Release Note
C0FE003A_xxxx.sff
MPI Version |
FPGA Version |
03.04.00
Production Release |
C0FE003A_0400 |
03.03.00
Production Release |
C0FE003A_0346 |
C0FE003A_0400
|
Remove drive reset complete from the fault and fault mask registers |
|
|
Reference Number: FP 466 |
|
|
Type: General Change |
|
|
FPGA Version: 0x0400 |
|
|
Problem/Cause:
As more drives require a driveBootComplete bit to indicate when they are finished booting, it was necessary to change the current bit to a more ideal location. |
|
|
Fix/Solution:
The FPGA bit, driveResetComplete, has been removed from the dedicatedIo Fault and Fault Mask registers. This bit has been replaced with driveBootComplete in the dedicatedIo Config register. This bit is monitored solely by the sqNodeLib utility. It is currently not visible through MPI calls or through Motion Console.
NOTE: This change causes compatibility issues between new FPGAs and old software (or old FPGAs with new software). FPGA version x0400 and later require the 03.04.00 MPI software release and later.
|
|
The watchdog fault was incorrectly being cleared after being activated |
|
|
Reference Number: FP 465 |
|
|
Type: Bug Fix |
|
|
FPGA Version: 0x0400 |
|
|
Problem/Cause:
The Slice I/O watchdog fault was incorrectly being cleared after being activated. If a watchdog fault occurred, the FPGA would send a stop command to all slices to halt cyclic operation. As a result of being in this stop mode, it would cause the watchdog fault to be disabled and incorrectly clear itself. Therefore, the fault would never be seen. |
|
|
Fix/Solution:
The FPGA was changed so that if the watchdog fault was the trigger to enter into the stop mode, the watchdog fault would not disable itself, thus allowing the fault to stay latched and also trigger the nodeIo fault.
|
|
sqMac Version 0300 |
|
|
Reference Number: FP 407 |
|
|
Type: New Feature |
|
|
FPGA Version: 0x0400 |
|
|
Description:
The sqNode x0400 FPGA release introduces an upgraded network module, sqMac version 0300. Upgraded features include:
- SynqNet HotReplace - allows one or more more nodes to be shutdown, replaced, and brought back on-line without interrupting other nodes.
- Revised Cable Length Discovery - a new node-to-node method was added to make discovery more accurate.
- sqPll shutdown - a controlled shutdown of individual nodes for HotReplace.
The sqMac 0300 version is NOT fully compatible with MPI versions prior to 03.04.00. Using 0300 sqMac with older software (for some configurations) might cause the "idle cable" check function to accidently corrupt normal packet traffic. But, the MPI will issue a warning if old/new FPGA/MPI are mixed by accident. |
|
Latch added on the notAmpPowered bit |
|
|
Reference Number: FP 402 |
|
|
Type: General Change |
|
|
FPGA Version: 0x0400 |
|
|
Problem/Cause:
In FPGA versions previous to x0400, the notAmpPowered fault bit was not latched. |
|
|
Fix/Solution:
In FPGA versions x0400 and later, the notAmpPowered fault bit is now latched. If active, it can be cleared by writing a "1" to bit 5 of the dedicatedIo fault status register.
|
C0FE003A_0346
|
Modications to gpioPinConfigReg |
|
|
Reference Number: FP 336 |
|
|
Type: General Change |
|
|
FPGA Version: 0x0346 |
|
|
Problem/Cause:
Prior to FPGA version x0346, gpio pins that were configured as an "output" did not resume "output" direction after an ioAbort event. Also, during the ioAbort event, if a service channel write to the gpio configuration register occurred, the gpio "output" pins would be set to "output" mode momentarily (and then cleared to "input" mode by the ioAbort event). After the ioAbort event was cleared, the gpio pin state would remain as in "input" mode unless software restored the gpio configuration register again. |
|
|
Fix/Solution:
In FPGA version x0346 and later, an ioAbort event still forces gpio "output" pins to "input" mode, which allows external resistors to pull pins from either high or low for a "safe" state. The resistors may or may not be present depending on the PCB assembly. However, during an ioAbort event, service writes to the gpio configuration register no longer cause momentary "output" direction. After the ioAbort event is cleared, gpio pins will resume their previous "output" state without software intervention.
|
|
Intermittent Dual String Discovery Failure |
|
|
Reference Number: FP 332 |
|
|
Type: Bug Fix |
|
|
FPGA Version: 0x0346 |
|
|
Problem/Cause:
Probe, resetRequest, and resetComplete were repeated on the wrong port (OUT port) when using the OUT port as the primary port during SynqNet node discovery. As a result, SynqNet node discovery may have failed. This problem was caused by the probe, resetRequest, and resetComplete being repeated on the wrong port. This issue existed in all FPGA versions prior to x0346. |
|
|
Fix/Solution:
The FPGA code was corrected in version x0346 and later releases. This fix is required for systems using the Dual String or IN Port String topologies. This fix is NOT required for systems using normal OUT Port String or Ring topologies.
|
|
sqMac - 0x00 will be read for dataType NULL |
|
|
Reference Number: FP 323 |
|
|
Type: New Feature |
|
|
FPGA Version: 0x0346 |
|
|
Description:
0x00 will be read for dataType NULL in the status and feedback packets.
|
|
Intermittent Fault Recovery failures |
|
|
Reference Number: FP 315 |
|
|
Type: Bug Fix |
|
|
FPGA Version: 0x0346 |
|
|
Problem/Cause:
Some nodes may have intermittently failed to recover from a fault in some network configurations. If a node failed to recover, it exited cyclic operation and correctly triggered ioAbort logic (forcing outputs into "safe" states). Any nodes "downstream" of the failing node (after it switched ports) were also likely to exit cyclic operation. For any given node, there was only a small schedule window where this error may have occurred. There were two causes of the problem. The rcvMiiBuffer did not clear itself after an overflow in some situations. Also, a packet fragment generated by the fault recovery that happened to have valid CRC (but no data) would confuse the packetFifo, and corrupt internal node configuration bits. The first issue existed in FPGA build versions x0341 through x0345 and their branches (MPI release 03.02.00). The second issue existed in all FPGA versions prior to x0346. |
|
|
Fix/Solution:
The fault recovery problems was corrected in FPGA build versions x0346 and later. The rcvMiiBuffer will correctly clear itself after an overflow. The xmitMiiReg will release xmitHoldRcvBuf within 8 clocks after a fault, preventing an rcvMiiBuffer overflow. 8 clocks will guarantee that there is minimum packet spacing between the packets. sqPll will ignore the second SYNQ packet after fault recovery (in case packet delays were effected in the cycle where fault recovery occurred). The rcvMii will correctly handle a valid packet with data length of 0. No software changes are required.
|
|