[Oberon] Maybe this would do [was Unlimited Oberon System for any board]

[Skulski, Wojciech]

Darek:

  a few general remarks concerning such boards, based on my experience with BeagleBone Black which is kinda the same idea. Please open the picture of the BBB on their website https://beagleboard.org/black. Then open our website FemtoDAQ.com and look at the photos. We designed a stack of two add-on daughter cards (named "capes" in BBB parlance) according to BBB specs. Then we put the instrument in a box. The BBB is at the bottom, then two "capes" on top. Based on this experience, here are the observations which are true for any such Single Board Computer; also for the Olimex.

1. The SBC is providing connectors (look at the photo) at certain locations. This defines the size of the application box and the holes. You have only that much space as provided by the SBC. Since the connectors are at two sides, you can only grow the box side wise. If the connectors are around, you cannot grow at all.

2. The BBB cannot be embedded deep inside a large box because then the connectors cannot be reached. For that reason I redesigned the BBB and stripped it of any connectors. I rather routed all the signals to multipin high performance Hirose connectors. The redesigned clone named MicroBone can now be embedded. At the same time, MicroBone cannot be used w/o a motherboard, what increases the cost. You can see the MicroBone on www.skutek.com, 5th item in the left menu titled "  ARM System-on-module". Look at the photos. Redesign was necessary if we wanted to use BBB in an instrument whose size exceeds the original dimensions. Exactly the same will be true with any such board. 

3. The foregoing means that you can easily apply the SBC for applications which the designers envisioned. You are out of luck if your application does not fit into their mindset. 

4. Investing your development effort into BBB or Olimex or RPi will keep you locked in the designer's box. This is manifestly true in case of Rpi, where cloning is not possible. You might say "but BBB or Olimex are both open HW, so there is way out". Let me tell you, barely so. I confirm the effort quoted on the Olimex page (half a year). It took me about that long to complete the MicroBone. Now, how many "hobbyists" are ready for this when their project outgrows the original box? I was ready for this and I chose BBB because I was planning the redesign. But I am kinda special. I am not afraid of such projects. Not many SBC users can consider the redesign of that complexity. I rather say "if you choose a particular SBC, you are stuck forever". 

5. Look at the pin expansion interfaces offered by the SBC. The BBB is similar to Rpi or Arduino. There are few pins and the 0.1" pin connectors are low performance. The expansion boards are severely limited by the 0.1" connectors. The rationale was that these expansion boards (capes or whatever they are called) are meant to be almost the breadboard quality limited to hobby projects, simple college robotics, etc. There is a certain disparity between the very high performance CPU chip and the low performance expansion boards which are permitted by the 0.1" connectors with very few slow GPIOs. Here the SBC designers basically decided what you can, and what you cannot do, even before you started your project.

I replaced the 0.1" crap with high performance Hirose connectors with integral shields. I could interface MicroBone with FPGAs and reach much better throughput than afforded with the 0.1" crappy original headers. I achieved this at the expense of heavily redesigning the SBC. Not many SBC users can complete such a redesign.

6. Modern CPU chips pack enormous collections of silicon cores on chip. The AM3358 is a good example. (This si the BBB chip.)  However, these cores are not available all at the same time, because the chip vendors save on the pins. The AM3358 uses a "pin mux" where most pins can be switched among on chip peripherals. This is (mis)used by the SBC designers. BBB is a good example. The BBB designer decided that it was neat to put the mass storage on the SBC, and for performance reasons he used eMMC. What a great idea! So the BBB comes with Linux on board, it is running great, and you do not have the memory interface even though it was connected to the expansion. But it was used for the eMMC, so these pins are dead for the user. The first thing I did with MicroBone was to remove the eMMC, so I could use the memory pins for the FPGA. This is an example how the SBC designer made decisions which are hard to change. 

In the original discussions the designer advised to remove the eMMC with a heat gun if you needed the memory interface. What a great advice! Later on some smart engineers figured out how to disable the eMMC and effectively take it out w/o desoldering. This was pretty extreme taht in order to use an obviously needed feature (memory interface) an important part of the SBC had to be dismantled. 

Other SBCs will come with their own load of similar problems. The reason is that the SBC designer is trying to address the most typical applications to gain on the volume. You can rest assured that *your* application will not be among the *most typical*, unless your application is fairly trivial. 

So these are my assessments of any SBC on the market. Yes, they are very useful for the most typical cases they are meant to serve. If your application is not the most typical, you better learn how to develop an eight layer board.

Thanks,
W.

________________________________________
From: Oberon [oberon-bounces at lists.inf.ethz.ch] on behalf of Darek Maksimiuk [darek.maksimiuk at gmail.com]
Sent: Friday, May 8, 2020 6:21 AM
To: ETH Oberon and related systems
Subject: [Oberon] Maybe this would do [was Unlimited Oberon System for any      board]

 Soon (I hope), there will be a new version of an ARM-based system from Olimex:
      https://olimex.wordpress.com/2020/04/15/the-industrial-grade-40125c-open-source-hardware-linux-board-which-is-hardware-compatible-with-a20-olinuxino-lime2-but-with-st-microelectronics-stm32mp1xx-soc-stmp1-olinuxino-routing-is-complete/

Could be this HW a good starting point for  porting the  Project Oberon  (or A2, V4, or even S3) ?

Olimex offers also boards that are based on the Open Source Hardware initiative (i.e. https://www.olimex.com/Products/ARM/ST/STM32-E407/open-source-hardware
Regards,
   Darek