Misplaced Pages

Revision as of 19:15, 30 October 2011

This article is written in British English, which has its own spelling conventions (colour, travelled, centre, defence, artefact, analyse) and some terms that are used in it may be different or absent from other varieties of English. According to the relevant style guide, this should not be changed without broad consensus.

This article was nominated for deletion on 9 June 2011. The result of the discussion was speedy keep.

Not a microcontroller

Something intended to be used for control purposes must power up and run a a control program. By the time you add hardware to a KIM 1 to store an application program, you might as well build a board with the 6502 on it alredy and skip the expensive and useless mask ROMs that your application isn't going to use anyway. We are needlessly confusing the reader by lumping everything since ENIAC in with boards designed for control systems that had at least an EPROM socket on them. If you were building a controller in the 1970's, you might have had a KIM 1 kicking around the lab for education purposes, but your production hardware would not have used it, or you would have spent at least as much on expansion as the KIM cost in the first place. Yes, acuracy is a point of view, I'm sorry this distresses some editors, but I really can't understand why they want to lump in demo boards like the KIM 1 with microcontrollers. Plesae find me a reference saying a KIM 1 is a micrconcontroller board. --Wtshymanski (talk) 13:36, 7 June 2011 (UTC)

Imagine a heart-lung machine controlled by a stock KIM 1. Switch on the power, then find the cassette deck and cables, plug into the KIM, press buttons on the keyboard (was it a single button or some arcane jump instruction you had to enter from the keyboard? I can no longer recall.), press "PLAY" on the cassette deck, then press "RUN" after the tape loads...yeah, that's the very epitome of process control. Even in 1977 we could do better. --Wtshymanski (talk) 13:40, 7 June 2011 (UTC)

The rather nice, but wordy, introduction says

A single-board microcontroller is a microcontroller pre-built onto a single printed circuit board. This board provides all of the circuitry necessary for a useful control task: microprocessor, I/O circuits, clock generator, RAM, stored program memory and any support ICs necessary. The intention is that the board is immediately useful to an application developer, without needing to spend time and effort in developing the controller board.

• • A 6502 is a general purpose microprocessor, not a microcontroller.
  • Yes, a KIM-1 is a single board...at least as far as it goes. Add an EPROM board to it and it's not single-board any more.
  • The KIM-1 board does NOT provide all the circuitry for a useful control task, partticlarly, it lacks any user-programmable EPROM. No real-time clock or battery backed RAM either but that's probably not essential.
  • The board is not immediately useful for control applications because you have to find an EPROM board and adapt it to the KIM-1 - and there's no provision for mounting any daughter boards to the KIM-1, so you've got to figure that out as well.

I'd also point out that if you were mad enough to have bought a KIM-1 for control purposes you're not likely to use all that mask ROM on the board anyway, and the configuration of the board doesn't lend itself to compact integration into any product.

I bought a KIM-1 31 years ago for use in my thesis project. I had a lot of these difficulties to overcome just to develop my system software on it. The thing that I wanted to build wouldn't have used any KIM hardware at all, particularly not the nasty expensive mask ROM/IO chips. We would have *killed* for an Arduino in those days; would have been a palace to us. --Wtshymanski (talk) 21:02, 7 June 2011 (UTC)

Your mistake is to equate 'microcontroller' with 'to be used for control purposes'. Most of them sold, certainly the single-boards of this period (the actual subject here), certainly the low-cost boards, weren't sold in the expectation that they'd ever be used for this purpose to the full, but rather that they'd be used for education and training. Whilst your basic point that the KIM 1 wasn't suitable for controlling anything vital is correct enough, it's simply irrelevant - no-one was suggesting that it was, nor that 'microcontroller' requires it to be so.

