/Solution:

MPI was changed so that the extended number of objects is also saved to flash memory. MPI now saves flash objects correctly.

Problem:

Kollmorgen NIM-V3 did not move into the 'sync' state of SynqNet on MPI04.

Cause:

Board initialization for NIM-V3 was missing in MPI04.

Fix/Solution:

Board initialization for NIM-V3 was included in MPI04.

Problem:

The MechaWare Noise block will report the 'Invalid MDL model file' error if the block does not specify the input block.

Cause:

The Noise block has an input pin. This pin is a Simulink option, so no error should be reported if the input pin connection is not specified.

Fix/Solution:

Even if an input pin connection is not specified, no error is reported.

Problem:

The EventManager of service routine returned “device error”. This error is from the device driver.

Cause:

03.04.xx MPI does not support pending interrupts.

Fix/Solution:

Support for pending interrupts was added. (Currently it is as same as MPI 04 driver.)

Problem:

The MechaWare block PIV_GT has a bug with processing velocity deltas in transitions.

Cause:

If the status is DISABLE, the internal previous position data for velocity calculation will be zero. If the status is now enabled, the position delta (= velocity) will be the current feedback position, causing a position jump.

Fix/Solution:

This problem was fixed on 920A3 FW. There is no position jump.

Problem:

mpiCaptureReset() returns "0x0204 (516) Capture: invalid feedback number." The application changes capture-feedback configuration memory using mpiControlMemorySet() directly after the first mpiCaptureConfigSet(). The second mpiCaptureConfigSet() returns the referenced error.

Confirmed version 04.03.18 working with same application code.

Cause:

A prior change caused this error. Reference: SVN 76958 "Add mpiCaptureConfigGet(), before calling mpiCaptureConfigSet()" All 0 is good.

Fix/Solution:

Reverted SVN 76958 source code. mpiCaptureReset() now bahaves as specified.

Problem:

When a motion sequenceis repeated, the second sequence returns an error. This is the motion sequence that contains an error:

1st command to mpiMotionCamMove() returns MPIMessageOK with MPIMotionCamAttrMaskREPEAT. After MotionActionE_STOP and a Reset, 2nd command mpiMotionCamMove returns MPIMotionMessageATTRIBUTE_INVALID.

Cause:

Internal validation was using the wrong variable.

Fix/Solution:

Validation is now correctly using the right variable. Both motion sequence work correctly now.

Problem:

If mpiFlashCreate is called with an invalid control handle, MPI will access bad memory.

Cause:

It is not checking for an invalid control handle.

Fix/Solution:

Code was added to check for an invalid Control handle. It returns an error status if an invalid control handle is passed.

Problem:

If the application called mpiControlMemoryToFile() function with NULL filename pointer, MPI might report an assertion.

Cause:

This cause is an uninitialized internal variable.

Fix/Solution:

It returns an error status if a NULL filename pointer is passed as argument.

Problem:

MPI library can leak memory.

Cause:

Some MPI methods were not freeing the allocated objects.

Fix/Solution:

All MPI methods now properly free allocated objects.

Problem:

mpiUserLimitEnableSet will report a spurious timeout in rare situations in a multi-threaded MPI client. This problem is related to Bugzilla-10393.

Cause:

If the same limit is controlled by two threads and operations occur at about the same time, the firmware has a race condition which causes it to report a timeout.

Fix/Solution:

The MPI will now use two separate locks for enable and disable operations.

Problem:

mpiUserLimitEnableSet will report a spurious timeout error if customer calls mpiUserLimitConfigSet when enable=true, followed quickly by mpiUserLimitEnable (argument=false).

Cause:

The timeout check was faulty in this situation.

Fix/Solution:

This logic for the check for timeout was fixed.

Problem:

A timeout error is returned when downloading new Firmware to a QMP (Flash.exe). This issue is only exhibited on QMP boards that have the newer MT28EW256ABA flash chip. This issue is related only to flash programming only and not normal operation.

Cause:

The flash timeout was traced to a software bug introduced in MPI 04.03.18 and MPI 04.04.03 – 04.04.08.

Fix/Solution:

Upgrade to MPI 4.04.09 or later.

Problem:

The ZMP command position may drift for SqDC4 drives. This may occur if the SqDC4 drive has some load while holding position. The rate of position drift depends on the motor and load.

Cause:

The solution was a change in the MPI code so that the ZMP PID will ignore the AmpActive status from SqDC4. The MPIMotorDedicatedInAMP_ACTIVE bit for SqDC4 SqNode library is NOT set, which will cause the ZMP to ignore the AmpActive status.

Fix/Solution:

The position drift occurs because the SqDC4 toggles the AmpActive state on/off/on even though the ZMP continuously requests AmpEnable. When AmpActive toggles off, the ZMP's PID is momentarily disabled and the "command = actual" disable feature may allow ZMP command position to drift. Note this use of the AmpActive was introduced in MPI 2628 "AKD-SQ Reduce Position Jump on Amp Enable" (MPI 04.03.05).

