On Wed, Dec 26, 2012 at 1:45 AM, Arnim Littek <arnim@actrix.gen.nz> wrote:

On Wednesday 26 December 2012 19:09:04 you wrote:

...

I wrote this reply with the idea it might post to the entire group. If you are ok with that, please make it so.

I'll leave that to you. The degree to which you release information at which stage, is your call. You have in mind how much you'll release to the Oberon group, and the Kickstarter group...

Ok, I'm going to copy the group, as I think discussion of an embedded project & Oberon is a natural interest area. I don't feel much need to be secretive with the Oberon group. If the KickStarter funding does make, it likely will be from people who follow KickStarter, not people who follow Oberon.

...

I carefully did not mention Oberon in the KickStarter project. I think other people on this list are much better compiler talent. I thought they might make suggestions for a board that would interest them. I thought they might suggest that I use a MIPS chip for my board, and make their own KickStarter project to deliver a compiler and system compatible with it. And if not their own KickStarter project, then let me know if they are interested, and how much calendar time and what funding it would take to make it real, and I could try fit that in and hope.

Hmm. Of course it is hard to know what other lurkers there are on the list, who (like me in this case) pop out of the woodwork on this occasion.

...

Yes. My thoughts are that the loader would put code into Flash, and most of the variables would live in RAM. I mis-spoke, I was using EEPROM as a shorthand for "stable but difficult to write to" memory. I am unsure where the initialized flag and the pointer to additional modules would need to be for the modules loaded by the bootloader.

There are tradeoffs to be made, and some of these (speaking in terms of C compilers for embedded systems) will involve what variable goes in what storage. For some micros, running from Flash is pragmatic, because it is nearly as fast as SRAM, but some micros really want stuff in SRAM if speed is an issue. Then loading variables from Flash to SRAM is part of the package - sometimes handled almost invisibly, with appropriate linker setup. I don't know how Chris Burrows' products handle this, but presumably he must have analogous memory management capability.

I have placed a .zip file on Google

...

Docs<https://docs.google.com/open?id=0Bx35YZ-0A4ffTTR1ZlVlQmhxOXM>. At the moment, anyone with this link can grab a copy. The files named RS232_Control_5_003 are the EAGLE files for the legacy (16F877) board from April 2004. This is what I will deliver for the KickStarter $5 reward.

I'll see if I can go over it tonight.

...

The files named RS232_Explorer_5_002 are the EAGLE files for the companion RS-232 explorer board. This is not promised in any of the KickStarter rewards. It is this board that is photographed toward the bottom of the project.

The new (revision) board that I will make if the KickStarter project funds does not exist. One of the KickStarter activities is that I will discuss the choice of CPU core before I make the board. That should be clear in the KickStarter language. If it is not clear, I need to hear about the confusion very clearly, and clarify the language so that no one can be confused. In principle, KickStarter is about a new way to cultivate creativity and not a store.

I've got that message, but that doesn't mean everyone else has.

...

...

I'm past through-hole electronics. For any prototype I'd build for me, I'd prefer surface mount. Unless you go for more exotic packaging like fine pitch QFN/DFN/BGA, they'll be easier to hand-place anyway. And any decent sensor packages I'd want to hang off it will be SMT anyhow.

I am very interested in the above comment. I made a daughter board for the PIC32 evaluation board. I used the Hirose connector that Microchip used, and ran the lines out to 0.1" thru holes. I tried using a Bi/Sn/Cu/Ag solder to have something low melting. I could not buy any solder paste with Bismuth in it, so I used homemade (I'm a chemist). I had a fair amount of difficulty. It was completely impossible for me to solder things that tiny. I always ended up with a blob that shorted everything. This was in 2008. I took the spare boards and had a board house assemble it for me. I have new appreciation for the term FUBAR.

I'm interested in a process that I can do in a home lab, or that students can do in a teaching lab. Do you think that 20 year old college students can work with SMT parts without needing high dollar equipment?

You mentioned a senior electronics tech or something earlier on. Talk to him about solder flux and what it does for fine pitch ICs. It helps if the soldering equipment is useful, but I use some fairly old gear at work, and get by. Good soldering iron tips that are designed for a particular role, make complex-looking tasks quite simple, especially when applied with decent solder flux. The above assumes you have some temperature control over your iron too.