As to the practical aspect of how to use a board like this as a controller without supplied NV memory, there are many ways round this. From memory, as all of these are ways I did it myself back then:

• Keep the power on. It's probably not a live controller that must come up clean from a cold boot, just a class demonstration. Battery back-up if needed.
• Replace the boot ROM chip with another, programmed for the task in hand. As the single-application ROM only needs to do one task, not to offer a full monitor, this is quite easy.
• Add another EPROM to the bus. It's a 6502 and there's no need for DRAM refresh, so this is an easy piece of breadboarding.
• Add an additional hardware board (quite possibly doing some A/D task or typical 'controller' function) and use this board for extra EPROM too. Many of these 3rd party add-on boards were already designed with just such a socket, often they could map their addresses so that they could take the boot ROM position (the original ROM was also pulled and moved to the add-on board).
• Duplicate the entire controller to a modified home-brew. Copyright wasn't such an observed issue in those days, so it was pretty common for a class studying with the full-monitor and keyboard version of the controller to also make their own minimalised copies of it. The boot ROM was copied, as was the PCB layout. The hex keypad, display etc. might be ignored in favour of adding some I/O drivers or an EPROM socket. A similar practice continues to this day with the Arduino and its minimalised Boarduino variants.

As to the issue of nomenclature, then that was the marketing of the period. The smallest of single-board machines were sold on the basis of vastly inflated claims of computing power, not on how small they were. Despite this, the topic is the topic, not the name attached to it in the past.

As a Brit, I'm not particularly familiar with the Kim 1 / AIM 65 - we'd be more likely to have the early Acorns and the Microtan 65. It's listed here because there's already some wiki coverage of the Kim. Mostly though, it's a good example of the two related machines were one was a minimal controller and the other a rather more sophisticated device aiming at being a general purpose computer. Not a clear distinction in those days, as even the larger machine was still pretty small, but it is a good and clear example of the point being discussed in the article. Andy Dingley (talk) 11:53, 9 June 2011 (UTC)

Look at the introduction to the article. Look at how much hardware you've described as added on. You basically descirbe throwing out the store-bought KIM or AIM and building your own controller. That's not what micrcontroller boards are aimed at. Something like the Arduino board or BASIC Stamp can be put into a small-volume product without any engineering to figure how how to expand it, power it, and provide non-volatile program storage. That's the point of a single-board micrcontroller. How does it serve the reader to equate antique development boards with something that is purpose built to fill the spot on the diagram labelled "Controller goes here" ? If a KIM 1 is a microcontroller, is an Apple II? (At least they had a spare EPROM socket.) A Commodore 64? An IBM PC? (Lot of spare sockets!) Not really.

It doesn't serve the reader's interests to claim that any random board is a microcontroller board. You might as well claim an IBM 1800 was a microcontroller - sure it needed an air conditioned room and weight 2000+ lbs, but by golly you could hook it up to digital and analog process signals! (That took another fridge-sized box of gear).

I don't know how you see this as being a good and clear example as you've tagged anythin with a microprocessor in it as being a microcontroller - and they simply aren't. Flip through the ads in BYTE or its successors - you'll see lots of "single board computers" with hex keypads, and you'll see ads for microcontroller boards (usualy in the back in 1/8th column ads, and later in BYTE's run). Distinct products, distinct purposes. YOu might have used a KIM 1 in 1977 to control your automated chicken plucker because you didn't have many other choices - but you would have spent months in the lab building or integrating various expansion boards and card cages, all of which would have cost several times the price of the KIM. You would have had to figure out how to package it. And if you planned to sell more than a handful of automatic chicken pluckers in a year, you would have thrown all that engineering out and started over with a custom board because you didn't want to pay for the custom mask ROMs you weren't using, for the cassette and teletype interfaces you weren't going to use, for all those card-edge connectors that would get full of feathers, for a hex keypad and display that weren't mounted in the right spot to be acutually usable. Whereas, with a real microcontroller, you bolt it to the box, run the wires in, and load code and run it. Totally different product. There's no need to talk about development boards that weren't intended to be used in controller applications when we can actually talk about boards intended to be used for control.

