NRF51 design/layout questions

[jon]

We are currently using a NRF51 module (HRM1017) for a device in production. We want to migrate to the latest version of NRF51822 and move away from using a module to having all the parts directly on the PCB. Due to limited PCB real estate, we are considering the WLCSP version.

A few questions:

1. Looking at the reference layout for CxAx (nRF51x22-CxAx Reference Layout 1_3), you use micro vias in pads for the chip itself, but regular vias for the rest. It seems like a bit of a strange choice since you use HDI tech anyway and the point of this package is to get a smaller design. If we pack it tighter and use microvias for everything, is there anything we should keep in mind?

2. What are the implications of leaving out the "optional" RTC crystal, from a BLE perspective? We're not using it for anything else.

3. When using a balun instead of a matching network, we can't trim it as a part of improving RF performance with our particular antenna, enclosure etc. In your opinion, how big a problem is this?

4. The ref design doesn't have any shielding box. The top layer is rather "busy" with a lot of IO (as is our design). Do you have any guidelines on how to place and connect a shielding box into your reference design? How big a problem is it if we leave it out altogether? Our device has a plastic enclosure.

5. It looks like the 256k versions of the WLCSP package (CEAA and CFAC) are footprint compatible and only differ in the size of the package itself. Is this correct, ie can we layout for CFAC and later drop in a CEAA if we don't need the extra RAM?

Best, Jon

[Asbjørn]

1. We use regular vias as a part of cost reduction. We have seen some PCB manufacturers charge by the amount of micro-vias and hence we have chosen to use regualr vias as much as possible. There's no problem with using micro vias all over the board. Just make sure you have enough all over the board.

2. The external 32 kHz crystal is not needed for BLE operations. You can make use of the internal 32 kHz RC oscillator. The internal RC is not as accurate and hence you will have to have slightly longer radio events to ensure you hit the time slot. A lot of our customers use the internal, especially in your situation where they have a small board.

3. Generally as long as you implement the balun as the manufacturer specifies it we see better overall performance from using baluns than using a discrete solution. In an integrated balun they tune the component values specifically to the chip and hence they can for example make a 8.75 nH inductor in the integrated balung instead of settling with either 8.2 nH or 10 nH. So generally speaking the overall performance is better.

4. Unless you have a power amplifier there's generally no need to shield the top component side. Just make sure that you are decoupling essential parts of your design and follow Nordics reference layout. Connect all parts of ground together through a lot of vias.

5. Yes, the footprints are the same, the AC version with 32 kB RAM is slightly bigger, but the footprint under is the same as the AA version. You can replace them later as long as the edge of the large one isn't interfering with anything.

[jon]

Thanks, good answers!

/J

[jon]

Btw, for question 3, when you say "tune the component values specifically to the chip" - which chip do you mean? Are the baluns normally "matched" to a specific part (ie the NRF51 or a specific chip antenna?)

[Asbjørn]

No, the antennas are generic, but their characteristics may change from design to design. So it's important if you want to optimize radio performance that you tune antenna in your application. The baluns are specific for the chip or for a set of chips, so they might have to be tuned for different chips.