0

I'm working in some projects with an RTC. I'm using the PCF85063A. The problem is: the RTC counter is advancing from real time about 2 minutes per week. I searched for hardware and firmware errors, and the only thing that caused me doubts was the oscillator's load capacitors.

This is the crystal I'm using: enter image description here

As you can see the crystal frequency tolerance is 20ppm, it means about 1 minute of error per month in clock counter.

The RTC have embedded oscillator load capacitors:

enter image description here

I'm actually using 7pF because it is the default value. I found nothing about oscillator load capacitor vs oscillating frequency variation.

Can someone explain the relationship about my problem and the load capacitors, or is there something else I'm not considering?

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
Maurício Tochetto
  • 41
  • 4
  • [This question and my answer](https://electronics.stackexchange.com/questions/320059/different-15-and-10-pf-load-capacitors-on-32-768-hz-quartz-crystal) should give some good insights into the problem. – Andy aka Dec 12 '18 at 10:29

2 Answers2

1

The tolerance of the crystal = 20 ppm doesn't mean that an oscillator should operate at a frequency that accurate. 20 ppm is only the tolerance of the placement of the certain resonant point at the impedance curve of the crystal. The oscillation frequency of an oscillator is that where the whole feedback loop has Nx360 degrees phase shift. All other parts affect, too a little, although the steepest phase shift frequency dependence belongs to the crystal.

You can tune the oscillator a little with external reactances, but that chip has software calibration scheme where a few pulses are added or bypassed now and then. Check it in the datasheet https://www.nxp.com/docs/en/data-sheet/PCF85063A.pdf.

Michel Keijzers
  • 13,867
  • 18
  • 69
  • 139
0

The frequency changes are always in the opposite direction to C load changes and more sensitive to change when the load gets smaller.

This type of crystal can easily be pulled +-100 ppm by C load changes and possibly +/-200 ppm.

Since the clock is operating fast by 2 minutes/week or 2/(7x24x60)= +198 ppm, the capacitance needs to increase so the settings should be set to the crystal spec , in this case 12.5pF and not left as default of 7pF. User error.

There is also a temperature error away from 25'C at +/-6 ppm per deg C, which is an unlikely culprit.

Other user errors may be attributed to soldering (thermal shock) or dropped (mechanical shock) as more likely than a good supplier, out of spec.

Layout differences from those recommended can contribute more or less stray capacitance to ground on the order of a few pF.

Tolerance errors of the load capacitance choices of 7 & 12.5pF of the RTC chip are both +/-40% max, so this can contribute greatly but not quite as much as the user error of using the wrong load setting.

If all datasheet specifications are adhered to, it is tested by the OEM and guaranteed to be within the tolerance spec @25'C here at +/-25 ppm.

By following the RTC datasheet, the measured error can be corrected digitally to less than 2 ppm.

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182