Open9x @ ersky9x crash

[Peter]

I had a crash yesterday with my EasyGlider Pro. It landed in a corner field! Been searching for an hour before I found it. Found it by looking at the RSSI! It was almost unharmed luckily.

What happend. I lost control at 100 meters, while it relatively slowly circled down. I looked at my transmitter and saw it was in USB mode! I quickly rebooted, but within a couple of seconds it went in USB mode again. After a couple of quick tries I had control for a couple of seconds but it was to late, it went out of sight behind a dike.

Been running revision 1195. Cpu temp was 70 degrees celcius. Outdoor temp 31 degrees. Radio was tested for hours before. At home I couldn't reproduce the problem within half an hour. In the radio is a frsky diy dht, haptic, bluetooth and a USB extension cable to the charge port made from a spark fun breakout board.

Does anyone know what could have caused it?

[bertrand35]

USB detected when in flight? Would it be possible there was a short somewhere? I could add some code to avoid USB from being detected when let's say 10seconds after the Tx is started.

Bertrand.

[Peter]

The USB was indeed detected inflight. I have to open transmitter to see, but I don't think a short caused it. All shrink tubed and hot glued. Could a little sweat on the connector caused it? But I don't even touch the connector itself, it is to deep.

How hot may the Cpu be?

The radio can be flashed with radio off, is it? Then maybe remove it when powered on.

[hendrix]

Well 70 degrees C is very hot and abnormal for the cpu i think.
What battery voltage do you use, i use a two cell lipo and at 39 degrees ambient temperature the cpu temperature rarely goes above 45!
Let us know what you find.

[Peter]

I use a 3 cell life. 9.8v yesterday. But it was very hot yesterday and the black case was in the sun for 2 hours.

[Rob Thomson]

Yesterday I had mine out - 30 degrees celcius - 2 hour cross country event. No issues at all!

I use a 2 cell lipo. Maybe the difference is in the voltage? Unless of course.. did you have a 'beta board' that did not have the heat sink on it?

Rob

[Peter]

I do have the one with a heatsink. And it seems like the standard operating voltage is up to 85C.
And could heat cause a USB cable detection?
Using 170mA by the way.

[Peter]

Here the photos of the usb connections. I think this rules out a shortage.

[MikeB]

I'm looking at the USB voltage detection circuit. The 5 volt power from the USB does 2 things:
1, Via a diode it powers the CPU
2. Via a potential divider it provides a 3.3V logic signal to indicate USB is connected.

It may be possible for the diode reverse current to cause the logic signal to become high enough to be detected as a logic 1. The diode reverse leakage current increases with increasing temperature, and is also higher when the battery voltage is higher.

Both ersky9x and open9x should do something to guard against this possibility.

a) Use Bertrand's idea, no USB after 10 seconds.
b) No USB if supply voltage is above 6 volts
c) Check for USB only at start up, i.e. only enter USB mode if power up is caused by plugging the USB in (USB power detected.

I think I favour (c), as this will also help with stopping interrupts used for normal operation, we can actually go straight to USB before we enable anything. I had been thinking about doing that anyway. We might include (b) as well just to cover the (unlikely) posibility of a transmitter re-boot while running.

Mike.

[bertrand35]

I am also in favor of (c). And perhaps if USB is connected after start up, it will be massstorage (as soon as it works).
Bertrand.

[MikeB]

I've just committed (c) and (a) on ersky9x firnmware as a precaution. (b) will take a bit more time. As a general recommendation for ALL transmitters, try ro keep them in the shade to keep them cooler, better for both the electronics and the batteries.

Mike.

[Peter]

Is a temperature alarm a good idea?
At what temperature should it alarm?

[bertrand35]

Yes it's a good idea. Could be done this way
CS1 v>OFS CPU 90°
CS1 => PlayTrack "bbq"
Bertrand.

PS: you would open an issue for that so that I think to it?

[Rob Thomson]

I would think this should be a system option, and not a custom switch? Mist people will not bother to code it as a switch ; but it really is an important alarm - just like the low battery alert.

Sent from my Nexus 7 using Tapatalk 2

[Romolo]

I would think this should be a system option, and not a custom switch? Mist people will not bother to code it as a switch ; but it really is an important alarm - just like the low battery alert.

