TXCO for RFM69

[perky]

Has anyone actually removed the crystal on the RFM69 and replaced it with a TXCO? If so which one did you use? The pinout seems to be for a can with the other two pads grounded, but TXCO's need a VCC. I've not yet taken one off or checked the pinout to see how easy it might be to bodge one on.

I need an accurate transceiver driven by +/-2ppm clock source or better for a server that's used for calibrating clients during manufacture, so not worried about power consumption.

Mark.

[WhiteHare]

No, I haven't done it, but the DS (3.2.1. Reference Oscillator) seems to suggest a different approach than a simple swap.  The way I read it, the existing XTAL is connected to XTA and XTB.  However, for a TCXO, it says:

An external clock can be used to replace the crystal oscillator, for instance a tight tolerance TCXO. To do so, bit 4 at address 0x59 should be set to 1, and the external clock has to be provided on XTA (pin 4). XTB (pin 5) should be left open.  The peak-peak amplitude of the input signal must never exceed 1.8 V.

Figure 4 shows the schematic.  Judging from that, deadbug bodging a HopeRF after extracting its XTAL looks like it at least has a chance.  I'm surprised no one has tried it yet.

[perky]

It appears that the two load capacitors are essentially pads that connect to all 4 pads of the crysta and could act as pads to solder to. As long as the output of the oscillator can be arranged to be on XTA it stands a chance ;_
Mark.

[WhiteHare]

I guess try it on an RFM69HW since it's not as cramped.

[perky]

I guess try it on an RFM69HW since it's not as cramped.

I can't do that, my design uses RFM69CW. I think it can probably be done though, the only thing that worries me is the 1.8V max input spec. It's also unclear whether AC coupling is allowed (some inputs have a DC bias).
Mark.