No luck on this 7135 thing yet. I hoped it was a software issue, but it doesn't seem to be. I went back to the old test rig with no back-side 7135's using latest firmware, and also tried an older version firmware with the new driver. The test board works fine, with the all7135s channel on, with no actual 7135s on it. Just to see, instead of shutting down all channels right before sleeping I turned on turbo instead. It really shouldn't matter, and it didn't.
About the best idea I have is get the driver to self trigger a shutdown. by sending input power through a second channel FET. Then it's possible to time the shutdown detection. Mostly that would just prove what already seems true though, but still wouldn't explain it. This is really puzzling.
a crud.. never mind. I thought I had progress on this 7135 thing, but I think I fooled myself. Anyway, I'll keep playing with it.
Anyway, I found diagnostic tool. I can sneak in a Vcc ADC read in the interrupt right between detecting the shutdown and powering off. It's working, although no time to interpret it now. It should give insight into how much time and charge is lost before the shutdown. I could make use a couple actually, like one after the first wake, and get a read on how much drain is happening during sleep that way too.
So some interesting results. Unless I'm getting completely fooled, like the interrupt is firing at all the wrong times (TURBO works so it can't be all broken) I think I'm measuring the voltage on C2 immdiately after the pin fall is detected.
I first had to calibrate my Vcc read real well (which uncovered a slight bug in the calibration function, but almost irrelevant really). Anyway
In turbo I'm getting:
Input Voltage Vcc Catch voltage
4.0 3.06
3.06 2.71
On All7135s I'm getting almost the same
Input Voltage Vcc Catch voltage
4.0 3.00
3.06 2.60
I find it interesting that the 4.0V reading does relatively badly compared to the 3 V readings in both cases. But what's more interesting is that there doesn't seem to be any serious problem here with the 7135s, and yet there is. This helps though, because if this holds up it at least points to what's not the problem, power down detection time, unless that 0.1V is really a big deal, I wouldn't think so.
This seems to be a good tool. I'll continue, trying to measure voltage at the first power-on and see what that looks like.
(just updated numbers to uncorrect for the diode, so these are real voltages on the Vcc pin now.)
Yes. I got it. Just had to find the right pin state to get the driver pins into after shutdown detection. Apparently output low doesn't work (or output PWM 0, should be the same thing). I have no idea why not. Input high is documented as a good option because it's high impedance and avoids supposedly problematic middle voltages (input low actually means input floating). And yet, the solutions was input low. It maybe helped that I shutdown the 7135's first pulling them down with a quick output low first, I don't know.
Anyway, it not only works, it works great. You can whistle a couple of line of Dixie before you turn you light back on. I've got the code in a bit of a mess at the moment from all kinds of testing, so I'll probably start over from the last release and add things and test, might take a couple of days, might not, but I think we'll have OTSM working great in bistro here shortly.
I keep recompiling and reflashing this same ROM and keeping checking that it's really in ALL7135s because I keep thinking I must be confused, but it's working.
Updated: It actually works over 4s at 3.0V without a bleeder (~4K of divider resistance) and 160C heat gun pointed at it!
I have no idea what you are talking about Flintrock but your writing tells me you are excited and that can only mean one thing. Your onto a winner. well done. :+1:
OTSM. (I think TA named it that). I think it means off-time sleep mode or something like that. It's the replacement for OTC in bistro HD. It means that the length of a medium click is as steady as the watchdog timer, which as far as I can tell is about 10% even when hot, which is much better than many caps. It's also not sensitive to all the resistor tuning that the OTC is, so should eliminate those headaches.
I've got the first low-capacitance bistro Texas Avenger OTSM light ever right here in my hand, working just fine. I thought the issues were worked out several weeks ago, but it hit a snag when applied to a real light. The snag is now defeated.
These fixes will of course get posted up to the bistro-HD thread shortly.
12v jobbers isn't a term I can google to fix my bricked chip /board (I tried to do a flash and I think I bricked my chip) What do I need to fix my driver board? It's a 17DD MTN Fet + 7135.
However, if it’s one driver you’re talking about it’s cheaper and less frustrating to just get a new one. I used the reset pin quite a lot on the ATtiny85 so I needed this in order to be able to flash after activating reset pin as IO.
Have to reanimate this thread, I guess.
Did anybody take the dependency between voltage and temperature readings on Attiny25/85 into account?
For my 2018 BLF Contest light I developed bluetooth remote control with telemetry. Voltage and temperature readings are sent to the app once per second. I noticed a significant dependency between voltage and temperature. Here a couple of readings for roughly the same temperature:
at 4.6 V temperature reading is 13 °C
at 3.6 V temperature reading is 27 °C
at 2.7 V temperature reading is 33 °C
As you can see the dependency is not linear. When I find some time I’ll do more research.
For now I’d say this dependency might interfere with temperature regulation.
I didn’t care much about it. On the other hand I don’t recall getting such wild variations, but since I haven’t touched the 85 in ages my brain has purged most of that stuff. I’m sitting with my ATtiny1634 drivers right now so I just tested: 20°C at 2.6V and 22°C at 4.4V, nothing for me to worry about.
No I don’t. Have you allowed a settling time when switching between voltage and temperature readings? At least on the 1634 it’s important if you are reading voltage by connecting internal voltage reference to the ADC and using MCU voltage as reference. I read voltage on 1S lights this way and if I don’t have a delay I’d get funky results. 1634 datasheet specifies 1ms settling time, I have a 2ms delay just to be sure. I’m not in a hurry.
This will probably not help if the results you posted are always consistent but at the moment I can’t think of anything else.
After some more research - including the noise reduction sleep mode - I checked the temperature readings again in my battery driven final light setup and the readings turned out normal (without noise reduction), without much deviation at different voltage levels. But the wrong readings are still there in my test setup on breadboard. So the reason must be somewhere in this test setup, don’t know yet if its the test setup itself or if its the USBASP connected to the breadboard which is causing the noise. And it must be noise causing the weird readings since they are normal even in my test setup when I use the noise reduction sleep mode.
Actually I do this delay only when measuring voltage internally since Atmel says in the specs it is required when measuring Vbg against Vcc. As I understand the specs this delay is not required for measuring temperature against Vbg. (I’m refering to Attiny85 specs).
Glad it’s sorted. You’re right about the delay, according to the datasheet it’s only required when switch to internal voltage reference, it says nothing about from. For me it doesn’t matter though, I have 2ms to spend after each conversion regardless if I’m reading voltage or temperature.
It shouldn’t be too hard, assuming there’s space in the ROM. Basically just take code from the candle mode’s EV_tick handler, put it in a loop, and add a 16ms delay between frames.
This, um, also assumes some other things though… like the existence of a function to set the output level in ~150 visually-linear steps, without the caller having to worry about how that’s actually implemented in the hardware.
Hey you atmel wizards. I want to reduce the PWM frequency of the 7135 channel of a D4 driver.
From doing a bit of reading in this thread I gather that doing this might involve using a divider with the clock frequency. On the engbedded page if I check the divider of 8 checkbox it changes the low fuse to 0x62. Would that work? Would that change the frequency for both channels, and if so is there a way to just change the 7135 channel?
If I understand correctly that would lower the frequency to around 2 kHz, which would work for my application. But is there a way to adjust the frequency more finely, rather than by a factor of 8?
