Attiny25/45/85 FW Development Thread

As it is right now, are the frequencies on all PWM channels the same? I remember TK mentioning somewhere, that running the 1xAMC7135 channel on PB3 maybe isn’t that great of an idea because the higher frequency maybe wouldn’t go that well with realizing nice moon modes.

Haven't checked it in that much detail. I was think'n of PWM'ing an indicator LED used as a 4th channel but didn't do the research, just recall seeing it discussed, mentioned, and think the Atmel specs mention PWM can be applied on 4 channels.

I don’t have a clue about most of this stuff but I did a bit of digging on the subject a while back. 4 pmw channels is definately doable, for sure on the 85, but it was mentioned that the reset pin would need to be forfeited with no reflash ability. I also remember something g about needing the 85 dev board to accomplish it. (At least this is what my notes say, I don’t really remember any of it…) :slight_smile:

Ohh - if you have the Atmel Dragon (a developer's kit), you can utilize the ability to have the reset pin as an I/O pin and not lose the ability to re-program the 85.

I'm pretty sure there are two clocks being used in bistro, and I think two are all that's available, but the clocks are just counters and an "output compare register" sets the duty cycle by matching a particular clock value during the cycle. I believe two OCR's are sharing one clock. What I didn't know is if the second clock can be used with two OCR's. From Tom's answer, it sounds like it can.

Yeah, I think there are ways to squeeze in 4 pins, it just depends on how creative the software is and/or what other features get compromised. two switches and voltage CAN be done on one pin. low (maybe < 0.5V) = tail switch power off; mid (from 0.9 to 1.8 maybe) = voltage sense range; high (Vcc) = eswitch engaged.. wired to short voltage pin to Vcc (R2 still means it's not a short to ground). I guess I'm kind of imagining a light where a tail switch cuts power and the eswitch doesn't. I don't know what really exists. But even just having two eswitches, the story should be the same.

That can be done if the ADC is used to detect OTSM power-off. It's just from a bistro perspective, not the way things are being done now. But if the eswitch instead shorts to ground and behaves the same as the tail switch, then that already would work the way things are now.

Hmm another creative way might be to have the eswitch still pull low but not as low.

So < 0.3V = main swtich. 0.5 to 0.9 = eswitch, 0.9 to 1.8V off limits due to pin change tolerance, > 1.8V = voltage sense. That would work with the present bistro software. The eswitch would wire from pin 7 to ground with an appropriate series resistor to pull it part way down. You still detect the switch press with a pin change, but then in the wake loop check the value with the ADC to see which switch hit. It avoids needing to implement a general ADC interrupt setup. It could work. Tail led's complicate things because they shift all the voltages a little, but it could probably all be made to work.

I kind of like it actually. Can we have resistance inline with the switch wire or it needs to be on the board?

I'm not seeing 4 PWM's. There are four output compare registers but looking at the pin diagram on page two of the data sheet I don't see how they can be applied to 4 separate pins. PB3 and PB4 only get OC1B and not OC1B and PB2 gets nothing. OCOA and OCOB are pb 0 and 1, but OC1A (and not OC1A) is also only on pb 0 and 1. So I don't see how to make use of OC1A on a new pin. Maybe I'm missing something.

Maybe you can hack this by placing an interrupt on OC1A and "manually" flipping the output of pb3 from an interrupt handler? Actually this does sound possible. I'm already running a secondary interrupt off one of the PWM clocks for idle timing, not sure why you couldn't rig something like that.

Regarding multiple channels:
There is an alternative to PWM if you need more channels: BAM (Bit Angle Modulation), 8 channels are easily doable (if you have the pins). I did some tests last year and it worked out quite well. It’s a bit tricky to program and I wouldn’t choose this method if I needed smooth ramping with 7135 at low levels.

