This silicon shortage really sucks.

As you know, shortly after designing my Si446x reference design board I discovered that the Si446x parts are out of stock everywhere. In fact, most EZRadio/EZRadioPro chips are out of stock well into 2021 or 2022.

I went ahead and ordered the PCB and parts anyway (I was able to snag a handful of Si4464 from Mouser). Last weekend I assembled a board:

Despite overbaking the PCB and sapping the silkscreen color, it actually works (at least for RX)!

 

Observations/notes (no particular order):

• I really wish I had watched this entire video (from the great Michael Ossmann) before designing my board: https://www.youtube.com/watch?v=TnRn3Kn_aXg. If I had, I would been able to greatly simplify my design by replacing most of the RF filtering and impedance matching capacitors and inductors (over a dozen in total) with a single part: https://johansontechnology.com/datasheets/0433BM41A0019/0433BM41A0019.pdf. That being said, this forum post casts doubt on usage of that part for Si4464: https://www.silabs.com/community/wireless/proprietary/forum.topic.html/si446x_is_there_an-uf6g
• Even with my OptiVisor manually picking and placing tiny 0402 parts is not fun. Especially when a little bit of solder paste gets on the tweezers and things start sticking.
• I’ve been researching alternative sub 1 GHz transceivers in the hopes of finding a compatible chip that is more readily available. The Texas Instruments line is interesting but I’m not sure it will work here. The frequency resolution of the Si4432 (used in the Tempur-Pedic remote) is 156.25 Hz. A chip like the TI CC1120 has a frequency resolution of 15 Hz. Clearly 156.25 is not a multiple of 15. While I think I could get away with tuning to a close-enough frequency for receiving packets, we’re interested in transmitting packets. I’m worried that “close enough” for transmitting will not actually be close enough.
• I had also been debating whether to switch to one of Silicon Labs’ combination microprocessors and proprietary wireless transceivers (many options here). Although they show as being in stock there is no guarantee they won’t also be depleted. Sticking to the planned ESP32 + separate transceiver provides flexibility.

 

Current plan

My plan is to tentatively stick with the Si4464. Although this means I won’t be able to actually start selling these devices until late 2021/early 2022 (unless manufacturing resumes before then), there’s a good chance development will take most of 2021 anyway. Should the rest of the project be completed much before then, I will look at alternative transceiver ICs. In the meantime I will focus on the non-RF portions of the project such as app development.