"Don't lie to the readers" should be one of the principles of Misplaced Pages. --Wtshymanski (talk) 13:27, 9 June 2011 (UTC)

As you've clearly not bother bothered to read it, I present the first para of my last replay once again:

Your mistake is to equate 'microcontroller' with 'to be used for control purposes'. Most of them sold, certainly the single-boards of this period (the actual subject here), certainly the low-cost boards, weren't sold in the expectation that they'd ever be used for this purpose to the full, but rather that they'd be used for education and training. Whilst your basic point that the KIM 1 wasn't suitable for controlling anything vital is correct enough, it's simply irrelevant - no-one was suggesting that it was, nor that 'microcontroller' requires it to be so. Andy Dingley (talk) 14:06, 9 June 2011 (UTC)

I didn't believe it the first time I read it, either. Please provide citations showing a KIM-1 was sold or promoted as a "microcontroller". It wasn't sold as a controller in ANY period. What is a "microcontroller" for, if not to "control" something? Or is this more of the curdled thinking that we've gotten used to after so many years of clicking on "Start" to shutdown the computer? "Single-board microcontroller" is not synonymous with "single-baord computer", but is only a subset. If you wanted to control something today, would you be reaching for something like a development board or microprocessor trainer that requires tons of extra glue to make it usable, or would you say "Oh S*T, I haven't got time for this, where's the ad I saw in "Circuit Cellar" ? --Wtshymanski (talk) 15:16, 9 June 2011 (UTC)

So just what is your point? - you've changed it again.

Most of the single-boards sold for controller purposes in the '70s weren't fit for production-grade use as such. This is not a problem, they were training devices instead. Trainers cost a hundred, reliable controllers based on the same processor family might cost five times that.

The Kim-1 was not the world's greatest computer or even controller. Not even in its day. However it was a popular and well known example. It also came in the "little and large" formats (with the AIM 65) that are why it's listed here. If it wasn't any different from the AIM 65, and wasn't intended for a rather more restricted purpose, then why two products?

A "single board" isn't restricted to a literal "one board and no more". Many had a separate board for PSUs. Even today with the Arduino, one of its great virtues is its easy expansion through a 'Shield' to a two-board device. The defining point about a single board machine at this time is that it wasn't a card-cage or bus machine and that it could execute some minimal code with just the single board in use (i.e. processor, bus control and some of the memory were co-located). This doesn't preclude further expansion for some tasks. The Kim 1 meets this.

Adding additional 'glue' to a single-board is a common task. As the glue often varies according to the arcane needs of each application, it's impractical to make one board that supports everything you might want. In the early '80s, there was also the important aspect that I could make a PCB to house a couple of 14- or 16- pin ICS, but laying out multiple 28- and 40-pin devices would be either beyond me, or certainly a serious effort. It made a lot of sense to buy in a working board with a working bus on it, and then add my own board for simple (but particular) IO tasks.

Were they called microcontrollers at this time? Not usually. I'd be interested to know just when that phrase was coined. However the concept, and their use, was around long before that word became the standard way to refer to them.

What is a "microcontroller" for, if not to "control" something?
Mostly to teach students how to control something, probably using a better and more robust of a similar device with a processor from the same family.

"Single-board microcontroller" is not synonymous with "single-baord computer", but is only a subset.
That would be why there are two articles: one here at Single-board microcontroller and one at Single-board computer.

If you wanted to control something today,
This isn't an article about what I'd do today, it's mostly about the historical situation of the late '70s and '80s. These days I might use an Arduino (cheap embeddable component, and its own dev board, all for £20) or else recycle an old PC with squiggaflops of processing power that I don't need, but already have gathering dust. Neither are relevant to the early developments of 30 years ago. Andy Dingley (talk) 15:54, 9 June 2011 (UTC)

it's mostly about the historical situation of the late '70s and '80s.

So that means yes, there was lots of PL/M in use for this task. On the 8080 platform rather than the 6502-based Kim 1, but those big blue cubes of the Intel SDK machines were very often cross-compiling PL/M for 8085 embedded controllers.

