LeafLabs Garden

A place to share, learn, and grow...

1. 

   tdc218
   Member

   So, just for grins, I wanted to learn how to read the Maple's internal temperature sensor. I came up with the following code:

   #include <stdio.h>
   #include <adc.h>
   #define TEMP_CHANNEL 16
   void setup(void)
   {
   }
   void loop(void)
   {
     uint16 temp;
     char buf[64];
     temp = adc_read(ADC1, TEMP_CHANNEL);
     sprintf(buf,"%04x", temp);
     SerialUSB.println(buf);
     delay(1000);
   }

   Now, according to RM0008 (section 11.10), the temp sensor is on ADC1 channel 16. When I run the above code, I get an output of 4 (occasionally 3, but mostly 4). If I make TEMP_CHANNEL 17, then I get numbers in the high 0x600/low 0x700 range that act like temperatures. When I put my finger on the CPU (warming it up) the numbers go down, and when I set a frozen bolt on it they go up, just like one would expect. Looking at the code for adc_read, I can't account for the off by one channel number. Anyone have an explanation for it?

   Posted 4 years ago #
2. 

   gbulmer
   Moderator

   tdc218 - (I took the liberty of replacing [code] and [/code] with back quotes to preserve your code layout)

   I have not checked the initialisation of Maple, but the temperature sensor must be switched on, so it may be channel 16, but it isn't switched on.

   Further, the process described in RM0008 (section 11.10 "Temperature sensor") says:
   Reading the temperature
   To use the sensor:
   1. Select the ADCx_IN16 input channel.
   2. Select a sample time of 17.1 μs
   3. Set the TSVREFE bit in the ADC control register 2 (ADC_CR2) to wake up the temperature sensor from power down mode.
   4. Start the ADC conversion by setting the ADON bit (or by external trigger).
   5. Read the resulting VSENSE data in the ADC data register
   6. Obtain the temperature using the following formula:
   Temperature (in °C) = {(V25 - VSENSE) / Avg_Slope} + 25.
   Where,
   V25 = VSENSE value for 25° C and
   Avg_Slope = Average Slope for curve between Temperature vs. VSENSE (given in mV/° C or μV/ °C).
   Refer to the Electrical characteristics section for the actual values of V25 and Avg_Slope.

   Looking at the STM32F103xB datasheet, section 5.3.19 "Temperature sensor characteristics", the slope is positive, so the value will increase with increasing temperature, and not decrease as you have explained.

   Also, the code you posted doesn't increase the sample time to 17.1us, which might also effect the results.
   I think that means:
   ADC1->regs->SMPR1 = (0b111 << (3*6)); // set channel 16 sample time to 239.5 cycles
   239.5 cycles of the ADC clock (72MHz/6=12MHz) is over 17.1us (about 20us), but no smaller sample time exceeds 17.1us.
   (The default sample time for all the ADC channels is initialised to 55.5 ADC cycles, about 4.6us, which section 11.10 says is too low)

   So I think the channel number should be 16, the temperature sensor switched on, and the sample delay increased to 17.1us delay.

   (full disclosure: I am not a member of LeafLabs staff.)

   Posted 4 years ago #
3. 

   tdc218
   Member

   Thanks for your comments. Based on that, some further reading, and studying the adc.* code, here's some code that seems to be working:

   #include <stdio.h>
   #include <adc.h>
   
   #define LED_PIN BOARD_LED_PIN
   
   uint16 read_temp(void)
   {
     adc_reg_map *regs = ADC1->regs;
   
   // 3. Set the TSVREFE bit in the ADC control register 2 (ADC_CR2) to wake up the
   //    temperature sensor from power down mode.  Do this first 'cause according to
   //    the Datasheet section 5.3.21 it takes from 4 to 10 uS to power up the sensor.
   
     regs->CR2 |= ADC_CR2_TSEREFE;
   
   // 1. Select the ADCx_IN16 input channel.
   
     regs->SQR1 = 0;	        // set regular channel sequence length to 1
     regs->SQR3 = 0b10000;		// select channel 16
   
   // 2. Select a sample time of 17.1 μs
   // per gbulmer: set channel 16 sample time to 239.5 cycles
   // 239.5 cycles of the ADC clock (72MHz/6=12MHz) is over 17.1us (about 20us), but no smaller
   // sample time exceeds 17.1us.
   
     regs->SMPR1 = (0b111 << (3*6));
   
   // 4. Start the ADC conversion by setting the ADON bit (or by external trigger).
   //  note by virture of bit 11 being zero returns right aligned results.
   // Aparently we also need SWSTART - tdc
   
     regs->CR2 |= (ADC_CR2_SWSTART | ADC_CR2_ADON);
   
   // wait for conversion to complete
   
       while (!(regs->SR & ADC_SR_EOC))
   	;
   
   // 5. Read the resulting VSENSE data in the ADC data register
   
     return (uint16)(regs->DR & ADC_DR_DATA);
   }
   
   void setup(void)
   {
   /* Set up the LED to blink  */
     pinMode(LED_PIN, OUTPUT);
   
