50 comments

[ 5.0 ms ] story [ 121 ms ] thread
It's always a bit disappointing when you have one of these pages and it's just "the designers left a footprint to fit the part" and no bodging required.

Still, seems a useful bit of forethought from the designers.

It's pretty rare designers don't leave a footprint for antenna connectors. The footprint is pretty small, and makes debugging and testing of the software stack far easier since you can plug the connector into other hardware which simulates other wireless devices, without requiring an expensive and hard to use RF chamber.
They use it in their part 15 compliance testing. Conducted power output is a standard test.

The alternative is bodging a connector onto the board.

The number of people doing this mod is probably less than 1 in 10,000 to 1 in 100,000 units shipped.

If the connector cost was $0.25, that could mean the Raspberry Pi foundation was spending $2,500 to $25,000 for every one person doing this mod. It would be a very poor financial decision.

Including the footprint is free, though.

I believe not including the connector factors into their FCC certification
> You'll be expected to never turn on your Pi Zero without an antenna attached again, should your transmitter get fatally confused by the mismatch of hardware-defined impedance assumptions.

Can the hardware really be damaged by using it without an antenna? I would not have expected that at all.

Indeed it can - if the impedances are not matched correctly, the transmitted signals can end up being reflected back down the antenna, causing damage to the transmitter.

The Wikipedia article on Standing Wave Ratios [1] has more on this topic.

[1] https://en.wikipedia.org/wiki/Standing_wave_ratio

It can, but I’d be a little surprised if the zero w makes enough power to destroy its output without an antenna.
> Indeed it can - if the impedances are not matched correctly, the transmitted signals can end up being reflected back down the antenna, causing damage to the transmitter.

I think I've seen this once in a documentary where this blew up a whole ship and then many others, too. I think it was called Independence Day or sth. like that.

AFAIK the same is true for any WLAN card with antennas, radio with antennas, HackRF, etc etc.
I know this is true for big multi-kilowatt power amplifiers.

But are there any on-chip silicon amplifiers which can be damaged transmitting into a short or open circuit? Logic to detect that case and power off is trivial and should make the manufacturing process easier too (for one thing, rework of a dry solder joint on an antenna no longer requires the SoC to be replaced).

Why does the person in the video cut off the PCB antenna with a dremel and short-cut the resistor pads, instead of simply taking the SMD resistor that is ~1cm left-bottom and connects the antenna trace to the PCB antenna?
That's explained in the comments on the original source. It turns out that the curved bit of trace is significant: if you leave it connected but unterminated it'll be resonant near the 5th harmonic of the wifi signal, which is bad.

Microwave RF is hard.

Oh, also it's a capacitor, not a resistor. I don't think that changes the explanation, though.
On the contrary, that makes all the difference: it's an LC circuit that is resonant at a frequency that you don't intend to amplify.
It is, but if the proposal is to desolder the component from the board, does it matter which it was?
By the time you get to that high a harmonic, the signal is way, way, way down.

I have to imagine these footprints, both the u.fl footprint and the 0402 footprint, were meant to be used during testing, and I can't imagine that their own testing also involved Dremeling off a trace. Removing the cap at the end of the curved stub is surely fine.

Good point about the testing. I guess it's cheap and easy enough to try out, right?
If you trust your test setup for 17.4 GHz to be within a few dB of what you'd get at a certified test lab, sure.

Yeah leaving the stub is bad practice because it could cause you trouble in FCC certification, not because it always will or even probably would. Let's put it this way, if you wanted to get enough energy into your fifth harmonic to be within 10 dB of what you were putting out on the fundamental by leaving a stub like that... you would tear your hair out for a good long while before giving up.

Nothing about testing at those frequencies is cheap, and very little of it is easy, but it's certainly doable for someone who already has the lab.
It's simpler than that, though - if you're pushing energy into the 17GHz band, that's energy which is not going into 2.4GHz. If the curved trace is significant, leaving it on should raise the noise floor and lower the signal level of the wifi signal we care about. That should be observable in the performance of wifi itself. You don't need to directly measure it unless you care about the FCC.
Logarithmic scales are devilish things, though.

Spurious emissions have to be at least 20dB down from the carrier, which means a factor of 100. And I'm highly skeptical that you can measure a 1% difference in wifi signal strength by looking at something like performance alone, given all the counfounding factors.

(comment deleted)
A good question... The trace looks like it is enclosed in ground up to that moment so isn't intended to radiate until after the SMD resistor, so removing that resistor should be as good as cutting the trace.

My hat's off to them though - that is some steady hand dremel/soldering work on something extremely tiny. Those little u.FL connectors are hard enough to mate with the connecting wire - I dread to think how difficult soldering them on by hand must be!

Raspberry Pi Zero is basically no longer available. Out of stock everywhere. With the chip shortage who knows how long till they’re generally available again.
Dunno where in the world you are but there are still a few on sheves. https://coolcomponents.co.uk/products/raspberry-pi-zero-2-w still have a few but they are limiting sales to 1 per customer atm.
Where I live they sell pretty quickly but they also get restocked quickly. Slightly more pent up demand, but nothing like a GPU.

