Interesting. It looks like the ADC stays on for 4 ticks, instead of 1.
I wonder if I can reduce that, or if that’s actually the shortest it can run and still work.
…
Looking through the code and the reference manuals, it looks like it could probably be reduced by temporarily using a different sleep mode during each measurement. I’ll have to try that sometime. I thought the ADC kept running in suspend mode, but it looks like it gets paused, and only runs during the active time for each sleep tick… and since the ADC needs like 1600 clock cycles for a measurement, and the sleep tick only runs for maybe 500 clock cycles, it takes 4 ticks. But I think it could instead use noise reduction mode to do the measurement all at once, which should (I think) reduce the time to just 1 or 2 ms, and reduce the total energy used from ~150 uA-seconds to ~4 uA-seconds.
So, for example, in aux low mode, avg power would drop from ~99 uA to ~81 uA. And it might simultaneously eliminate the spurious readings it occasionally gets. And it could probably measure more often without making a significant impact on total power used. Like, maybe every 2 seconds instead of every 8.
That is, if I understand it correctly, and if I did the math right.
Because the colors change with voltage, and right after the light gets shut off, the voltage is probably sagging from the load it had a moment before. So I want the color to “bounce back” to its actual voltage fairly quickly.
It currently waits 8 seconds between measurements, so people have to wait about 8 seconds to find out if their battery is actually low or if it was just sagging. But if it could check every second or two, it would give useful feedback much faster.
It shouldn’t be too often, of course… because the goal is to keep power use very, very low. But that’s a matter of diminishing returns. Like, if the change I have in mind works, it might use 0.5 uA on average, when it checks every 8 seconds… but the chip’s standby mode uses at least 30 uA anyway, so 30.5 uA is virtually indistinguishable. If I raised it to once per second, it would go up to 34 uA… but that’s still pretty negligible.
Or with the aux LEDs on low, it might go from 80 uA to 84 uA… and either one would be an improvement over the ~100 uA it uses now.
TL;DR: I think I might be able to check more often and reduce total power use.
Well then I am glad I can ask a bunch of dumb questions to help along innovation by facilitating conversation. I still hate mom mode and will be changing it to 5H. so its hard to accidently enter it. also is there a new chart that shows the new tact modes everyone is talking about? I still got a ton of stuff to go over. Like putting together the PCB for the h25lr for starters. I need to get that done by aug.
Nobody’s updated any of the charts yet but it’s fairly simple. 6H to enter/exit, 7H then enter a value for each slot for the 1st/2nd/3rd option, and 1-3H to use each slot.
Omg my wukkos light is so much better than the fw3a I just got btw. So much smoother(if thats what you call it), no weird flashes as its ramping, the button clicks are better( already designed a new PCB with a tact switch.) Like I notice weird flickers in candle mode compared to the wurkoos. Like its just not as smooth. I think with the tack button and the attiny1616 and cmy aux it will become a decent light.
WOO, tact switch. Also I already updated the design. for the vias
I am currently putting together tlv333 board with the at 1616 for the fwXXs . I know its like putting in an LS6 into a 66 fastback. or would it be a 70 bug.
What R and C size do you suggest? Also what pmos do you suggest for the polarity protection?
0603 is a good size, small enough for everything to fit on one side for such driver while still being relatively easy to manipulate.
The miliOhm sense resistor can be 0805 (0603 also works).
The RPP is just for the MCU, so any package will do, I like SOT-723 because it’s small, but a little bigger (SOT-523 ?) might be easier to manipulate.
The attiny1616 can do PWM at any speed from 2 to 65535 steps per cycle, adjustable on the fly. I use this to make the low modes lower and make the ramp generally smoother.
The FW3A used attiny85, which has only 8-bit PWM. So its ramp doesn’t go as low, and it tends to get stair-steppy at the bottom. This could be alleviated by implementing delta-sigma modulation, but that’s tricky to do so I haven’t. It requires very tight timing.
The blink while ramping is just a reference point. It’s mostly relevant on FET+1 lights, to let the user know when they have shifted “gears” from the 7135 chip to the DD FET. But on the FW3A it got carried over even though it’s not as relevant there. OTOH, today’s FW3As seem to be downgraded to FET+1, so… meh.
Nope… the one I got is full fet + 7 statis. I literally have had it for 4 days. Which I plan to redo that for the 1616 and rgb but proper board. As soon as I build out @thefreeman hdr board for the fwXX. then do one for that h25lr… its current driver is lacking. as its 5 levels only. BTW do you suggest a particular set up. I dont know if a 2 channel 3 melt7135 or what would be best for that type of set up. I am only driving one led for each of the 3 leds.
Also I figured the blinks might be for the ramping… but its like super jank looking. also I do get some weird flicker out of it at higher outputs. I am pretty sure the 1616 will fix the weird flicker. I could have a bad programing too
The HDR FET Q2 used to switch between the two sense resistors needs to have very low and consistant RdsON, since it’s included in HP-Rsense (R5+Q2).
I tested a bunch of them and HSCE2530 is a good alternative :
SISS80DN also works but it varies a bit more with Vgs (=Vbatt)
I wasnt sure what values to look for so thats good to know. I will take and incorporate your suggestions into my design. Though my choices were not to to far off from the mark. like the one I choose for q1 was going up to 1000a on the SOA were the ones you suggested where up to 100a… so it makes it more difficult to compare. I do wonder how it will behave at lower Amps.
All three MOSFET mentioned have similar power dissipation at DC load Tc=25°, 70W for the Nexperia (SOA diagram) and 80W for the Diodes. Of course 25° Tcase is unrealistic, our drivers are thermally regulated at 45° by default, but the FET Tc is going to be hotter, depending on the thermal resistance between the FET and Attiny, how much we don’t know, let’s say 100°, according to fig.1 ( PSMNR58-30YLHX) that’s 50% = 35W. assuming we have a maximum of 1Vds that’s 35A, much more than the 5A we want (or more), event if Tc=150° it’s still safe under 17.5%*70=12W.
So yeah good choice, but it’s only available in full reels that’s why I mentionned the other two, and they’re cheaper.
