From IC guys:

" After a thorough accuracy study of the Temp sense circuit, we found that with our current equation, an average part can read 7 degrees lower than a known forced temperature. We also found out that the standard variance was around 2C; which is the tightest distribution that we could create. We need to change the temp sense equation to center the average part around the target temperature. "

New equation: Tmeas = (Nmeas - n1) / slope + t1 + offset n1= fused room count t1= 25 offset=3.580661 slope= 0.4148468 – 0.0015423*n1

According the new equation, update the thermal driver. c1 and c2 changed to u64 type and update comments.

Signed-off-by: Peng Fan peng.fan@nxp.com Cc: Stefano Babic sbabic@denx.de --- drivers/thermal/imx_thermal.c | 38 +++++++++++++++++++++----------------- 1 file changed, 21 insertions(+), 17 deletions(-)

diff --git a/drivers/thermal/imx_thermal.c b/drivers/thermal/imx_thermal.c index 0509094..b159464 100644 --- a/drivers/thermal/imx_thermal.c +++ b/drivers/thermal/imx_thermal.c @@ -22,8 +22,9 @@ /* board will busyloop until this many degrees C below CPU max temperature */ #define TEMPERATURE_HOT_DELTA 5 /* CPU maxT - 5C */ #define FACTOR0 10000000 -#define FACTOR1 15976 -#define FACTOR2 4297157 +#define FACTOR1 15423 +#define FACTOR2 4148468 +#define OFFSET 3580661 #define MEASURE_FREQ 327 #define TEMPERATURE_MIN -40 #define TEMPERATURE_HOT 85 @@ -54,39 +55,42 @@ static int read_cpu_temperature(struct udevice *dev) struct thermal_data *priv = dev_get_priv(dev); u32 fuse = priv->fuse; int t1, n1; - u32 c1, c2; + u64 c1, c2; u64 temp64;

/* * Sensor data layout: * [31:20] - sensor value @ 25C * We use universal formula now and only need sensor value @ 25C - * slope = 0.4297157 - (0.0015976 * 25C fuse) + * slope = 0.4445388 - (0.0016549 * 25C fuse) */ n1 = fuse >> 20; t1 = 25; /* t1 always 25C */

/* * Derived from linear interpolation: - * slope = 0.4297157 - (0.0015976 * 25C fuse) + * slope = 0.4445388 - (0.0016549 * 25C fuse) * slope = (FACTOR2 - FACTOR1 * n1) / FACTOR0 - * (Nmeas - n1) / (Tmeas - t1) = slope + * offset = 3.580661 + * offset = OFFSET / 1000000 + * (Nmeas - n1) / (Tmeas - t1 - offset) = slope * We want to reduce this down to the minimum computation necessary * for each temperature read. Also, we want Tmeas in millicelsius * and we don't want to lose precision from integer division. So... - * Tmeas = (Nmeas - n1) / slope + t1 - * milli_Tmeas = 1000 * (Nmeas - n1) / slope + 1000 * t1 - * milli_Tmeas = -1000 * (n1 - Nmeas) / slope + 1000 * t1 - * Let constant c1 = (-1000 / slope) - * milli_Tmeas = (n1 - Nmeas) * c1 + 1000 * t1 - * Let constant c2 = n1 *c1 + 1000 * t1 - * milli_Tmeas = c2 - Nmeas * c1 + * Tmeas = (Nmeas - n1) / slope + t1 + offset + * milli_Tmeas = 1000000 * (Nmeas - n1) / slope + 1000000 * t1 + OFFSET + * milli_Tmeas = -1000000 * (n1 - Nmeas) / slope + 1000000 * t1 + OFFSET + * Let constant c1 = (-1000000 / slope) + * milli_Tmeas = (n1 - Nmeas) * c1 + 1000000 * t1 + OFFSET + * Let constant c2 = n1 *c1 + 1000000 * t1 + * milli_Tmeas = (c2 - Nmeas * c1) + OFFSET + * Tmeas = ((c2 - Nmeas * c1) + OFFSET) / 1000000 */ temp64 = FACTOR0; - temp64 *= 1000; + temp64 *= 1000000; do_div(temp64, FACTOR1 * n1 - FACTOR2); c1 = temp64; - c2 = n1 * c1 + 1000 * t1; + c2 = n1 * c1 + 1000000 * t1;

/* * now we only use single measure, every time we read @@ -118,8 +122,8 @@ static int read_cpu_temperature(struct udevice *dev) >> TEMPSENSE0_TEMP_CNT_SHIFT; writel(TEMPSENSE0_FINISHED, &anatop->tempsense0_clr);

- /* milli_Tmeas = c2 - Nmeas * c1 */ - temperature = (long)(c2 - n_meas * c1)/1000; + /* Tmeas = (c2 - Nmeas * c1 + OFFSET) / 1000000 */ + temperature = lldiv(c2 - n_meas * c1 + OFFSET, 1000000);

/* power down anatop thermal sensor */ writel(TEMPSENSE0_POWER_DOWN, &anatop->tempsense0_set);