Low Speed mode using Temperature sensor

Hi Man

     I design a project that's low power system, the MCU(RL78L1C) goes to STOP mode at most time. Now I want to use Temperature sensor and REGC pin output stable voltage gives Op-amp. So the MCU must work at Low Speed Mode, but the problem is the hardware user manual said temperature senor only work at High Speed mode. I forced read the channel of temperature sensor at LS mode. It works! My prototype board includes a 3V battery, RL78L1C, Op-Amp(power from REGC pin, that's current is only 2uA) and  a 4 number segment LCD panel to show the temperature every second. MCU's Option byte 2 is set to 000C2H: 0xA9(LS mode, 16MHZ). I run the prototype board for 2 weeks, it works good.

    So I want to ask MCU runs in LS mode and read the temperature sensor channel, does it have any risk?

Thanks

Steven

  • Hi Steven,

    Yes, this has much risk.

    Renesas does not guarantee the Temp sensor will work correctly in LOW SPEED Mode or for VDD <2.4V.  It may work on some RL78/L1C units at 25C, but may not work at extreme temp limits or if you try across different date codes (the temp errors could be larger).  Also, using 16MHZ in LOW SPEED Mode and VDD <2.4V violates the RL78/G1C specification, so we strongly discourage it.  The Renesas RL78/G1C design team sets the Electrical specifications to guarantee high reliability over -40C to +85C, and with slightly varying silicon process limits.  We cannot guarantee 16MHZ operation in LOW SPEED mode.

    When RL78 CPU/system clock frequencies > 8MHZ are used, HIGH SPEED Mode should be set.  The HIGH SPEED mode turns on an internal charge pump that allows Code Flash memory fetch.  In LOW SPEED mode, the charge pump is not turned on, and the Code flash memory access could fail at any speed >8MHZ.  Then in HIGH SPEED Mode, any CPU/system frequency from 1MHZ-16MHZ can be used as long as VDD = 2.4V or greater.

    I have been discussing these issues with Bob M., so he is getting these answers from me.

    Regards,

    Mike

  • BGR_RL78.PDF

    Hello Steven and Mike,

    I have another idea.

    BGR of REGC-pin is 1.8V at LS (low speed main) mode.
    On the other hand, the temperature sensor voltage becomes 1.284V temperature reaches -40 degrees, as temperature coefficient is "-3.6mV/°C". (1.05V+3.6mV*65)
    There is a margin of 0.5V. (=1.8V-1.284V)
    The voltage variation by the temperature of Vbe(Vf) inside the BGR circuit seems not to be so large.

    I tried changing the power supply voltage in order to reduce the voltage margin. I attach the results.(REGC,VCONST,VTMPS)
    Renesas does not guarantee this result, but I believe that no problem operationally.

    P.S.
    mod. value 1.284V from 1.1945V

  • Hi Mike and Kirin

       Thank you for helping.

       My product operation temperature: 0ºC to 40ºC, the Max temperature is 55ºC, this is only for UL test. The prototype board is base on RL78L1C(R5F110MJ), one 3V battery supports the power. VDD=AVDD=2.4V(Min.) to 3.1V(Max.). The fHOHO=fIH=16MHZ(I will drop to 8MHZ).  A 4 digital LCD panel shows the temperature value. AREF+ connects to REGC pin and VREF- connect to external GND. MCU weeks up to read the value from temperature sensor and Op-Amp output every second. The current of PCBA is only 4uA. The temperature sensor works good in LS mode at 55ºC.

       At the beginning design, the MCU works in HS mode, but the REGC pin outputs a square wave  from 1.795V to 2.097V. Because Op-amp is powered by REGC and IN+ input connect REGC too by two resistors, so Op-Amp output is square too. That will affect this ADC channel reading. That's why I set MCU working in LS mode and read temperature sensor channel. By the way, the Internal voltage reference(1.45V) is bad than REGC when PCBA operating temperature is from 0 to 55ºC. I put 5 samples PCBA(MCU is RL78G1A) to a chamber and set the temperature 55ºC. Their BGR voltage is from 1.43V to 1.46V. So I didn't use BGR in my design.  

       I just want to know if VDD>2.4V, MCU running at LS mode(8MHZ), temperature operating range is 0ºC to 55ºC, does the temperature sensor work OK?  

       Mike: Bob only told me MCU is only HS mode using temperature sensor and he didn't explain why.

    Steven Ma

  • Hey Steven,

    55ºC is OK.
    Since the temperature coefficient is minus, it is stable when high-temperature.

    By the way, the switching noise of REGC is large, as REGC is CPU core power supply.

  • Hi Kirin,

    Thank you for the detailed PDF graphs.  Your information is interesting.

    Do you have an email address so I could contact you?  My email address is listed in the RenesasRulz forum.  I support RL78 technical issues at Renesas.

    Thanks.

    Hi Steven,

    Now I understand why you changed from HIGH SPEED Mode to LOW SPEED Mode setting.  Yes, the square wave on REGC is because:

    a. In HIGH SPEED Mode (CPU RUN and CPU HALT) the REGC output = 2.1V (within +/-0.1V or less)

    b. In STOP mode (for HIGH SPEED and LOW SPEED setting in 000C2H option byte), the REGC output = 1.8V (within +/-0.1V or less)

    I think you are mostly in STOP mode and wake up periodically to CPU RUN mode, so the REGC voltage will switch from 1.8V to 2.1V every time the CPU starts up in HIGH SPEED Mode.

    I could see that affects your OP-AMP output stability, if REGC is changing.