Your rhetorical question "Are there really microcontroller development systems out there currently using PL/M? " in Misplaced Pages:Articles for deletion/Single-board microcontroller is an irrelevance because this article is mostly about the historical development of single board controllers, and "currently" has nothign to do with it. Andy Dingley (talk) 19:05, 9 June 2011 (UTC)

History of microcontroller boards

It wasn't at all clear to me that you were talking about history only; certainly the lead illustration is a pretty contemporary example of the microcontroller. I don't know that history alone is all the article needs; single-board microcontrollers are quite popular (Arduino, and so on) even now. --Wtshymanski (talk) 19:46, 9 June 2011 (UTC)

If there's going to be a history, then there should be a section titled "History". References are going to be a drag. When did controller-type boards (as opposed to demo boards) start being advertised? I have a few BYTE magazines from the late '70s and early '80s, I suspect the ads will start at some point.

Maybe the history of single-board microcontrollers is coterminous with the history of single-chip microprocessors. Here is Intel in 1976 claiming the SBC80/10 as the "first" single-board computer, targeted at integration into embedded systems. There's a lot more chips on that board than you'd find on a modern system but it is definitely being marketed to OEMs as a bolt-down solution instead of roling their own 8080-based board. --Wtshymanski (talk) 14:17, 10 June 2011 (UTC)

The section you didn't bother to read is called "Origins".

Your Intel ref is interesting, because not much else of 1976 is so clearly on line. You might note that (just like the Kim 1) it doesn't use the term "controller" anywhere. It does however discuss PL/M.

This isn't the sort of device I was really talking about in this article, because it's too complex and too expensive. It's certainly within most definitions: it's single board and it can be used for control, but it's on the fringes. Even in 1976 there were cheaper ways of doing this - why spend for bus drivers that you'll never need for a single board installation? Can you even use this board on its own, without also buying a card cage that costs more than a Kim 1 does altogether? With its bus, cage, range of expansion boards and its UARTs, this is a device that is really aspiring to be a computer not merely a controller, and that's something that the article has identified as a separate piece of kit, right from the start. Andy Dingley (talk) 14:36, 10 June 2011 (UTC)

It would help me to climb down the ladder of abstraction. Could you point at some things you mean by "Single board microcontrollers", if possible, highlighting how they differ from what Intel is selling in this brochure? You don't need to use a card cage to use this board (but you can if you want). Unlike the KIM-1, you are paying at most for a few cheap bus driver chips that you may not use, compared to the expensive mask ROM/IO chips that contained the "Keyboard Input Monitor" program in the KIM. The SBC 80/10 has 4 sockets for user-programmable EPROMs, which you could burn with your Intellec development system. I haven't downloaded the user manual for this board yet, but I'm pretty confident it will show a stand-alone configuration where as soon as the power comes up, the chicken-plucking application program starts up with no user intervention. Intel is targeting this board at embedded systems devlopers who otherwise would have to use a minicomputer or minicomputer board or roll their own microprocessor controller. And Intel says in this pamphlet that if you're find yourself selling a lot of machines using this board, they will license the PCB artwork to you so you can build your own board and drop any features you're not using from the Intel SBC or add things you do need.

Google Books snippets show a few publications that talk about using this exact board as an embedded system. There's also a few snippets that use the words "single board microcontroller" but some of those are before 1975 when micrprocessors were uncommon. --Wtshymanski (talk) 15:06, 10 June 2011 (UTC)

"Aspirations"?

Beware the pathetic fallacy - I don't really think my KIM-1 languishes in a box in the basement dreaming of having a graphics display and a mouse. There must be a better way to phrase this. --Wtshymanski (talk) 15:25, 6 July 2011 (UTC)

Which bus is the internal bus?

You've got things like, say, an 8080, which has a full-up for-real external(off-chip) address and data bus. It's a Von Neumann architecture, if that explains anything, so you can put either program memory or data memory on the bus (in fact, you must, becuase there's no internal memory on-chip). Then you've got things like an 8051, which can also have an off-chip address and data bus, which might be used for data memory, or program memory (but there's also on-chip program memory), and it happens to be a Harvard architecture (though I'm sure with enough glue logic you could put programs and data into the same battery-backed static RAM if you want to.)