Regarding the reset pin:
if you use this pin you don’t need expensive programmers for further reprogramming. I built a cheap fuse resetter with this instructions:
http://www.rickety.us/2010/03/arduino-avr-high-voltage-serial-programmer/
I used a arduino nano and a cheap MT3608 boost converter for the 12 V ouput.

I'm not sure what you guys are say'n. Certainly we do 3 PWM outputs now, but you see only 2 are possible?

PB4 (pin #3) is PWM'ing now for the FET in BistroTriple and narsilTriple.

I'm not saying only two are possible. However, only 3 are possible using the hardware waveform generators. There are 4 output compare registers but two of them have waveform generators that attach to the same pins.

Anyway, yes, I've figured out how to get 4 in principle. You need to leave the 4th compare register detached from the pin, and enable its(the register not the pin) interrupt, and use the interrupt to bit-bang the OTC pin. In phase mode you need to keep track of ups and downs. (why are we using phase-correct mode though?)

It's not too bad really. The datasheet recommends against it because it will keep the CPU busy. On the other hand, what else is this CPU busy doing hunh? It's not like we're going to connect these lights to SETI are we? I shouldn't ask.

Sigh.. the tripple version was clearly whipped up a littel quickly to work and was no longer compatible with single and double versions anyway, so I've started cleaning that all up and adding capability for a fourth channel too. It does mean ramps don't have nice names anymore but are just RAMP_PWM1 ... etc, but calling it 7135s just inhibits keeping the code flexible.

I've also simplified the defines to maybe make configuring creative driver setups a bit easier. I don't promise TK would approve of all the changes. Maybe someday I'll even finish something.

Oh someone asked about frequency. As far as I can tell bistro tripple is running at CPU frequency. The PWM ramps up to 255, but the 7135 and 7135s are in phase mode so it ramps up and down, so that's a frequency of 8Mhz/500 so 16khz. The FET doesn't have a phase correct option and appears to me to be running at 32 khz! Unless I'm missing something.

This is one of the examples I found that created four pwm outputs on the 85.

http://www.technoblogy.com/show?LE0

This stuff gets a little silly now:

#ifdef PWM4_LVL
inline void set_output(uint8_t pwm1, uint8_t pwm2, uint8_t pwm3, uint8_t pwm4) {
#elif defined(PWM3_LVL)
inline void set_output(uint8_t pwm1, uint8_t pwm2, uint8_t pwm3) {
#elif defined(PWM2_LVL)
inline void set_output(uint8_t pwm1, uint8_t pwm2) {
#else
inline void set_output(uint8_t pwm1,) {
#endif


but it actually does save a few bytes over using globals. It kind of just went from one to two, and then someone hacked a third on, and now it's kind of silly.

Exactly the way I've headed with it. Interrupt for the 4th, like I said. Apparently I can figure that out, but not how to end a quote after a link.

Ok, it's now configurable to use any combination of 4 outputs, trying to get the hardware defines more simplified and flexible too.

Oh ha, and this was another example of software PMW on four channel.

Seems you have surpassed this info already though. Great work!

You had one too many end-quotes. :wink:

This all sounds great Flintrock! I gotta get a looksy when you post it all. Quick read of the article as well...

Well, distracted with this

https://budgetlightforum.com/t/-/44174

And by the way, things are bit broken now. I said I implemented it, but I've done a bunch of "clean" up and reorganizing and who knows how long it will take me to recover from that. I haven't even tried to run it yet. I already know of one thing I broke (a feature that I added though), but I'm not going backwards. I'll twiddle at getting it all in shape as I get time.

thats way over my head lol

:FACEPALM: It ended up being just a dumb mistake. After all that I deleted it. I just didn't have optimization on at the time, doh.

Some days are like that programming. You know like when waste a bunch of time just to figure out the program you're running is in a different directory than the one you're compiling, which explains why the last 10 ways you tried to fix it didn't work. I'm pretty sure anyone who's programmed much has done that one before... or maybe it's just me.