    But using LOW speed mode, the REGC pin will stay almost constant at 1.8V (within tolerance).

    Another consideration is the 12bit ADC is only specified for VDD = 2.4V minimum to 3.6V max.  Below VDD = 2.4V we just show 10bit ADC operation.  Does that work for you?

    Regarding " I just want to know if VDD>2.4V, MCU running at LS mode(8MHZ), temperature operating range is 0ºC to 55ºC, does the temperature sensor work OK? ", I will contact Renesas Japan." for a better explanation.

    If running LS mode it is a good idea to use 8MHZ speed or lower.  16MHZ operation in LOW SPEED mode is not guaranteed and may fail.

    Regards,

    Mike

  • Hi Kirin

       There's a 1uF capacitor connect to REGC  and if in LS mode, the noise on pin REGC is very low(<1mV). It doesn't affect ADC part. In my experiment, the REGC output (1.8V) is batter than internal voltage reference(1.45V).  Thank you very much!

    Hi Mike

       In my design, if the battery's voltage is lower than 2.5V, there will show a "Low battery voltage" icon on LCD panel. So in the actual application, the VDD is great than 2.4V.

       Thanks

    Steven

  • RL78G13_temp_vs_VDD.PDF

    Hello Steven,
    I have additional information.
    I have measured the sensor values ​​with changing the temperature from -18 ℃ to 59 ℃.
    If the power supply voltage is constant, characteristics of from 0 ℃ to 59 ℃ is good. (LS was the same characteristics as HS.)
    The more voltage is low, the more it seems to be stable.
    VDD=2.4V has best linearity in more than 0 ℃.

  • Hi Kirin

       Thank you very much for your chart.

       Mike has asked Japanese engineer in Renesas. Please see these it's from Mike's email.

    The reason the RL78/L1C only shows Temp sensor operation in HIGH SPEED mode is that Temp Sensor was designed with REGC (internal CPU core voltage) = 2.1V level, which only occurs for HIGH SPEED Mode.

    In LOW SPEED Mode (or in STOP mode) the REGC voltage is always 1.8V for active CPU voltage.  Renesas Japan said that the Temp sensor errors can become very large when TEMP Sensor is powered at 1.8V, but they cannot tell us how large the errors are.  (they didn’t test enough to give a guarantee spec.)

    If you take a risk, and use the TEMP SENSOR in LOW SPEED Mode, Renesas cannot guarantee good/consistent accuracy for temperature measurements.  So if large temp errors occur in your final product, Renesas cannot be liable for the results.

    However, if you do use LOW SPEED mode, make sure to use option byte 000C2H setting for 8MHZ or lower speed.  If >8MHZ speed is used in LOW POWER mode, the CPU can malfunction which could cause your end product to fail.

    Actually, Renesas Japan told us 3 years ago not to use Temperature sensor in LOW POWER Mode, since the Temp sensor circuit was designed to use internal REGC = 2.1V and potentially can have large errors when powered at 1.8V.  Just because you see good accuracy on a few RL78/L1C units doesn’t mean it will always have good accuracy.

    I think the problem is we know the Temp sensor error is greater at 1.8V supply, but we don’t know the total range of error.  This is the situation for all RL78 MCU versions except a newer one called RL78/I1D series.  The I1D has modified internal design so Temp sensor can work well in LOW POWER mode down to VDD = 1.8V.  Unfortunately, RL78/I1D does not have LCD drive outputs and the largest pin count is 48pins.  RL78/I1D does have 12bit ADC and 2-4 Op-amp channels.

    Kirin:

        I made a prototype board for RL78L1C, it looks linear when temperature is over 0C. It only needs to do the calibration for temperature offset error. Actually I'm only care about the range 40C to 55C, because the prototype board is very sensitive at high humidity and high temperature. My prototype board works very good, has a high accuracy with temperature when MCU run in LS mode at 3V. But I don't know if there are more  samples to do these test, how is the accuracy of temperature. I will make more than 10 samples PCB in PCB manufacture, then do these test again, if you like please leave you email, I will sent the test result to you. My email: steven@ppi-stech.com.

      Thanks

    Steven

  • Hey Steven,

    Thank you for much information.

    I have tried on the evaluation board of G13 and G14.(QB-R5F100LE-TB, QB-R5F104PJ-TB)

    I did not know that internal structure of the temperature sensor in a newer L1C and others changed.

    When practicing what is not indicated to a user's manual, it is performed at my own risk.
    After performing enough evaluation, of course.

    Thank you, I take care of the combination of 000C2H(clock speed) and 000C1H(LVD) in order to ensure normal oparation.

    Now, you are testing.
    I think that the sample of same lot has the similar characteristic.

    When measuring with the dispersion(margin) of the characteristic,

    it must be necessary to get the sample in which Transistor Vth of Pch and Nch differ intentionally.

    Least five types of samples.

    • Pch,Nch=High,High
    • Pch,Nch=High,Low
    • Pch,Nch=Low,High
    • Pch,Nch=Low,Low
    • Pch,Nch=Mid,Mid

    Even if you can not get the variation sample,
    if the number of samples is increased, the dispersion will be able to be fixed statistically.(as F-test)

    Best Regards,

  • Hi Kirin

      What's your mean about these:

    " it must be necessary to get the sample in which Transistor Vth of Pch and Nch differ intentionally.

    Least five types of samples.

       Pch,Nch=High,High

       Pch,Nch=High,Low

       Pch,Nch=Low,High

       Pch,Nch=Low,Low

       Pch,Nch=Mid,Mid

    "

    Is setup MCU's I/O output status?

    Steven