Reading internal temperature sensor

When reading the ADC channel 16 (temperature sensor) I get constant readings like 1880 even I change temperature of the package by say 15degrees.
The sensor is enabled (from Read Vcc post):
void setup_vcc_sensor() {
adc_reg_map *regs = ADC1->regs;
regs->CR2 |= ADC_CR2_TSEREFE; // enable VREFINT and temp sensor
regs->SMPR1 = ADC_SMPR1_SMP17; // sample rate for VREFINT ADC channel
}

You might take a look at http://forums.leaflabs.com/forums.leafl … ml?id=1120

Andy thanks!
Now it works with
// Reading Vdd and Temperature Sensor
// Pito 8/2016
// Temperature sensor at ADC16, VREFINT at ADC17
// BluePill and Maple Mini

void setup_vdd_tempr_sensor() {
adc_reg_map *regs = ADC1->regs;
regs->CR2 |= ADC_CR2_TSEREFE; // enable VREFINT and Temperature sensor
// sample rate for VREFINT ADC channel and for Temperature sensor
regs->SMPR1 |= (0b111 << 18); // sample rate temperature
regs->SMPR1 |= (0b111 << 21); // sample rate vrefint
adc_calibrate(ADC1);
}

void setup(){
setup_vdd_tempr_sensor();
Serial.begin(115200);
delay(1);
}

void loop() {
float tempr, vdd;
// reading Vdd by utilising the internal 1.20V VREF
vdd = 1.20 * 4096.0 / adc_read(ADC1, 17);
// following 1.43 and 0.0043 parameters come from F103 datasheet - ch. 5.9.13
// and need to be calibrated for every chip (large fab parameters variance)
tempr = (1.43 - (vdd / 4096.0 * adc_read(ADC1, 16))) / 0.0043 + 25.0;
Serial.print("Vdd= ");
Serial.print(vdd);
Serial.println(" V");
Serial.print("Temp= ");
Serial.print(tempr);
Serial.println(" C");
delay(500);
}

@pito and Andy

Perhaps we should add this function somewhere inside libmaple ??

(any suggestions welcome

The refman claims the spread of the two params in the tempr calc function is large, so the sensor has to be calibrated for each chip separately. Moreover, the temperature measured depends on the CPU utilization (it measures temperature of the mcu’s silicon die), so it is not intended for measuring an ambient temperature..

Btw – do we Calibrate the ADCs somewhere in the ADC init code? (there is an ADC self-calibration procedure for STM3210x).
A: there is adc_calibrate(ADC1); available..

RogerClark wrote:@pito and Andy

Perhaps we should add this function somewhere inside libmaple ??

(any suggestions welcome

FYI – my ambient is 24C, after power on it shows 30C, after 5min 35-36C while running the above sketch.
There is only 2.3bit/degC, what is inside the noise background on the BluePill, so the measurement is really rather indicative.

Pito wrote:FYI – my ambient is 24C, after power on it shows 30C, after 5min 35-36C while running the above sketch.
There is only 2.3bit/degC, what is inside the noise background on the BluePill, so the measurement is really rather indicative.

I placed the BPill into my fridge (-16C inside). Let it freeze there for 20minutes. 1.5m long usb cable (50cm inside the fridge).
Then measured (BP still in the fridge):
10seconds after power on = -13C
5min after power on = -5C
The readings are noisy, jump +/- 1.5C, so the numbers are rather average.
Interestingly, the Vdd measurement is rock stable, much stable than under ambient temp.

And finally I dipped the BP into boiling water. It stopped sending data almost immediately (last data 40C). Immersing only the crystal part/side of the pcb into the boiling water stops the data, so it seems to me this part is critical. Or the water from my water tap is too conductive
I did with hair dryer and I got 60C reading (the max of my dryer) while BP worked. I can try harder with my hot air soldering gun but I do not want to remove the chip off the BP’s pcb

So subtracting 10C from my readings may indicate the ambient temperature with say +/- 2degC accuracy. That is valid for my chip only, of course..

LOL moment when you said you put the board in boiling water and it stopped working

I have an infrared thermometer and I measured the temperature of the GD32 at different clock rates ( I think i probably posted the data to the forum, but it would be ages ago)

I also have a thermocouple thermometer ( but I don’t trust it that much), but I could clip the thermocouple to the top of the STM32 and take a reading if you want.

Long time back I had almost unlimited access to liquid nitrogen and deionized water. Plunging electronics into LN was a clear NoGo (semiconductors stop working at that temperature), boiling distilled (and best deionized) water cannot harm much with modern devices when properly designed.