The way the computer is organized has nothing to do with the expansion bus. You can't tell from the card-edge connector that the chip on board is a Harvard architecture or just a very well behaved program on a Von Neumann architecture that has agreed not to try to write to its program memory. The thing that Joe Gas Pump Kiosk designer cares about is that if he spends an extra $4.95 per controller board, he doesn't have to worry about sweating his program down to 4K memory on-chip and can put it on an external ROM. --Wtshymanski (talk) 19:03, 21 October 2011 (UTC)

Don't lie to the readers

"Universally" Von Neumann? Pfui. An 8048 or an 8051 was not Von Neumann. A PIC is not Von Neumann. I wouldn't even want to guess which philosophy is more common in single-board microcontrollers unless I had a nice recent market report close at hand. --Wtshymanski (talk) 19:23, 21 October 2011 (UTC)

Of course not - that's the whole point. However these are the next generation, the single-chip controllers. The generation here, when board-level integration became available for a controller's budget, but before chip-level became technically practical, were the Z80s, the 6800s and the 6502s. These have to be a Von Neumann bus, because it's the only way to get their (single) bus out through the limited pins of a DIP package. A Harvard bus just isn't possible without either a board level processor that's already in multiple packages, a many-pin surface mount package, or else (as happened with the 8048 et al) a closer integration so that the buses and system RAM/ROM can remain wholly inside the chip package, and all that's needed externally is the I/O.

If this section is unclear, or you've failed to see the point, then by all means either fix it or tag it for someone else to fix it. This is an issue of English, or prose expression - fair enough. Yet what you're doing, and what you always do, is to take the line that "If Wtshymanski doesn't understand it, then it's Wrong" and you then delete the entire section. You are not the sole arbiter of technical accuracy. Andy Dingley (talk) 20:18, 21 October 2011 (UTC)

It's not prose expression, it's just not accurate and it's certianly got too many buzz words. When I laid out printed circuit boards for an 8048 processor, I could count 40 pins - and I knew it was a Harvard architecture. 8048s go back to (nearly) the "dawn of time", as far as microprocessors go. The limitations of DIP lead frames has very little to do with the design of microcontroller boards and nothing to do with "Harvard" or "Von Neumann" architecture. When the 4004 was built, they multiplexed *everything* to get it into a 16 pin DIP. The 68000 has "supervisor" and "user" modes available on the status bits, in principle there's no reason you couldn't have used the same chip pins for both program memory and data memory, with a status bit to decide which chips got the signals off the 40 pin DIP package. And today's single-board microcontrollers may use either Von Neumann or Harvard and that is the *least* important factor in selecting which board to buy.

If Wtshymanski doesn't understand it, it's certainly not clearly explained in the article he's just read; and very often it is wrong. I don't touch (or read) the maths articles, I can't tell when they are bullshitting me, but I can count to 21 without taking my socks and pants off. In that zone I will continue to be bold, as the encyclopedia exhorts us. --Wtshymanski (talk) 20:38, 21 October 2011 (UTC)

The 8048 certainly doesn't go back to the "dawn of time", it had to wait for chip scale integration to get to the stage where RAM & ROM could be on-chip (and so the bus didn't need to go off package). There's a whole generation, that was important for years, before this. Once again, everything has to be viewed through the "Wtshymanski goggles" - your generation had the 8048, so nothing different can have existed before this and the previous approaches can only have been doing just this, but in black & white. Andy Dingley (talk) 20:53, 21 October 2011 (UTC)

"the Z80s, the 6800s and the 6502s. These have to be a Von Neumann bus, because it's the only way to get their (single) bus out through the limited pins of a DIP package." Ah, no, not really. You just designate one pin, formerly an address pin, to denote I-space vs. D-space, in the manner of the PDP-11. Your program ROM responds to one set of addresses and your data RAM to another. Jeh (talk) 05:43, 22 October 2011 (UTC)

That said, I know of no single-board microcontrollers OR single-chip microcontrollers that implement a true Harvard architecture with its performance benefits. Yes, they commonly have separate flash RAM for code vs. volatile RAM for data... but that doesn't mean they can perform code and data access at the same time even though there appear to be "separate" buses serving different address ranges. Nor is it a given that a uP of this sort can't execute code out of the volatile RAM, unless they implement separate I and D space and provide no way to turn that off. This therefore becomes more a quirk of the implementation's topology rather than a true distinction between Harvard and von Neumann architecture. Jeh (talk) 04:01, 22 October 2011 (UTC)

Tell us more of this PDP-11 based single-board microcontroller.

