How to build a simple low cost reliable antenna switching circuit

Question

I'm looking to build a simple (MVP: minimal viable product) BLE antenna array circuit to test BLE direction finding (AoA/AoD) functionality.

I'm not wanting precision just yet. So, for my MVP I'm looking at just 3 maybe 4 antennas. For that I will need an RF switch and if I look on any of the major distributor websites I can find plenty single pole 3 or 4 throw (SP3T/SP4T) options.

My question is how to build the input control side, which usually consists of logic table like this one (for a SP4T RF switch):

Input 1	Input 2	Output Mode
0	0	RF1 on
0	1	RF2 on
1	0	RF3 on
1	1	RF4 on

The switching frequency is at the microsecond(s) level.

My BLE chip is a single 3V3 MCU so I don't want to use this MCU to handle the switching as well as the other stuff. It's not a dual core PSoc, for example, where I could've used their hardware timer blocks etc.

So what IC options do I have with a power rail of 3V3.

Is it a case of using a precision Timer IC with some logic gates (e.g. OR, AND, NAND etc.) or would something like a tiny 8-bit MCU do the trick. Basically looking for as few components as possible which is low cost and reliable etc.

Looking to experts for some practical advice.

Thanks

shabaz · Answer

Hi Colin, By coincidence we're thinking on the same track : ) I've been looking at a silabs reference design, and aim to work as closely as possible to it, although it will take longer and is a pain due to the component sizes (smaller than I've ever fitted, so I might get it done for me). However, generally for RF switches, traditional popular choices are Peregrine Semiconductor and Skyworks, although there are options from others now too, e.g. NXP may have some. They are easy-ish to use, provided a few rules are followed. They mostly work with 3.3V logic. The inputs/outputs would need to be microstrip, that would be easiest, and there are online calculators for it. Some of the Peregrine parts are hand-solderable, and others may be too. For things like SP4T as you mention, one technique is to just use SPDT, and arrange one to feed two of them, i.e. a total of three SPDT can make a SP4T. The benefit being that sometimes it is easier to more symmetrically make the paths if you can spread them across separate chips, so that all paths are the same length (if that's what is desired depending on usecase). The chips can be controlled by normal logic gates, they are compatible with that, so a microcontroller to directly control them is fine too. Other than that, some basic RF prototyping rules can be followed, like complete ground plane, and any top side ground (generally just around the perimeter) must have lots of vias (spaced a couple of millimetres apart at a guess) to the underside ground. You might also want to consider 4-layer, in case it helps with the microstrip, since the width of that is determined by the distance to the groundplane, as you'll see with the calculators or use thinner PCB. Also, boards will warp, so with 2-layer you might want to consider having slightly more ground plane on top again with lots of vias, it's a balancing act. 4-layer and using the closest-to-top layer as ground plane is an option. Personally I'd go with 4-layer for this.

Fred27 · Answer

I recently needed to test out a NFC reader with multiple antennas. I grabbed a Peregrine Semi evaluation board (EK42441-03). It worked very well and it seems to fit your requirements, including the control logic. The development board wasn't cheap (£83 incl. VAT) but the PE42441 seems more reasonably priced if you're spinning up your own board.

BigG · Answer

Just to add that when it costs less than $10, including postage, to get some prototype boards made I simply could not help myself... So I went ahead with a basic design. My limited understanding is that if your RF traces are very short on the PCB then width of trace is not too critical. I went for u.Fl connectors like those used on Hackaday project rather than the more common SMA connector if you're not using pcb trace antennas. This is a 2 layer board too. My intention was to use a 1mm board thickness as this provides me with a ground plane closer to the traces but it would've cost more so went with 1.6mm. I'll find out soon enough if this was a waste of $10 (I like to learn by my mistakes and still only know electrical impedance and EMI/EMC theory). I'm using a Skyworks SP3T on this board. They are still tiny! https://www.skyworksinc.com/-/media/SkyWorks/Documents/Products/301-400/SKY13317_373LF_200914K.pdf I'll check out the ones from Peregrine Semiconductor as you suggested as it looks like the Skyworks stock levels are very low.