With the updated MPI, the ZMP will instead use the AmpEnable state to control disable behavior of ZMP PID.

Problem:

In the previous specification, the controller init would called motor config set if the motor had a gear ratio under certain conditions. This is for re-calculation of the gear position because if contoller initialization was done first time after power on, and gear ratio has been flashed, the intial position might be wrong position if resolution is very high and it had ABS initial position. That original cause was overflow of firmware register.

Cause:

However, the firmware overflow issue was fixed in MP2728 already. Therefore motor re-initialization process was no longer useful. And this re-initial process might become an unexpected operatioin if customer configures the controller with unique motor link. (This case is posted by MPI2765)

Fix/Solution:

The motor re-configuration was removed from controller initial process. MPI and controller should be correct operation in any case.

Problem:

If customer configured unique axis-motor link (like an Axis0 referred Motor1 encoder as feedback), and linked motor has been configured with a gear ratio, Axis0 encoder will jump if controller is initialized.

This problem occured when Motion Console was refreshed because it initialized the controller.

Cause:

This cause is in motor re-configuration in controller initialization process. In expected operation, MPI will skip this motor configuraton if the motor enabled. However, this skip condition was incomplete and only the same axis and motor numbers were checked. Therefore, if customer configured unique link between axis and motor, the encoder position jumped.

Fix/Solution:

The cause of this route is reconfiguration, but this feature was no longer needed.
(See MPI2766). Because we removed re-configuration from controller initialization process, encoder jump issue is also fixed now.

Problem:

Motion will unexpected done event if axis list has been changed in another MS.

If customer is using multi thread or multi process application, and if these program are running on paralell that one task changed axis map configuration and other task called motion start and wait an event, the FW might notifed the unexpected done event.

Cause:

This cause is in axis map change process. So far, MPI could changed the axis map in anytime.
But FW process is refering mapped axis count for the loop for MS status control internaly. Therefore if MPI changed axis list to smaller than previous during MS processing of FW, FW refered to unexpected axis list as '0'. This was cause of unexpected event notified.

Fix/Solution:

For axis map control, we changed the MPI to call the MPIActionMAP internally. In new MPI, FW will set the axis count to ZERO by MPIActionMAP first. And if FW conpleted that process, MPI will set the new axis map and new axis count. This process allows you to synchronize between MPI and FW so you can fix unexpected events.

Problem:

The mpiMotionMemoryGet and mpiFilterMemoryGet/Set did not work correctly.

Cause:

These function's address validation was broken and it would return the MPIMessageARG_INVALID unpredictably.

Fix/Solution:

Now the functions sould be work collectly.

Problem:

In Windows 64 OS, customer could not open the drive map with sqDriveConfig tool.

Cause:

This cause is in MPI source code. So far, sqDriveConfig could be run on WIn32 only, not in Win64. This was the bug.

Fix/Solution:

Now, we can run this tool on Win64 too.

Problem:

If customer is using certain MPI methods as both Service and Application, controller will lose synchronization sometimes.

Cause:

The MPI lock control didn't support global lock control. Since, MPI lock did not support global control, If customer was using a Windows Service, the lock control between service and application layer was not functional.

Fix/Solution:

Now we can use MPI on both Service and Application at the same time.

Problem:

Because there is no LOCK and UNLOCK, if multiple threads call those functions at the same time, they collide and the functions can not return the correct value.

Cause:

LOCK and UNLOCK missing from the source code.

Fix/Solution:

Added LOCK and UNLOCK code to the mpiAxisCommandPositionGet(), the mpiAxisActualPositionGet() and other functions.

Problem:

The Compute command of sequencer could not work.

Cause:

A bug was found in the firmware. The compute command could not refer to the target address.

Fix/Solution:

This problem was fixed by a firmware code change. The compute command should work from 918A5.

Problem:

MPI had a wrong operation of the address in compare configuration.

Cause:

Coding error.

Fix/Solution:

The correct operation was implemented.

Problem:

To flush the last flash command, the last address was read back. This code had a bug and it was reading back a different address. Sometimes the address would be close enough to flush the command, and other times it would be completely different from the last address written.

So sometimes the erase was successful, other times it failed.

Cause:

The wrong address was used to read back the last command sent to the flash. i.e. the flash chip did not get the last command to erase it and presumably times out.

Fix/Solution:

The boot sector erases correctly.

Problem:

The downstream error count increases every few seconds on all SynqNet nodes. No other errors are visible, and the system continues to run normally.

Cause:

The root cause is in the ZMP controller FPGA version 324_A301.fpg and has been resolved in version 325_A301.fpg. This fix is recommended per customer request if the downstream error issue occurs on any MPI 04.xx.xx system. Note these downstream errors are caused by transmit timing errors and should be eliminated if they occur.

Fix/Solution:

FPGA version 325_xxxx.fpg has been modified to prevent this issue.

The version 325_xxxx.fpg FPGAs are recommended for all MPI 04.xx.xx versions, and all controller hardware versions.