Pito wrote:Long time back I had almost unlimited access to liquid nitrogen and deionized water. Plunging electronics into LN was a clear NoGo (semiconductors stop working at that temperature), boiling distilled (and best deionized) water cannot harm much with modern devices when properly designed.

In 80ties I had access to vacuum chamber(s), we used tons of LN in order to create ultra high vacuum. BPill in vacuum may overheat when running demanding sketches – as there is only cooling via radiation available.. In vacuum (outer space) the measuring the internal chip temperature would be recommended..
There is an option to spend winter holidays in Oymyakon, Russia, and open your window for a while when calibrating your BPill

Pito wrote:In 80ties I had access to vacuum chamber(s), we used tons of LN in order to create ultra high vacuum. BPill in vacuum may overheat when running demanding sketches – as there is only cooling via radiation available.. In vacuum (outer space) the measuring the internal chip temperature would be recommended..

You are right, closing the window during the measurement would be a good idea. Otherwise your LCD screen stops working, sure..

LOL. You need a very long USB cable to test in outer space

In all seriousness… I suspect there is probably some way to test in a vacuum using fairly readily available items.

Perhaps using a vacuum pump used in commissioning air conditioning systems.
( probably not as good as the vacuum in space, but probably good enough for testing )

But of course, irradiating with heat from the sun while testing in a vacuum would be challenging.

In all seriousness… I suspect there is probably some way to test in a vacuum using fairly readily available items.
For example try your hoover
“Vacuum” is a common term mostly used for describing “lower than atmospheric pressure”. For example 10000 Pascal.
Vacuum in space (say interstellar space) is something like 0.00000000000000001 Pascal.

Thanks Pito

I just wondered how good a vacuum you get from a $50 pump used to install air conditioning.

Probably not that good, but I wonder what vacuum you would need to achieve to give a good simulation of orbital space etc.

Unfortunately, I can’t find the specification of the vacuum used when commissioning air conditioning systems.

On the flip side I wonder how accurate it would be as a temperature sensor where the processor only wakes for a second or so to read the temperature transmit and shutdown. Like the lux sensor super low power, Super cheap.

simonf wrote:On the flip side I wonder how accurate it would be as a temperature sensor where the processor only wakes for a second or so to read the temperature transmit and shutdown. Like the lux sensor super low power, Super cheap.

Pito wrote:Long time back I had almost unlimited access to liquid nitrogen and deionized water. Plunging electronics into LN was a clear NoGo (semiconductors stop working at that temperature), boiling distilled (and best deionized) water cannot harm much with modern devices when properly designed.

RogerClark wrote:
Good point.

However the 32khz crystal side of the device is still running, so some heat will be generated.. Just , a small fraction as much as when its running at full speed.

I thought I’d have a quick play with the internal temperature sensor, so I grabbed the example from

../hardware/STM32F1/2016.11.18/libraries/A_STM32_Examples/examples/General/InternalTempSensor/InternalTempSensor.ino

.. but it fails to compile for the generic stm2f103 board with the following error

/home/ahull/.arduino15/packages/stm32duino/hardware/STM32F1/2016.11.18/libraries/A_STM32_Examples/examples/General/InternalTempSensor/InternalTempSensor.ino: In function 'void setup_temperature_sensor()':
InternalTempSensor:22: error: 'ADC_CR2_TSEREFE' was not declared in this scope
regs->CR2 |= ADC_CR2_TSEREFE;
^
exit status 1
'ADC_CR2_TSEREFE' was not declared in this scope

Andy

I recall there was a PR / Fix for incorrect definitions in that area. That ADC_CR2_TSEREFE may have been one of them

OK If I change to

regs->CR2 |= ADC_CR2_TSVREFE;


Andy

The Git PR was

https://github.com/rogerclarkmelbourne/ … /132/files

@stevstrong proposed it, so you could PM him, as he is active at the moment

RogerClark wrote:Andy

The Git PR was

https://github.com/rogerclarkmelbourne/ … /132/files

@stevstrong proposed it, so you could PM him, as he is active at the moment

Hi Andy

OK.

I’ll try to update it later.

Unfortunately my local copy of the repo has a load of other changes for various things I was hacking about with, so I’ll need to get rid of those (or save them etc) before I can update the example

Andy

I’ve edited the sketch file online in github, and committed the change. Can you download it again and test for me as my local repo isnt working at the moment