Routing Oscillators Between IC

[EETwidget]

Hello EEVBLOG Forum. I am a long time watcher of EEVBLOG however this is my first post. I am working on a project and could use some advise on routing the oscillator.

PCB Tools: Eagle V6.4
IC1: ADAU1701 - analog devices DSP
IC2: CS4272 - Cirrus CODEC
Crystal: 12.288MHz

Overview:
I am designing a DSP system that will replace our old TI stuff. I need to be able to have one PCB that can be stuffed with or without the CODEC IC. The ADAU1701 does not have a MCLK out but requires MCLK be synchronized with Cirrus. Trying to keep the BOM low, I don't want to include a buffer if I can avoid it (Meaning DSP cant drive CODEC). Cirrus CODEC has a bidirectional MCLK that I assume I can use to drive the ADAU1701, so...

One idea I have is to use only one jumper to switch in the MCLKI between CS4272 and ADAU1701, although I do worry about floating PBC traces effecting the oscillator circuit. Is this approach possible and reliable?

[wraper]

First of all you cannot connect two oscillator outputs together. IMO you should use CS4272 oscillator. Then output clock from MCLK pin (buffered clock output) of the CS4272 to MCLKI pin of the ADAU1701, if I read datasheets correctly.

[EETwidget]

JP9 is the jumper that selects if the oscillator feeds the ADAU1701 or CS4272, it will never be connected together. If JP9 is feeding CS4272, the output MCLKO from CS4272 is feeding ADAU1701. If JP9 is feeding ADAU1701, then CS4272 is not even stuffed.
I was not sure if this would work, I pulled it from the EVAL-ADAU1701 data sheet, with their eval board a switch is used to select between an external master CLK and the on board oscillator.

[wraper]

JP9 is the jumper that selects if the oscillator feeds the ADAU1701 or CS4272, it will never be connected together. If JP9 is feeding CS4272, the output MCLKO from CS4272 is feeding ADAU1701. If JP9 is feeding ADAU1701, then CS4272 is not even stuffed.
I was not sure if this would work, I pulled it from the EVAL-ADAU1701 data sheet, with their eval board a switch is used to select between an external master CLK and the on board oscillator.

Maybe JP8? On your first schematic I see that both outputs are connected when codec assembled. It will work if R1 is not soldered when codec is present. However traces will likely be too long. I would suggest to make places for crystal near both chips and assemble only one of them depending on board variant.

[EETwidget]

Yes, JP8 not JP9.. fat fingers and a small keyboard. I think I found a better solution though.

The CS4272 can operate with its internal oscillator when XTO/XTI are left floating. So, I think I can just wire the circuit ..  Crystal to ADAU1701 normally, with a 0ohm between ADAU1701 MCKI and where the MCLKO attaches from CS4272.
When not using the CS4272, just a small trace is attached to the clock. I’ll test this before production but it should be a short enough trace, not disturbing the circuit.

When we use the CS4272, we would not stuff the crystal or R1/R2.. I don’t know if OSCO from the ADAU1701 should be left floating, I assume it should be grounded?

[wraper]

It might be solution, however didn't find in the datasheet default frequency before configuring the IC and clock stability. If that is not a problem then go for it. However it might be a good idea to place one more 0R resistor to cut off that trace when using crystal. BTW according to datasheet of the CS4272 allowed crystal frequency range is 16.384-25.600 Mhz.

I don’t know if OSCO from the ADAU1701 should be left floating, I assume it should be grounded?

Lazy to search now but there was clearly written about that in the datasheet.