Note: controller FPGAs are downloaded every time flash.exe updates firmware. The default FPGA version is loaded unless the command line option is used, for example:
cmd> flash.exe -fpga0 325_A301.fpg ZMP913B3.bin
An MPI change is required to modify the default FPGA version.

Problem:

Depending on the sequence program, QMP will emit an alignment violation exception error then stop.

Cause:

QMP accesses 32 bit data twice and makes 64 bit data now.

Fix/Solution:

QMP accessed 64 bit data regardless of 32 bit data and 64 bit data at condition wait and calculation.

Problem:

mpiMotionPositionGet() receives incorrect position data when mpiAxisOrginSet() is called.
using mpiAxisCommandPositionGet() or mpiAxisActualPositionGet() returns correct position data.

Cause:

After calling mpiAxisOriginSet(), the Axis object executes a sample count wait. In MPI 03, the Motion objects have related Axis objects, so no problem exists. However in MPI 04, the Motion objects does not have related Axis objects and poses a problem as mpiMotionPositionGet() does not know whether mpiAxisOriginSet() is called or not.

Fix/Solution:

After calling mpiAxisOriginSet (), the Control object also executes a sample wait. Additionally, the motion objects can call a sample wait function using the control object and axis number.

Problem:

MPI has a certain initialization function and the argument interface was changed in 04.04.02. However the S200 Tool is unaware of this change, the MPI returned an error by argument mismatch when the function was called.

Cause:

This cause is the MPI interface change in 04.04.02.

Fix/Solution:

The correct initialization function is now called.

Problem:

When data is exported as hex to file, the values are bad.

Cause:

Eg.: When controller data was 0x80000001, exported data was 0x80000000.

Fix/Solution:

Also: Hex format display wasn't implemented correctly.

Problem:

This feature checks and validates the size of firmware address for 64-bit MPI builds.

The size of the firmware address is 32-bit. But if a customer made a mistake and set the wrong software build option, the firmware address becomes a 64-bit value.

This feature is introduced to catch this error. However, it will be 64-bit when customer sets wrong software build option.

Cause:

NA

Fix/Solution:

Check firmware address size when application call the mpiControlCreate().
If it is bad, it will return "Control: Firmware address size mismatch.".

Call yjr mpiControlCreate()

Problem:

If the application uses the 'Capture State' address by MPIUserLimitConditionTypeCUSTOM and sets configuration using mpiUserLimitConfigSet, then, when calling mpiUserLimitConfigGet(), the type is returned as MPIUserLimitConditionTypeCAPTURE_STATE.

Cause:

mpiUserLimitConfigGet().

It is determined only by the address that set the condition type. Additionally, the custom setting is changed to the simple setting of each condition type.

Fix/Solution:

The condition type is stored in a firmware field and retrieved when mpiUserLimitConfigGet is called.

The types that are set are read properly in the get functions.

Problem:

For AKD & S200 drives, if the user generates an Abort action (Disable action), the 'Near Target' status will be TRUE.

Cause:

The motor status data was incorrectly getting reflected in the axis status and causing a spurious status.

Fix/Solution:

When motion is aborted, the wrong bit from motor status is not copied to the axis status.

Problem:

In a certain situations, the motor gearing function did not work well.Example is when using a high value for the resolution and spinning the motor shaft at high speed.

Cause:

This problem was in unit size of firmware calculation.

Fix/Solution:

The calculation was fixed and the motor gearing works as expected.

Problem:

Axes are assigned to the following:

1. MS - Axis map
MS#0: Axis#0
MS#1: Axis#1
MS#3: Axis#0 + Axis#1

2. Motion start
MS#0: start.

3. EStopAbort
MS#2: EStopAbort.

4. Problem
Axis#0 abort soon.

Cause:

This problem occurs when two or more axes are assigned to a motion supervisor (MS). The axis staus of the 2 or more axes were not combined properly.

Fix/Solution:

When multiple axes are being assigned to a motion supervisor, each axis staus is properly combined to set the motion supervisor status.

It works correctly to abort all the axes at the right time.

Problem:

When using MPIRecorderRecordTypeAXIS in mpiRecorderRecordConfig(), the
DAC output value was always 0.

Cause:

DAC output was assigned filter output and MechaWare firmware does not have filter.

Fix/Solution:

Check Mechaware or Standard firmware. If it is Mechaware firmware, DAC output is assigned for Motor commutation input.

Problem:

Map: symbol not found is returned when setting the map pointer as MPICapture.Data.State address and calling the mpiMapPtrToSymbol() function.

Cause:

Improper coding of the mpiMapPtrToSymbol() function.

Fix/Solution:

The mpiMapPtrToSymbol() function was corrected.

Description:

If installed in the default installation directory (C:\Program files), Motion Console and Motion Scope require administrator privileges to run. This is due to increased security measures by Microsoft in Windows 7 and above. Installing to an alternate directory will circumvent this requirement.

Description:

At the time of the release, the development of SimServer had not been completed. This feature is available as a standalone package and is not included in the MPI distributables.