   // kill time until serial monitor is hooked up
     while (!(SerialUSB.isConnected() && (SerialUSB.getDTR() || SerialUSB.getRTS())))
     {
       toggleLED();
       delay(100);
     }
   // show start
     if(SerialUSB.isConnected() && (SerialUSB.getDTR() || SerialUSB.getRTS()))
       SerialUSB.println("Temp mon startup");
   }
   
   void loop(void)
   {
     uint32 tstart, tend;
     uint32 result;
     char buf[64];
   
     tstart = micros();
     result = read_temp();
     tend   = micros();
   
     sprintf(buf, "%04x  %08x", result, tend-tstart);
     if(SerialUSB.isConnected() && (SerialUSB.getDTR() || SerialUSB.getRTS()))
       SerialUSB.println(buf);
   
     toggleLED();
     delay(999);
   }

   A couple of additional things that seem be necessary:

   1. according to the data sheet section 5.3.21 it takes from 4 to 10 uS to power up the sensor. So I set the TSEREFE bit in CR2 first.

   2. I had to set the SWSTART bit in addition to the ADON bit to get the thing to convert. I dug this detail out of looking at the code in adc_read().

   It's about 72 deg F (22.2C) in my work space right now and I'm getting pretty consistent 0x06b1/0x06b2. If I put my finger on the chip, it steps it's way down to 0x06a7, and the frozen bolt test causes it to go higher. This makes sense because Temperature (in °C) = {(V25 - VSENSE) / Avg_Slope} + 25. Doing the math, VSENSE = V25 - Avg_Slope * (Temp - 25) so as the temperature goes up, VSENSE goes down, which is exactly what I'm seeing here.

   At some point I'm gonna stick the Maple in the refrigerator and see what kind of numbers I get. I'd be curious to hear what results others get from running this code.

   Posted 4 years ago #
4. 

   gbulmer
   Moderator

   tdc218 - Hot, and Very Cool (did you see what I did there? ;-)

   Thank you for posting the code. It seems like a great demo for anyone to try.

   I apologise for misreading the temperature calculation :-(

   As a thought ... Some of the work could get done by the functions in adc.h & adc.c
   So I think the unique parts could be trimmed to:
   

   void setup_temperature_sensor() {
     adc_reg_map *regs = ADC1->regs;
   
   // 3. Set the TSVREFE bit in the ADC control register 2 (ADC_CR2) to wake up the
   //    temperature sensor from power down mode.  Do this first 'cause according to
   //    the Datasheet section 5.3.21 it takes from 4 to 10 uS to power up the sensor.
   
     regs->CR2 |= ADC_CR2_TSEREFE;
   
   // 2. Select a sample time of 17.1 μs
   // set channel 16 sample time to 239.5 cycles
   // 239.5 cycles of the ADC clock (72MHz/6=12MHz) is over 17.1us (about 20us), but no smaller
   // sample time exceeds 17.1us.
   
     regs->SMPR1 = (0b111 << (3*6));     // set channel 16, the temp. sensor
   }

   Using adc_enable(ADC1); in setup() would do the rest of setup, though I also think that is called before setup anyway.

   Then
   result = adc_read(ADC1, 16);
   would get the reading.

   (full disclosure: I am not a member of LeafLabs staff.)

   Posted 4 years ago #
5. 

   tdc218
   Member

   gbulmer - I see what you did there. :D

   Your suggested code worked perfectly first try. I had a slight concern that there might be some interaction between the setup for the temp sensor and normal analog reads so I tossed one into loop and no problem.

   It occurs that if one only wanted to sample the temp at long intervals, one might shut off the sensor when not in use and save a few milliamps. Possible (but as yet untested code might be):

   void shutdown_temperature_sensor(void)
   {
     ADC->regs->CR2 &= ~ADC_CR2_TSEREFE;
   }

   Posted 4 years ago #
6. 

   gbulmer
   Moderator

   tdc218 - It looks like you are The Temperature Sensing Expert !-)

   Thank you for sharing your experiences.
   It is something I've never thought to dig into (so thank you), and it looks like a very nifty *free* feature (I love free:-)

   If you can find time, maybe you could add a wiki page?

   Posted 4 years ago #
7. 

   mbolivar
   LeafLabs

   If you can find time, maybe you could add a wiki page?

   Please do! http://wiki.leaflabs.com

   Posted 4 years ago #
8. 

   tdc218
   Member

   I can do that. Where would be an appropriate place to put it in the wiki?

   Posted 4 years ago #
9. 

   mbolivar
   LeafLabs

   Thanks! How about putting it in the How To section?

   http://wiki.leaflabs.com/index.php?title=How_To_Guides

   Posted 4 years ago #
10. 

    tdc218
    Member

    Done! I'll go back and clean it up a bit more when I have a bit of time (if someone else doesn't beat me to it first).

    Posted 4 years ago #
11. 

    mbolivar
    LeafLabs

    Thanks! Looks great.

    Posted 4 years ago #

RSS feed for this topic

Reply

You must log in to post.