Jan Cumps · Answer

shabaz wrote: ... the BLE stack on the chip is in control of the switching pins,... Not the exact same use case, but in the ballpark: On the 2 LoRa boards you gave me, there are two antenna connectors. They are for 433 MHz and the sub-1 GHz operation. They use PE4259 3 MHz switches. What's different: the frequency is lower than BLE, but it's getting there .. the destination are two different pins in the radio IC.

shabaz · Answer

Hi Colin, Even worst-case if it doesn't work to optimal performance, it's still a handy test board for testing out your code, confirming the switch behaviour, and checking footprints and so on.. definitely makes sense to test things cheaply quickly too. By the way, for the next prototype iteration, one more benefit of 4-layer, since you're using very small parts, the low-cost manufacturers produce 4-layer boards to slightly tighter specs (I feel/believe from what I've seen so far, but I have no confirmed evidence!).

BigG · Answer

Hi Shabaz There's a saying "great minds think alike" but I think in my case it's just coincidence Yes, the Silabs reference design is very handy alright. I've also come across a reference design from TI. http://software-dl.ti.com/simplelink/esd/simplelink_cc2640r2_sdk/2.30.00.28_new/exports/examples/rtos/CC2640R2_LAUNCHXL/… https://www.ti.com/lit/an/tida029/tida029.pdf And this project on Hackaday: https://hackaday.io/project/174175-micro-music My question is very basic and simply highlights my lack of electronics design know how and my uncanny knack of over complicating things. I'm simply wanting to evaluate or look at different hardware design options to offload the clock timer signal to trigger the RF switch logic as I have this notion that a micro second interrupt timer on my MCU will get in the way of doing other things. C://

shabaz · Answer

Hi Colin, As I understand (I could be wrong) the BLE stack on the chip is in control of the switching pins, so there's no need for any further consideration, apart from (perhaps) if you need to do any translation in logic to map to the signals required by the switches. If you're maybe aiming to do this yourself in your own code (wouldn't it need to be sync'd with the BLE chip tx or rx somehow? I don't know enough to be sure) then as far as microcontrollers go, as a ballpark, I think a 16MHz clocked processor would take about 1usec to get to a state from a timer interrupt where you have saved off enough registers and are ready to begin performing whatever you needed to do (of course, this is dependant on processor, RISC/CISC etc), so realistically you'd want at least a 48 MHz processor if not more. This is assuming you want to be using C and not assembler. I think I'd be wanting to process all three or 4 measurements within one timer interrupt when it occurs (and then you don't need as speedy a processor or as frequent timer interrupts). But this is still assuming that the BLE stack really cannot do it, so it's worth double-checking. Finally if you're just looking at repeatedly switching between antennas asynchronously that could be done with logic gates but I don't think this will work, so I guess I've misunderstood the requirement.

shabaz · Answer

I only briefly scanned the doc, but if that chart is from their antenna chamber testbed, then that's quite unrealistic conditions too. From what little I know, even if they can achieve 1 metre of accuracy they are doing extremely well. 2-3 metres of accuracy for 50% of measurements would still be considered good (i.e. many measurements can be taken and averaged, or throw out the outliers). 30 metres is pretty far, and room features/objects might also start impacting a lot (but I don't know for sure) so it might get significantly more untrustworthy at such a distance than what they have in the chart. I'm really speculating though! At such distances probably multiple devices are needed if the use-case required accuracy in those locations.

shabaz · Answer

From what I read, the timeslots are of that order (microsec), so 2us parts may be too slow. Some parts are faster in larger packages, e.g. PE4220 has 200nsec typical in MSOP package, so that might have been an option, except it's obsolete and hard to find : ( I've not found a convenient package part. The one that Silabs uses is 0.37mm pitch QFN.

shabaz · Answer

Hi Colin, Nice find. They do look (especially the Macom one) a lot more convenient to use. Another option to add to the list is PE4251 in MSOP package (0.65 mm pitch). It's relatively expensive ($1.16 in single qty) compared to other options, and only seems to be available at Digikey. It has very good isolation, and only slightly more insertion loss (perhaps not a major issue).

How to build a simple low cost reliable antenna switching circuit

Top Replies