I bought one to set up some home automation.

Dunno what the performance of the Pi Zero 2 is compared to the Pi3 (iirc they have the same number of cores, but the Pi Zero 2 has less ram), but I happily run Home Assistant on a Pi3 even with a couple of DOODS detectors running on it processing camera feeds (however it takes about 500ms to process them iirc - its been a while since I set them up and I actually store the recordings of those cameras off the Pi).

Have fun, its a rabbit hole, but a fun one :-)

EDIT: Also if you are in the UK. Poundland have some Tuya/SmartLife compatable Wifi sockets for £5 a piece - https://www.poundland.co.uk/446890-ultrabrite-wifi-controlle... You can open them and reflash then with Tasmota but they are a pain to open.

The EASY option is to stick with using Tuya/Smartlife and add Tuya/Smartlife to HA, just remember to remove the Smartlife skill from Alexa otherwise you can end up with dupicates and Alexa getting confused when you want to turn things on/off.

ATM (imo) the "tuya old" intergration is better than the current offical tuya version (They changed things when updating it, which removed some compatibility but they said they would re add it at a later date, but duno if Tuya actually got round to it yet).

Also if you do use Home Assisant, get into the habbit of reading the patch notes before updating, quiet often there will be breaking changes which mess up with your set up.

Some Tuya/SmartLife sockets can't be flashed with tasmota. I've got 2 that don't use Espressif SoC's.
Yeah I have one RGB light that uses a ThingsTurn module (which iirc is based on a w600) and I couldn't be arsed to swap out the module for a ESP.

But I have 10 (8 in use, two on hand incase I ever want another lamp/secuirty light around the home, well atm the spares are in use for the christmas lights, but will go back into standby status after the lights come down.) of these ultrabight poundland sockets and all of them are ESP based. (though the ones I've recently added to my collection couldn't be flashed with tuyaconvert and needed some surgery, the last 2 in my collection I couldn't be arsed and just left them with the Tuya firmware and use the tuya API to control them, but thats my own lazyness not that they cna't be done.)

For ceiling lights I use shellys so I could retain the use of the wall switch to control them.

>You can open them and reflash then with Tasmota but they are a pain to open.

Not sure if you tried it, but depending on the age of the stock or the tardiness of the OEM in applying security patches, it may be possible to flash them without opening them first.

https://github.com/ct-Open-Source/tuya-convert#procedure

It's not out of stock, you just have to buy it at the market price, which may be more than you're willing to pay.

I just bought raspberry pi 4 on amazon through canakit. It came with a bunch of other stuff that was useful like sd card, case, cable, hdmi mini to hdmi converter, etc, so it really wasn't that bad.

Here's the equivalent for rpi zero for $50

https://smile.amazon.com/CanaKit-Raspberry-Wireless-Complete...

The 1st-gen Zero had the same resonance cavity antenna on-board,[1][2] but it had a nicer design for the external antenna, namely a 0-Ohms(?) resistor that could be desoldered and resoldered at an angle (and be reverted if need be),[3] rather than crudely cutting the trace.

[1] https://i.redd.it/fido28zdv1331.jpg

[2] https://www.embeddedcomputing.com/technology/analog-and-powe...

[3] https://www.briandorey.com/post/raspberry-pi-zero-w-external...

I'm curious if you could desolder and use one of the resistors further down the trace to avoid having to scrape off the mask.
That's so cool! Should be one hell of a wireless router. Maybe it would even work on bonding multiple connections.
Put it up a watertight enclosure Hook it up to a directional antenna and it would make a good PTP set up.
Okay, technically cool, but is it legal? Seems like something the FCC (or local equivalent) would have an opinion about
It's been a while since I've read FCC rules, but as far as I know they usually regulate power output, not antenna design.
They regulate effective radiated power (except for the AM broadcast band), which is a function of both conducted power output and antenna design. Any part 15 stuff like WiFi that has replaceable antennas will typically be certified for use with specific antennas. (there are plenty other rules too)

So this hack is not FCC certified (since it isn't what was tested for the zero W 2 certification) and may or may not be compliant.

My understanding, from experience of finding similar footprints on other hardware, was that this footprint is there for testing in the factory and not actually intended to be a u.FL footprint for an external antenna.

I think the first time I saw it was on a GL.iNet mini router which I wanted to add an external antenna to. It's possible, but you have to cut the trace to the on-board antenna.

Can anyone confirm or deny this?

This is just a supplier generating traffic to their website. The actual post is here and has the real details: https://www.briandorey.com/post/raspberry-pi-zero-2-w-extern...
"to fit a zero ohm resistor (0201 size 0.6mm x 0.3mm)"

I know what they mean here, but I had to re-read "zero ohm resistor" a few times.

0 Ω links are commonly used on SMT boards as bridges, to avoid having to either have another layer or make routing easier when you have relatively hard constraints (e.g. ground planes behind antennas). The added marginal cost of a 0 Ω 0201 SMT link is basically $0 to about 3 s.f.. The added cost of another layer in the board is decidedly nonzero. Yes, they made me blink the first time I encountered them too...