Sent from my Nexus 7 using Tapatalk 2

+1
and, I will add,this alarm should not use the voice, which increase cpu load and current usage, but the stock beeper.

[Clivew]

Hang on, not so fast
How is the temperature going to be monitored?
I have 2 Ersky9x boards.
One shows me 2 deg c, the other 47 deg c when first switched on in the same room!
(Cpu temp) I suspect it's probably about 25 deg in here now
Unless they can be calibrated, any temperature alarm will be meaningless.
Clive

[Romolo]

Well I agree with Clivew, I was rather astonished by my radio temperature, 67°, but here is hot, so I considered it almost normal, but today I did a test: I put the radio with the case open on the table in my server room (18°) with cold air flowing on it.
The reported temperature is still 54°, with a infrared thermometer I have measured the MCU external temperature: 21°

For the records: I have one of the first boards, modified for 5V on audio amplifier, and a small thermal radiator on 3.3V stabilizer.

[MikeB]

From the data sheet:

Code: Select all

The VT voltage equals 0.8V at 27°C, with a ±15% accuracy. The VT slope versus temperature
dVT/dT = 2.65 mV/°C only shows a ±5% slight variation over process, mismatch and supply
voltage.
The user needs to calibrate it (offset calibration) at ambient temperature in order to get rid of the
VT spread at ambient temperature (+/-15%).

15% of 0.8V is 120mV. Divide this by 2.65mV and you get a possible +/-45 degree error. We either need to calibrate this out, or make use of the temperature sensor in the Tiny167 co-processor. This sensor, when calibrated using the manufacturer data, is +/- 10 degrees C.

Mike.

[Romolo]

Thanks Mike, this really explains a lot.
Btw we can use the Tiny167 to roughly check whenever the ARM sensor is calibrated at start time. but not to calibrate it as +/- 10° is still a lot. But at least we do know the sensor is not calibrated at all.

[MikeB]

I have the following figures:
Thermal resistance of LQFP100 package 46 deg C/Watt
Current consumption of my Tx (no module, backlight off) 80mA - assume this is all CPU current.
Power dissipation of CPU, 3.3*0.08 = .264 Watt.
Temperature rise of CPU junction to case 12.1 deg C (3.3*.264)
Datasheet expected current consumption at 36 MHz is about 40mA with all peripherals running - temp rise would be 6 deg C.

At power on, after the Tx has been off for some time, the CPU core temp will be at ambient, and will increase slowly.
Since the 2.65mV/degC is +/-5%, then when the CPU is 60 deg higher, the sensor might only read 54 deg C higher.

So if the offsetis calibrated, I would suggest a temperature reading limit for an alarm of no more than 75 deg C (chip specified to 85 deg C, this is usually ambient).

To calibrate we could ask for ambient temp, then store a calibration value that gives a reading of 6 deg higher.

Mike.

[Romolo]

And what about adding a lm35 and read it using a free adc port ?

About your figures: 7mA probably will go in the pots but very little changes.

[MikeB]

Yes, and I have a bluetooth module in there, then the Tiny167, EEPROM and MAX232 will all be taking current.

Mike.

[bertrand35]

I just commited a fix: the bootloader is launched only when USB is detected AND Tx is off.
Mike / Brent do you think that it's enough as a protection, or the temperature could make both conditions true?
Bertrand.

[SkyNorth]

I like the 10 sec timeout after power - on , then a logic 1 on the pin would not mater.
Can we log this to a ERROR flag? or warning . so we can see if it is happening a lot.
-Brent

[bertrand35]

I think that if we test the Tx main switch, it's ok as well, unless somebody reports the issue that the Tx shutdowned by itself with the swith ON!
Once the massstorage support will be implemented, I think we will show something on screen to indicate that the service is started (and in this scenario, it should not interrupt the mixer / pulses).
Bertrand.

[jhsa]

I think that if we test the Tx main switch, it's ok as well, unless somebody reports the issue that the Tx shutdowned by itself with the swith ON!

Scary..

[MikeB]

With the Tx power switch ON the Tx power does not rely on the soft power switch, it should simply be ON.

Mike.

[bertrand35]

Indeed! Then it's safe
Bertrand.