August News

Freescale K60 Kinetis

Freescale Kinetis K60 board

Freescale K60

We have been working between projects on the Freescale K60 Tower kit for almost a year. The board was shown here in August, 2011, although purchased a little earlier. Our real-time tick based kernel, some experimental profiling, and instrumentation are running on the K60.

The initial proposal of testing the Ethernet has been abandoned, as our distributed I/O will be CAN based. Software has been written, but we need to develop some small interface boards with transceivers on them, as most of our CAN evaluation boards do not have transceivers. (A saving of just over a dollar!) Any Ethernet we select will not be real-time, as we have no way of testing that without source code to the various automation vendors' packages, or “hooks” into the networking microcode for instrumentation. The scheduling of very fast response times on EtherCAT for the TI Sitara PRU required code to be written by TI (As mentioned in July news). Even prior Freescale networking in the QorIQ series was microcoded by Freescale, which looks like it will move to C in the Layerscape System Architecture as the QorIQ family gets an ARM core besides the PowerPC.

Tales from the woods

Almond blossom with blue sky

The almond trees are blossoming in McLaren Vale. We visited this spot last year—amazing how quickly a year goes by. We have been writing embedded software for just over a year in McLaren Vale. (Also K60 for a year).

We have programmed a large number of evaluation boards and done some projects for customers, however, debugging these boards over a JTAG is frustrating to say the least. The “Blue Sky” on the left is what is needed to try and get some sense in this ARM world, as there is very little that was portable between projects. Admittedly, most of the work was board support and peripherals, but the ARM cores, interrupt structures and a host of other evils are inconsistent.

The Freescale Kinetis work was resurrected due to the Vybrid announcement and the iMX6 shipping. We will resurrect our iMX53 work as well, so that when the Vybrid ships, we will be able to leverage our prior work. The TI Concerto has dual cores, one an ARM, the other a DSP. When we first looked into Concerto work, the learning curve looked manageable, but with the Freescale Vybrid announcement and greater similarities between the Cortex M and A cores, it looks like a better bet, plus the clock rates are a lot higher. TI only ships the multi-core OMAP4 and 5 to high volume wireless OEMs and ODMs, so we are not going to invest much more in the single core Sitara. The Samsung quad core devices ship with $129 evaluation boards and Android software, but their Ethernet is external on USB. The iMX6 is shipping on Qseven modules, which will hopefully be at ElectroneX from congatec, in Sydney next month.

We had one last look at the Xilinx Zynq before choosing a controller for the level above the distributed I/O. The Zynq-7000 Technical Reference Manual weighs in at 1691 pages, but we are not too sure how to communicate between cores (only skimmed through the manual and three other documents). The boards are starting to ship, plus next month is ElectroneX in Sydney, so hopefully we can visit someone with a board before plunging into another deep distraction. The Sitara ICE board at five bottles of red wine was easy to put away into it's box at the bottom of the evaluation cupboard, but the Zynq boards cost a touch more and if anything like the Virtex-II Pro with PowerPCs, then will take a bit more than two weeks investment.

The distributed I/O is pretty much decided. More of this later.