Yes, you can do all sorts of things that aren't listed here. However the scope of this article is (or at least used to be) fairly clearly defined as those single-board controllers of predominantly the 1980s where board-level worked and was affordable with chips like the Z80, but before the single-chip controllers appeared and reduced the need to use a whole board. An Arduino isn't quite the same thing as a Microtan. This is an encyclopedia after all, the main goal is to communicate, not to exhaustively define. No doubt there are many arcane multiplexed architectures that would be possible, but if these weren't used or (for a generalist article like this) weren't used enough to become part of the contemporary single-board microcontroller landscape, then including every possible option and variant loses clarity and focus at a damaging cost to article readability (look at the IEP engineering project articles right now). Maybe there was an LSI-11 microcontroller. Maybe HP built something out of their in-house processors as used in the HP85. If there really was something in that space, then it would be worth including. If it's just a theoretical possibility for an architecture though, it shouldn't be. The point here is that first generation chips (counting from the Z80) were Von Neumann, later single-chip generations like the AVR used started to use Harvard (or at least, a Modified Harvard). Andy Dingley (talk) 09:30, 22 October 2011 (UTC)

Hmm. As usual with our parts list articles, Intel MCS-48 has no dates for introduction of the chip but clicking on one of the links attached to the article gives us a user manual copyright 1978 - which is bragging about the new parts being faster than before. Zilog Z80 says it was sold starting July 1976. One and half years isn't very much time between releases, we can wait longer than that for a new version of Windows. Are you *sure* that the 8048 was so much later on than the 8085 and its ilk? --Wtshymanski (talk) 21:36, 22 October 2011 (UTC)

The distinction is less between the processor families and more about the amount of onboard RAM available. It was only about 5 years between the 8080 (a better start point than the Z80) and the 8048, but the first single-chip controllers had so little RAM available (only a couple of hundred bytes) that their use was extremely limited. Many controller tasks, even quite simple ones, needed more RAM than could be fitted into a controller, so had to stay with the bus-based boards. Andy Dingley (talk) 21:45, 22 October 2011 (UTC)

There were several different uCs built around the J11 (11/23 on a chip) and F11 (11/44, IIRC) chips. And if you argue that the KIM-1 was a "microcontroller" (even though lacking peripherals) then you have to admit DEC's own LSI-11/03 module qualifies too.

My real point though is that the situation is more complex than the article suggests. It isn't really Harvard unless the machine flatly can't execute instructions out of data space; this is only loosely associated with whether there are separate buses for I and D. I think the current wording too much links the concepts of "Harvard architecture" and "separate buses". As an aside, a set of signals that just runs from one element (the processor core) to one and only one other element (the program store ROM or flash RAM) shouldn't really be called a "bus" at all. Jeh (talk) 22:52, 22 October 2011 (UTC)

PDP-8s had a long career as controllers (Anyone else remember the "Aha! The old computer in the rumble seat" mad scientist advert that ran for years?). It's hardly surprising that DEC should have made a more dedicated controller version with the LSI 11 chipsets. If you can add anything here, please do. As to the KIM-1, then doesn't a lack of peripherals indicate a bias towards the controller role rather than computer?

I don't see your issue over Harvard architectures. Which ones are you thinking of, where the buses are separated but they aren't a true Harvard? Even for the modified Harvard, it's more the case that program space can be written to with a specialised write instruction, so as to permit in-circuit programming. The memory spaces are still separate and program memory never really becomes available as a general data space.

