MSP430F5438A RTC calibration

Tiago Bucker said:

For example, setting RTCCALS = 0 and RTCCAL = 0h would result in a -4 ppm decrease in frequency.

I just checked an older version of the users guide and there is no mentioning of the minimum calibration.

Apparently, it was originally meant to be -4+8* RTCCAL (+-0 by default), and it was later discovered that there is an offset of 2 on RTCCAL in the exact implementation, resulting in a calibration even if RTCCAL=0.

Now it is a very rary case that a crystal exactly has the correct frequency. All I ever had were too fast. because their temperature coefficient changes over temperature (usually being 0 at 25°C). So you alwas have to slow them down a bit. And if not, then jsut pick a different load capacitance to adjust the freuency. :)

If this note is correct, setting RTCCAL = 0 or RTCCAL = 1 have the same efect. I summarized some situations in the tables below. I appreciate if someone of TI confirm I'm corretc.

1) RTCCALS = 1

RTCCAL | Calibration
0               | +8,138 ppm
1               | +8,138 ppm
2               | +12,207 ppm
3               | +16,276 ppm
 ...             | ...
 63            | +260,411 ppm

2) RTCCALS = 0

RTCCAL | Calibration
0               | -4,07 ppm
1               | -4,07 ppm
2               | -6,105 ppm
3               | -8,14 ppm
...              | ...
63            | -130,24 ppm

Tiago Bucker said:

If this note is correct, setting RTCCAL = 0 or RTCCAL = 1 have the same efect.

If there is an offset of +8, then setting RTCCAL to n would result in RTCCAL*4 (by description of the RTCCAL value)  +8 (apparent offset, as described by the note). Setting it to 1 would give +12, 2 gives +16. It jsu tbegins with +8 and not with 0.
At least that's what description and note say. What they mean can be of course different and has to be verified either by experiments or by an official statement (whcih of course should be based on experiments or other in-depth knowledge too)

Jens-Michael Gross said:

If there is an offset of +8, then setting RTCCAL to n would result in RTCCAL*4 (by description of the RTCCAL value)  +8 (apparent offset, as described by the note). Setting it to 1 would give +12, 2 gives +16. It jsu tbegins with +8 and not with 0.

Strictly speaking, the note mentions a *minimum*, not an *offset*.

The description in point 4 still says:

4. Adjust the frequency, by performing the following:
(a) If the frequency is too low, set RTCALS = 1 and apply the appropriate RTCCALx bits, where
RTCCALx = (Absolute Error) / 4.069, rounded to the nearest integer.
(b) If the frequency is too high, clear RTCALS = 0 and apply the appropriate RTCCALx bits, where
RTCCALx = (Absolute Error) / 2.035, rounded to the nearest integer.

Hasn't TI replied anything about that yet?

I'm discussing directly with TI technical staff about this RTC module. Thank you.

If anyone are interested these are the results I have observed on my tests. The RTC was configured in hexadecimal mode. When I use BCD mode the results are completely different.

// RTCCAL = 1 | Delta
// 0  | 0
// 1  | +8,138 ppm
// 2  | +12,207 ppm
// 3  | +16,276 ppm
// ...  | ...
// 63  | +260,416 ppm

// RTCCAL = 0 | Delta
// 0  | 0
// 1  | -2,035 ppm
// 2  | -4,070 ppm
// 3  | -6,105 ppm
// ...  | ...
// 63  | -130,25 ppm

Besides that, there is an errata about RTC module of MSP430F5438A, where they say the RTC step calibration are +8/-4ppm, but I did not observe this behavior at all.

Care to share what news you have about the minimum calibration so far?

My measurements show no active calibration for RTCCAL = 0, I do not see the effect of this new NOTE (that appeared in august 2012).

Thanks for the update. My device has RTC6 in the errata as well, but I do see the 8/-4 ppm steps.

I am in BCD mode, though.

To be exact, the errata says "4ppm/-8ppm", but I think it is a typo (as it is not the double of the steps in the datasheet, rather minus the double. It does not match my measurements either, while 8/-4 does).