ST Micro M41T11M6 RTC Drift Issue

[MattBerberich]

I’m currently having an issue w/ ST Micro P/N M4T11M6 RTC w/ clock drift. Specifically, I’ve performed verification testing on 4 units to this point and I’m seeing over 8 seconds/day (Extrapolating to a value of 4.25 minutes/month) of drift from NIST time. The RTC clock calibration register only allows for a maximum of -2.75 minutes/month which isn’t enough range to allow for the drift observed to be calibrated out. This makes me question my PCA schematic and perhaps layout design. The XTAL used meets the required 32.768 KHz XTAL w/ 12.5 pF load capacitance. I have cut-away the GND plane under the XTAL as well (However, it turns-out the +3.3V plane isn’t cut-away, but is in the middle of the 8 layer stack-up and the RTC and XTAL are on the top layer). 

What am I potentially doing wrong that would cause such a drift issue?

We are in the middle of verification on a high-priority project w/ a tight schedule, so your urgency on this matter is truly appreciated. Thanks in advance for your assistance w/ resolving this RTC issue.

[AScha.3]

Do you use recommended crystal ?

+

>I’m seeing over 8 seconds/day

What then ? crystal/clock is to fast or too slow ?

[MattBerberich]

Yes, if you mean a 32.768 KHz XTAL w/ 12.5 pF load capacitance. I attached the XTAL datasheet for the XTAL I'm using in this design.

[Peter BENSCH]

@MattBerberich

You haven't answered the second question of @AScha.3, yet.

In addition, you mentioned the crystal, but what load capacitors do you have on pins 1 and 2?

Regards
/Peter

[MattBerberich]

AScha.3, 

I'm using ST Micro P/N M41T11M6F which is the SOIC8.  The datasheet specifically states that the load capacitances are integrated internal to the IC.  For the XTAL, I'm using Citizen P/N CM1610H32768DZFT which is a 12.5pF 32.768 KHz XTAL meeting the specification requirements in the M41T11M6F datasheet.  Do you know if the specified 35ppm tolerance in the M41T11M6F specification includes the XTAL or do I need to stack the PPM of the XTAL such that the total PPM will be 35PPM + XTAL PPM (20PPM for what I'm using) = 55 PPM???  Just trying to understand why I'm seeing > 8 seconds/day RTC drift.  I've tested 4 PCAs and unpowered, I'm seeing 8.7, 8.1, 8.2, and 8.4 seconds/day drift. I've attached both the XTAL I'm using and the M41T11M6F datasheet I'm referencing on page 20.

[AScha.3]

Still no info:

>I’m seeing over 8 seconds/day

What then ? crystal/clock is too fast or too slow ?

+

Did you try other crystals (types) ?

+

Did you try setting the calibration/correction ?

[MattBerberich]

Using FT/OUT pin on M41T11M6F on one PCA, I measured 512.0508 Hz, so +0.0508 Hz using a frequency counter. This I calculated to be +99.21875 PPM off from 512 Hz. When I measured the 32.768 KHz XTAL manually using a passive oscilloscope probe, I measured was 32,615.786 Hz.  The XTAL I'm using is 20PPM which is +/- 65.536 Hz, so that's a range of 32,702.464 - 32,833.536 Hz, which shows the XTAL is within tolerance.  Now, I don't think the oscilloscope method of measurement is as accurate as the FT/OUT pin, so take that w/ a grain of salt.  Not sure what to make of the XTAL frequency measurement directly using an oscilloscope.

I initially attempted to perform drift calibration, but was unsuccessful because we had a FW bug that was resetting the register 7 calibration register when using the command to set the time.  That FW bug has recently been corrected, and I am actually in-process to attempt maximum subtraction calibration by programming decimal 95 into register 7.  However, I only expect the calibration to account for roughly half of the RTC drift observed:  I'm seeing > +4 minute/month drift and calibration register is only capable of -2.75 minute/month calibration.

No.  I have NOT tried other XTALs.  NOT because I don't want to, but XTALs all come in different package sizes and layout footprints, so it's impossible to take some new XTAL and place it into the board design.  ST Micro doesn't offer a evaluation board for the M41T11M6F, so I'm stuck there as well.  I did consider this, but there's no realistic way to put in a new XTAL w/o adding stray C,L from rework.  It would take months to order a new board.  Why do that when I can find a different RTC IC that has an evaluation board that I can do proof of concept with.

[Hal3]

The too high frequency suggests crystal capacitance is too low. The data sheet Table 8 is ambiguous when it says in note 1 that capacitance is external for SOIC8, but note 3 says provided internally. And, @PeterBensch asks if you have external caps. Perhaps he can clarify.

Its a bit of delicate soldering, but  if no Bensch response I suggest adding increasingly larger external caps starting at about 3 pf until you get the needed accuracy.

[AScha.3]

Ok,

so your crystal is too fast ! 

1. play around  : put little drop of hot melting glue on crystal and tracks (to increase stray capacitance -> frequ. will go down a little) try. (glue can be removed easy....if you dont like.)

2. try other crystal...in same case :

3. or other clock chip... a very good old standard is 700-DS3232MZ+ (mouser) , +/-5 ppm (!).

4. Try compensation in software, if your device is usually powered all time:

I made a clock with 50mm LED 7-segment numbers, just for fun. :)

rtc clock is old DS1338 + crystal ; as this chip has no temperature compensation etc, i made the correction myself:

i set the correction in software, lets say: -28 ppm needed. The software just calculates, how many seconds is this, then every 35714 seconds from actual time one second is subtracted and - time drift corrected.

Since last spring ( 8 months or so) clock now is about 20 seconds drifted, because temperature is not compensated.

But anyway - very good, i think, for the simple way, the correction is done. And without any adjustment, just using crystal + chip from recycled cpu board...  :)

[MattBerberich]

I don't have external capacitors based on the datasheet information.  Again, I'm using the SOIC8 package.  The datasheet says that capacitors are externally supplied for the SO8 package.  Also, the datasheet includes no information for calculating and populating a C1 & C2 external load capacitors.  In my experience w/ NXP, ST MICRO MCUs, if load capacitors are required, there's a section on them w/ formulas provided for calculating the values and layout requirements, ect.  This doesn't exist in the datasheet for the M41T11M6F, which also reenforces the idea that they're not needed as they're integrated within the IC so long as you're using a 12.5pF XTAL.  I am using a 12.5pF XTAL.  See my layout below.  I have very little stray C as the XTAL is connected straight to the pins w/ close 4 mil traces.

If I solder in load caps as you suggest, to what GND would I attach on the right side?  There's nothing there so I would need to add a lead wire and add additional stray C on one side vs the other.