A bus with just one passenger doesn't stop being a bus and become a taxi. If the bus can only reach one peripheral then there's maybe a claim that it isn't a multidrop bus, but a multidrop bus that just happens to only have one peripheral connected is still a bus, just one with little occupancy. Andy Dingley (talk) 23:28, 22 October 2011 (UTC)

Once again we've got the pompous gas-baggery that Misplaced Pages is so justly famous for. What is the *significance* of the compsci buzzwords in this context? It's just pointless obfuscation that does nothing to explain the topic. --Wtshymanski (talk) 03:42, 28 October 2011 (UTC)

Please do not attack other editors. If you continue, you may be blocked from editing Misplaced Pages. --Guy Macon (talk) 06:14, 28 October 2011 (UTC)

Potential Edit War

Recently, Wtshymanski removed content about early single-board microcontrollers (see page history) and replaced it with an unsourced claim that some early single-board microcontrollers had a Harvard architecture. As justification for the removal, he cited the Basic Stamp as a counterexample.

I reverted Wtshymanski's edit with the comment "You cannot use attributes of a modern microcontroller-based board as justification for removing content about early single-board devices." Basic stamps are not early single-board microcontrollers. Arduinos are not early single-board microcontrollers. The portion that Wtshymanski deleted clearly says: "When single-chip microcontrollers became available later on, the bus no longer needed to be exposed outside the package and so the Harvard architecture of separate program and data buses (both internal to the chip) became popular." You can't use examples using single-chip microcontrollers to dispute a statement that specifically excludes single-chip microcontrollers.

There were a couple on intermediate edits by Jeh, which I retained. Of course, the edits he made changing the wording of Wtshymanski's unsourced claim were lost, but I kept all of Jeh's other edits intact.

Then Jeh reverted back to the unsourced information with the edit comment "I removed nothing! I moved a heading. otoh I see no justification for having two paragraphs that say almost exactly the same thing)"

Because this appears to be on the verge of becoming an edit war, I restored it to the last known good version, and invited Jeh to discuss it here. My edit comment was "Reverted to version as edited by Andy Dingley at 22:10, 22 October 2011. Please discuss on talk page before re-adding unsourced claims about Harvard architecture."

Please read Misplaced Pages:Edit warring and especially WP:3RR before reverting again --Guy Macon (talk) 05:31, 23 October 2011 (UTC)

Well this is certainly puzzling and embarrassing. When I looked at this diff I completely missed the point of your changes to the first graf under "Internal bus". I'm not going to edit or revert that graf again... but I would suggest that the word "universally" is just begging for trouble.

Regarding moving the heading - I will say again that the paragraph you added immediately before the L2 heading "external bus expansion" is almost identical to the paragraph immediately follows that heading. I moved the heading originally because I thought a description of an expansion connector belonged under "external bus expansion", not "internal bus". I guess you thought I'd deleted the paragraph completely (rather than moving the heading up a graf) so you "restored" it? I really see no point in having two paragraphs, one right after another, that say nearly the same thing. Maybe the heading should be made a level 3 and renamed "bus expansion", and the graf in question appear only after that heading? Jeh (talk) 02:25, 24 October 2011 (UTC)

On the other hand - the first graf under "internal bus" says that in later uCs "the bus no longer needed to be exposed". Both versions of the next paragraph then talk about an expansion connector, which will be kinda tough if the bus never gets out of the chip in the first place. Jeh (talk) 02:28, 24 October 2011 (UTC)

A clearer distinction (for Wyshymaski's benefit) might be to say that the Harvard bus _wasn't used on the board_. Von Neumann buses were, Harvard buses were _inside_ the 8051 and others, but these controllers didn't have a Harvard bus that came outside of the package and ran around the board. I presume this is what Andy Dingley meant?

