Reading the CPU temperature

[elahav]

I have been trying, unsuccessfully, to read the temperature from the TS-ADC controller. I have configured the various TSADC registers and. as far as I can tell, the relevant clocks are all enabled. Still, any attempt to read TSADC_DATA0 results in 0. Also, if I understand correctly, the value of TSADC_USER_CON shows that the analog-to-digital conversion never finishes.

Here are the values of the various registers that I believe are related to the TSADC:

Code:

CRU_CLKSEL_CON27=1f (DIV=1f SEL=0)
CRU_CLKGATE_CON9=0 (TSADC=0)
CRU_CLKGATE_CON22=0 (PCLK_TSADC=0)
CRU_CLKSEL_CON25=187
CRU_CLKGATE_CON8=0 (PERILP1=0)
GRF_SARADC_TESTBIT=4
GRF_TSADC_TESTBIT_L=80
GRF_TSADC_TESTBIT_H=84
TSADC_USER_CON=1200
TSADC_AUTO_CON=10133
TSADC_AUTO_PERIOD=753
TSADC_AUTO_PERIOD_HT=753
TSADC_HIGHT_INT_DEBOUNCE=4
TSADC_HIGHT_TSHUT_DEBOUNCE=4

Am I missing anything?

Thanks,
--Elad

[xalius]

Are you doing some bare metal programming? I see you have clock, is there maybe also a power register somewhere? Maybe have a look at the Linux driver ...

[elahav]

Are you doing some bare metal programming? I see you have clock, is there maybe also a power register somewhere? Maybe have a look at the Linux driver ...

Yes, I looked at both the Linux and NetBSD drivers, and couldn't find anything beyond what I have. The drivers themselves rely on the device tree for clocks, but I followed the tree up as much as I could, and believe that all relevant clocks are enabled by default.
I'm running QNX on the board. Got the basics up (all 6 cores, MMU, GIC, timers, UART) but I'm afraid of ramping up the core frequency on the two clusters without monitoring the CPU temperature.

---

OK, got it to work. Probably just needed to be more careful with all the clocks:

Code:

// Assert reset.
out32(cru_regs + 0x438, (0x10001 << 8));
printf("CRU_SOFTRST_CON14=%x\n", in32(cru_regs + 0x438));

// Set up clocks
// HCLK_PERILP1, PCLK_PERILP1
out32(cru_regs + CRU_CLKSEL_CON25, 0x079f0187);
out32(cru_regs + CRU_CLKGATE_CON8, (0x10000 << 2));

// SCLK_TSADC
out32(cru_regs + CRU_CLKSEL_CON27, 0x83ff001f);
out32(cru_regs + CRU_CLKGATE_CON9, (0x10000 << 10));

// PCLK_TSADC
out32(cru_regs + CRU_CLKGATE_CON22, (0x10000 << 13));

// De-assert reset
usleep(20);
out32(cru_regs + 0x438, (0x10000 << 8));
printf("CRU_SOFTRST_CON14=%x\n", in32(cru_regs + 0x438));

// Whatever this is...
out32(grf_regs + 0x0e648, (0x10001 << 7));
out32(grf_regs + 0x0e64c, (0x10001 << 7));
usleep(20);
out32(grf_regs + 0x0e644, (0x10001 << 2));
out32(grf_regs + 0x0e64c, (0x10001 << 2));
usleep(100);

// Set up auto mode.
out32(tsadc_regs + TSADC_AUTO_PERIOD, 1875);
out32(tsadc_regs + TSADC_AUTO_PERIOD_HT, 1875);
out32(tsadc_regs + TSADC_HIGHT_INT_DEBOUNCE, 4);
out32(tsadc_regs + TSADC_HIGHT_TSHUT_DEBOUNCE, 4);

[xalius]

Good also if you are worried about your hardware, there is a fail-safe circuit that connects the over temperature output to reset, so as long as the overtemperature condition exists, the RK3399 will be held in reset.