Decent solder helps. All good solder for fine pitch ICs has solder in the core - which runs out and tins your tip, as well as the electronics you're soldering. Get fine solder too - fat solder is never used by professionals. The roll on my shelf here is 0.56mm, and I'd not use any bigger stuff for my private work.

hmm. I suppose I should buy a cheap bag of out of spec parts and have another go at it. I suppose I could make a board that just repeats lots and lots of the PIC32 & the USB part artwork. Does anyone sell "practice your soldering" kits or does everyone John Wayne it?

...

...

Have you done anything Oberon-ish on the PIC16F877 target yet? I can't imagine that kind of pain.... But if you've done your previous work on that board in PIC asm, you have undergone all of it, esp. if you've done enough code to start fiddling with their bank switching - Ugh.

I have hand compiled Oberon code to run on the PIC16F877, to try to develop the right code patterns. This was before I read Wirth's paper that announced an even smaller language for a different PIC part. I was ok with all of the bank switching, and I had interrupt routines that worked in that environment. (I think I feel a headache coming on with the memory.) I could probably make a compiler, but my skill level is so low I would take the rest of my life to finish it.

For my money, almost any micro on the planet, 8 or 16 or 32 bit, has a better architecture and instruction set than PIC. The PIC architecture propagates so many bad habits - it should have been killed off at least a decade ago.

And the C compiler tools are hobbled by the limitations of the architecture, and generally not of the highest quality, in my experience. Even the best of the lot, Hi-Tech, purchased by Microchip a year or two ago, has it's problems. I report bugs every time I do any real amount of work with it.

I like regularity, simplicity, completeness. I think the MIPS has that, but I have not done a survey of the other chips in the market. It is probably not good engineering to choose a chip based only on the cpu core. It is tempting to think that an Oberon system can push quality code onto the part in a clean environment. It is also tempting to think that it is ok to write all the code from the silicon on up in an embedded environment, so that the "lack of legacy code" does not matter. It is tempting to think that Component Pascal, with its paradoxical smaller number of productions but also extensions for large projects would be a good choice. Yet there is not proof of concept in the marketplace.

...

I did bit bang RS-232 on 2 channels off the daughterboard. I monitored all sorts of RS-232 signal lines, not only RX/TX. I don't remember if it maxed out at 19,200 or 9,600. I recognize that this is not good performance, and it demands too many signals to make a commercial project. The thing that it did do was to convince me that, for a short time, I actually understood RS-232.

bitbash has it's place, but UARTs are cheap on a decent micro. PICs have never had them in oversupply.... like they never wanted to do any serious comms. Some AVRs can be set up to have over half a dozen USARTs, all with DMA support, if you really want. They don't cost any more than PICs... Same goes for the ARM Cortex parts available from a dozen different silicon vendors, if you can climb the learning curve fast enough. That's where Chris B has hands-on experience....

...

I think that is what I am supposed to give students. They should understand stuff when they finish a project, so that they can have a deeper understanding of the fancy stuff they use later in their career.

Interesting that you, as a chemist, are foisting good hands-on embedded systems experience on your students.

...

...

I don' t know how you'd get Chris Burrows to chime in - he's been there and done a lot of this, it seems.

I bet Chris would answer any question actually directed to him. Hi Chris !

He does seem open and cooperative, rather than competitive/secretive sort.

...

Fire/Police/Emergency responders. I planned to use a display like the B&N Nook, the ePaper stuff. This is all broad brush conceptual, I don't know if the PIC32 has enough output signals to interface with the ePaper displays. I have lived most of my life in Hurricane country,

The PIC32 family has grown since 2008, and LCD displays are on their list, although I am not familiar, since I don't design in displays in my normal course of activities, work or otherwise.

...

Thank you for your comments. And, as you have noticed, my KickStarter project is absolutely silent on Oberon. I want to keep the KickStarter promises fairly minimal.

Your comments on managing scope apply also in the commercial world. Managing expectations is not a trivial task.

Yeah. Opening up design questions in a KickStarter project runs the risk of making some people think I'm indecisive while I'm hearding cats and pulling out my hair.

I'll go look at your schematic shortly,

Arnim.

-- Aubrey McIntosh, Ph.D. 211 E. 5th St. Morris MN 56267 (512)-348-7401