The DEC processors sound interesting if anyone can turn something up. There was also the HP 9915 that was a cheaper, rack-mount version of the HP 85 with no screen or keyboard. This was a cased machine though and not a single board. Given the markets they were both selling into, I'd guess that any DEC product would have been similar and so neither are really "single-board microcontrollers". 213.249.204.90 (talk) 13:15, 24 October 2011 (UTC)

Both of the above comments by Jeh and 213.249.204.90 make a lot of sense, and I would have no problems at all if the article were edited to reflect the above, just as long as Wyshymaski's removal of content isn't put back in with a revert to one of his versions. Guy Macon (talk) 16:17, 24 October 2011 (UTC)

I missed several days worth of churning on this article. You see the trouble that the irrelevant buzzwords have caused? No-one but a CompSci student cares about "Von Neumann" or "Harvard" architecture. We're conflating chips with boards, modern with ancient, and confusing the whole point. Yes, you could add a memory board to some single-board microcontrollers; how the processor thought about that memory was its private business and of little concern to the user, who ROMed all his code anyway and didn't much mind about separate address spaces for instructions or data. (Of course, once you had added a memory board to a single-board controller, you lost one of the key advantages, viz, its single-boarded-ness. But sometimes you couldn't sweat the code down to 1K, 2K, etc. ) --Wtshymanski (talk) 04:32, 27 October 2011 (UTC)

Please explain why, other than "I don't like it", you are once again removing content without consensus to do so. Specifically, you have removed the content that explained that all early single-board microcontroller were Von Neumann architecture, and you have removed content that explains why (Harvard needs more pins than were available on the 40-pin DIPS that the microprocessors used.) Guy Macon (talk) 08:00, 27 October 2011 (UTC)

If you don't understand what a "Von Neumann" architecture is, you could try reading the linked article about it rather than deleting it? Isn't this how an encyclopedia is supposed to work? It covers _the stuff _you don't know_ so you can go learn about it. This sounds like an interesting point anyway:- one of the first times that a rather theoretical aspect of the buses becomes obvious at the practical product level. 213.249.204.90 (talk) 10:00, 28 October 2011 (UTC)

Although Wtshymanski does at times remove content that he doesn't understand, he often removes content that he understands perfectly. I have never been able to figure out his motivations, but he has a "thing" about deleting content, from paragraphs to entire articles. He will throw out alleged reasons why he wants to delete specific content, but they do not appear to be his real reasons, as evidenced by the fact that whenever someone refutes his argument he simply makes up a new one, never ever changing his mind based upon discussion. You are right about it being an interesting point: earlier computers made of vacuum tubes and transistors or even 7400-series logic had no restriction on how many pins a buss could have, and latter ones had high pin count packages. It is interesting and useful information to know that in the era of 40-pin DIP microprocessors Von Neumann architecture was always chosen - because of the 40-pin DIPs. -Guy Macon (talk) 11:00, 28 October 2011 (UTC)

How many pins on an 8748? 40? And it's which CompSci buzzword, now? At least it's Turing-complete, whatever *that* means. --Wtshymanski (talk) 04:00, 30 October 2011 (UTC)

49 40 pins on a 8748. See . If you don't like "CompSci buzzwords" you might want to consider not reading Misplaced Pages pages about Computer Science. To learn what "Turing-complete" means, start here: Turing completeness. As for your repeated attempts to remove sourced content from Misplaced Pages and you continued refusal to work cooperatively with other editors, there is a strong consensus that you are on the wrong track. Rather than wasting all of our time undoing the damage you are doing, why don't you find another hobby? --Guy Macon (talk) 08:05, 30 October 2011 (UTC)

Did you look at figure 3 in that data sheet? How many pins do you see? Now I know you're playing with me. Harvard/Von Neumann has nothing to do with the number of pins on a DIP and certainly has nothing terribly significant to say about single-board microcontroller expansion busses. --Wtshymanski (talk) 18:32, 30 October 2011 (UTC)

Typo corrected. Harvard/Von Neumann has everything to do with the number of pins on a DIP. Just because you cannot understand an engineering concept, that's no reason to delete it. Harvard, with dual busses, requires twice as many pins as Von Neumann -- too many for a 40-pin DIP. This has been explained to you before, and is explained quite well in the material you tried to delete. Again I ask, why don't you find another hobbyhorse? Guy Macon (talk) 19:15, 30 October 2011 (UTC)

• Misplaced Pages articles that use British English