AVR Freaks

I really do think the tiny's suit my needs perfectly... I am making a few "basic building blocks" to be driven from a master using i2c (in this case a PC, not even a microcontroller). One feature is a very flexible pwm implementation with 3-4 channels for light (led) dimming, another one is a set of 16 bit counters (also 3 or 4 channels) that don't miss a single pulse. Both of these cannot be implemented purely in hardware, whatever avr mcu I've examined. So if I'll have to implement things (partly) in software, what use will of this extended hardware features have to me?

So I think both of these can be perfectly implemented using attiny's.

In the future I may also add simple digital i/o's and adc's (again, over i2c) to the repertoire.

Now the crux is whether to use attiny's with very little output pins that only perform one of the features (very flexibel) or to use attiny's with more pins, performing some of the features combined. I favour the first option, but the two/three remaining i/o pins is really too little.

It's really a shame atmel goes through so much trouble making the functional differences between families and models so difficult to compare :-(

Yes that is exactly what I do now. We're some time further now, I've bought a programmer in the meantime, I've studied the datasheets of the '4s, '5s and 61's and found almost no functional differences (apart from the obvious pin count). I other words, as long as I keep using these, I can port code very easily from one to another, it's no issue at all. I don't like the idea of #defines from device-specific port's to "generic" port names, by the way, and I don't mind adapting the code a bit from one device to another.

The important issue was: what programmer to buy (not all of them can program all mcu versions...) and what actual devices to buy. So I bought I programmer (dragon) that can program the '4, '5 and '61 and bought a few of all of these, problem solved :-)

Besides the '61's ability to switch USI to another set of pins and the '61's having a few minor extra timer features, all of these are more or less the same, and so it doesn't matter that much, I found out.

8k flash and 8k ram is really enough for me, at the moment, so I'll stick to 84,85 and 861's for now. They're only a few cents more expensive than the smaller ones.

Well actually Atmel should supply this information of course. But they prefer marketing-fog instead :-(

Urgh, legacy technology :-(

(rant mode on)
The difference is that if you only use numbers that fit into 8 bits (which is very often the case in these kinds of projects), you will see no difference compared to 16, 32 or 64 bits. They may even run slower due to their bigger int sizes when stored into 8 bit memory and use more memory. I think 8 bits is fine.

On the other hand, rs232c is something once very useful and state of the art, but today it's obsolete. Nobody is using it according the specifications (right from the start, actually), which makes it more difficult to use between one or two unknown (black box) devices. I've spent too many hours making sets of two devices using rs232c try to communicate with each other and having no clue where it went wrong. Also, may be even more important, it needs a very stable and accurate clock on both sides, even though it is called "asynchronous". And even though it doesn't reach real high speeds, it's use limited to a few meters. Last but not least, there is no way to know if it kind-of-works without having very specific communication with the endpoint (compare this to usb, where each device identifies itself immediately and voluntarily).

I'd say: the worst of all words (of all available bus systems).

What doesn't help as well, is that many electronics engineers tend to use rs232c interfaces on computers for bit banging, which they're not suitable for. You cannot get the timing right, as pc's nowadays tend to have real operating systems that feature hardware abstraction (which I think is a very good thing). Most of the time the operating system's device drivers don't even allow direct bit banging.

I think it's time to jump on the train before it has gathered to much speed (i.o.w. no single piece of computer has a "usable" rs232c connection anymore).

One may of course disagree, but that's how I see it.
(rant mode off)

Ah better.