July News


Texas Instruments Sitara

Fedex parcels

Here is the “grocery” shopping we promised to show in our June news. Several boxes from Texas Instruments (three 24VDC → 3,3VDC switchmode supplies), the Sitara-based ICE, and a Luminary Micro B92 kit for a joint project with a North American customer.

TI box kits

The three large boxes at the top had single PCBs about the size of a matchbox, but thankfully, TI is generous and pays the freight. The bottom left is the Sitara-based ICE in a box a bit small for a software disk, while the bottom right is the Luminary Micro B92 kit for Ethernet and CAN work.

The Luminary B92 board was tested with the usual “out of the box” demo, connected to our CrossConnect JTAG probe and the Rowley Associates toolchain. It was also connected to the evaluation versions of the Code Composer Studio 5 toolchain from Texas Instruments (using the built-in JTAG debugger). Like the other ARM Cortex M cores with on-board Flash—nice and simple.

Industrial Communications Engine

ICE board top

The low-cost Sitara-based ICE board (top view).

The AM3359 is a Cortex-A8 core running from 275 to 720 MHz, well beyond the Cortex M3 devices. It has external memory which removes the memory wall for devices that only support internal Flash and SRAM.

The board has a built-in JTAG debugger, but no trace port, however, it is definitely one of the best sub $100 boards ($99) we have seen in a while.

The dual Ethernet connectors are for a switch that allows these boards to be daisy-chained as per EtherCAT cabling. The PRU communications microcode for the internal networking cores was written by Texas Instruments, however, long-term commitment would need to check up on licencing fees. (We think that the microcode is part of the AM3359 and included in the price of the device, however, the EtherCAT derivative products will require an investigation into certification and fees).

For testing EtherCAT, we would need to invest heavily in time and equipment; certainly more than the two weeks we dedicated to date. Additional networking interfaces include isolated Profibus and CAN. The 24VDC supply was useful for industrial interfaces, but the 24V → 5V power converter did get a bit warm. We bought several 24VDC evaluation kits (TI) to compare their efficiencies, but we may choose a transformer based design where the drops are due to winding ratios rather than duty cycles. This is far down the road, and will be external to the processor board.

The Sitara device has been marketed as a $5 long-term solution by TI, which should show up in plenty of new industrial designs.

For more info, see http://www.ti.com/tool/tmdxice3359


ICE board top

ICE board bottom view.

Note to self: Put in ground pins for connecting a scope when making any board. We have done this since 1980, however, the practice seems to have faded with evaluation board vendors. (Including Freescale on the Tower boards with a lack of decent probe points—particularly the serial board).

For software development, we used the Code Composer Studio toolchain, which is available for evaluation from Texas Instruments. We used the supplied GEL files for initialization, and wrote software from the main entry point. For startup code, there is an internal ROM (and a lot of reading required to get the code to run from startup in Flash).

The Sitara has a graphics core, but a different board would need to be purchased. We were kept busy with this board, but will not be tackling Sitara projects yet without external funding.


Available Design Notes and Software

Be warned, the AM335x manual is comprehensive at 4301 pages (SPRUH73F, June 2012). You will still need to do extra reading—the documents are free to download. There are Wiki pages on the TI website, where you can ask questions which are answered by TI staff.

The software requires a download with a paper insert in the ICE box giving the link. We could not find the software specific to the ICE board, even though the forums said it shipped with the product. According to the website, the software is a download, and not in the box. We looked on the website but the software for the SYSBIOS Industrial SDK at http://www.ti.com/tool/sysbiossdk-ind-sitara is for the IDK, not the ICE. (The download was dated 22 June 2012 when accessed to update this page on 16th August). The IDK software compiles; we can connect to the JTAG debugger; the GEL files initialize the device;—but as you single step into the weeds, you realise that there is no CPLD on the ICE board, nor several other peripherals like motors and a DSP.

The forums claim that the software is the same, but with the limited time we get to evaluate designs, if they don't work within the first few days of opening the box, they get shelved. It would be possible to edit the sample source, which is already a huge help compared to starting from scratch with the 4301 page manual, but all we ask is that the source be provided in one place as per the example startup code that was flashed into the ICE (sequence the LEDs on a I2C peripheral), and that it can be stepped without wandering into the weeds. For serious consideration of the AM3359, we suggest purchasing the more expensive Industrial Development Kit ($895), as the $800 difference quickly disappears when you are wandering aimlessly in the weeds. The ICE board is not probe friendly and you should really consider a trace port for justifying work at this level.

Our Impressions

The processor is impressive, particularly at $5 for the device (in volume). It is similar to the AM3358 device used on the BeagleBone board at BeagleBone.org which sells for $89 from Digikey (plus shipping). The BeagleBone expansion connectors are also more hacker friendly without the need for the hands of a neurosurgeon and a couple of magnifications short of an electron microscope to see where to cut tracks.


Tales from the woods


Wooden desktop

Part of a 3m pine slab sanded and coated with a two-part polyurethane. Therapeutic work that gets as far away from writing software as possible, and it does not date as fast either. In between the sanding and coating, we did two weeks of work on the Sitara board, but have to admit that the 4301 page manual was only fleetingly glanced through, so we would be a bit wary of accepting any work based on this device in the short term.

We are looking at various ARM-based devices for the controller above the distributed I/O level. If you have similar ideas and would like to share costs, please contact us.

Other work was looking into various patents relating to I/O interfaces. If you thought the Apple round corner patents were granted by some disinterested patent officer, then the gentleman/ lady handling the I/O patents simply ignored loads of prior art. We will be using integrated devices where the ice is a bit thin and rely on Freescale, ST Micro, or Texas Instruments to take care of the patent trolls.

For interested readers, the patents related to contact wetting, multiplexing to reduce power, and then for some prior art. The numbers of the US Patents were US3229040, US4745262, US5621250, US5729167, US6584197, US6600242, US6800965, US20030038017, US20070040638, US20070206731, US20070230686, and US20110264834.

Eventually this product will see light of day (between other projects)!