32KiHz external oscillator

[aurmer]

I have referenced a few application notes and spec sheets to come up with this schematic and layout for my external oscillator.

If anyone sees an obvious mistake please let me know... this is my first timing circuit. But I do not know the proper load capacitor value.

Crystal Spec sheet says 15pF C_L
This document (section 3.3) has what looks like my formula.

But I am a bit under-experienced when it comes to calculating trace capacitance. I assume it isn't negligible. How do i calculate it?

EDIT: Blerg, the schematic shows "U2" for the crystal and it should read "Y1".

[uncle_bob]

Hi

You have two constraints and two components. One constraint is the gain required to reliably start the oscillator. The second constraint is lhe load capacitance that puts the crystal on frequency. Both are dependent on the capacitors and the oscillator internals.

How much do you care about accuracy?

How much simulation do you want to do?

How good a model of the oscillator internals do you have?

In most cases the answer to all three is "not much". In that case, just drop a couple 15 pf's in there and move on. It should work ok.

Bob

[aurmer]

Timing is essential. The problem that I am trying to solve here is how to make this design always as time-accurate as possible every time it is built.

The oscillator has an inherent error of up to 10ppm (~26 seconds of drift every 30 days).
The MCU is capable of adjusting for error up to around 200-300ppm.
So I will have a calibration routine for each unit, but that just means that the error must stay within the "adjustable range" of +/- 200ppm.

[uncle_bob]

Hi

Some basics:

The part you have on the PCB board is a crystal. It plus the capacitors plus the internals of the main chip create an oscillator. The crystal has a finish tolerance at a specified load and temperature. In addition the crystal has an aging rate and a temperature coefecient.

So:

To hit the specified finish tolerance, you need the circuit to present exactly the correct load. There are two ways to do this:

1) Build one and change the caps until it is correct

2) Get the IC manufacturer to give wou enough data to analyze it up front

If you succeed at number 2 the only way you will *know* is to go back and do number one.

Ultimately, it you really want "maximum accuracy" you will need to calibrate each one. This includes calibrating the temperature performance. A 32 KHz crystal normally has a pretty major TC.

Bob

[bitslice]

Wouldn't a GT cut crystal help, then you'll have a 5ppm tolerance to start with

[uncle_bob]

Wouldn't a GT cut crystal help, then you'll have a 5ppm tolerance to start with

Hi

The GT (if you can get them) would have a better temperature tolerance. It still would have an aging rate and a set tolerance. The combination of all of that is going to be well past the 5 ppm level.

The real answer for accurate timing it so ditch the crystal. Switch over to one of the low power TCXO's at the same frequency. That takes out all of the fiddling and likely costs less than an exotic crystal.

Bob