One thing the article entirely neglects to mention is that the well engineered design came, via licensing, from an antenna specialist company called "Proant"
There's a huge number of people involved in the design of the rpi. The antenna is a great example of a subsystem designed by a separate team; another is the new PMIC for the 3B+[1].
I would really like to have access to an antenna chamber like that one :-). Lately I've been building/programming/characterizing SDRs and that generally also includes antenna. I realized I had forgotten most of what I had learned in school about electrodynamics so I've been refreshing all of those neurons as well.
The simplest solvers for Antennas do a great job on dipoles or other simple structures, and less well on antennae like the one on the Pi. I found it humorous that a well used tool was called 'Microwave office'[1].
I tried to do some genetic antenna design using a plotter, some conductive ink, and a simple S meter but it is really difficult to reliably connect the plotted antenna to the test fixture.
What I have learned, which matches the author, is that there is a lot of subtlety in antenna design that is not obvious by the traces they use on FR-4 or other substrates. That rabbit hole goes pretty far down.
I'm far more impressed by Raspberry Pi Zero W than by the original or even the latest Raspberry Pi. I'm using one as a streaming server for a security camera, and also a web server, and its working beyond my expectations.
Can you give me a brief walk through on how you accomplished this, or a resource you used to guide you? I have been interested in doing streaming security camera for a while but haven't seen a setup i like.
Me too, thanks for that set of links. Any links on connecting it up to Google's Cloud AI stuff to train it to recognize the crack-heads when they are peeking in the windows of peoples cars and rattling the doors of their tool sheds?
Agreed, I built a group of waterproof outdoor cameras with the Pi Zero W, that thing is a true gem.
One is bolted to a utility pole outdoors, 200+ft away from the nearest WiFi access point through several walls. It's not a high link speed, but it works with nothing but that little triangle PCB trace cutout for the antenna.
There's a medium-sized market for processor modules like this, as they're plenty good enough for quite a few applications.
For example if you need a big PCB, cheaper to make a big 2- or 4-layer PCB and put a module on it than to make your entire PCB 10 layers so you can break out a tiny BGA.
It has been out for a few years now but is never in stock anywhere for the advertised price of $5 ,it's sometimes available bundled with loads of crap you dont need for 10x this price, also you cant buy a qty of more than one anywhere.
If they've stuffed up the pricing and trying to reduce their exposure I am pretty sure a lot of people would be happy paying $10 instead.
I'm pretty sure there are more Tesla Model 3's being produced than Raspberry Pi Zeros :-)
Its definitely a loss leader with very limited availability, but thankfully it did spur a wave of sub-$10 single board computers, most of which are on par with much more powerful Raspis.
My local Micro Center usually has them in stock, but there's a quantity price penalty if you buy more than one, and the more you buy, the greater the penalty.
Yeah, it's a really fun set of tradeoffs to play with. You're correct to a first approximation. Channel bandwidth is is easier to come by at higher frequencies (a 20MHz wide channel at 150MHz would be a giant swath of the band, but a 20MHz channel at 5GHz is very small relative to the centre frequency).
But... then you start encountering higher path loss. To make up for that, you need to increase transmitter power (whether by increasing actual power, or by using higher gain antennas, etc) or by adding redundancy to the data stream (e.g. FEC). The added redundancy chips away at your bit rate, but corrects errors.
In practice, whether you get a better net transfer speed on a narrow lower-frequency band or a wider higher-frequency band is going to depend on a lot of factors. Sometimes it pans out, sometimes not.
According to [1] (which has more info on the antenna design), the resonant cavity antenna was designed by a company called Proant [2]. I remember reading the the antenna is the same on the Pi 3 and Zero W. There are a few options for PCB antenna design software out there, but I had never seen this type of PCB antenna before the Zero W. Not sure what software they use, or whether it's off the shelf or proprietary. Any antenna designers out there?
Typically you'd use a ceramic antenna on a product like this. What is the difference in radiated efficiency between this antenna and a ceramic one? I'm guessing there are other design considerations though, such as cost and circuit board area consumed.
A non-expert doesn't care about antenna gain - he just wants his wireless connection to work (or his radio signal to be clear). When he starts researching antenna gains he delves into the hobbyist scene. I know that hobbyists are not experts but understanding how dB work is simple enough after you understand the practical rules:
1. A difference of 3 in dB means times 2, a difference of 10 means times 10.
2. Because dB are in logarithmic scale adding dB multiplies the effect.
3. Negative numbers work the same but with loss instead of gain.
Thus a 3 dB gain antenna will double your signal strength while a 9 dB antenna will make it (9=3+3+3) 8 times stronger (8=2x2x2). Another example: 23 dB is a 200 times gain (23=10+10+3).
>1. A difference of 3 in dB means times 2, a difference of 10 means times 10.
How can this be right? Aren't you kind of fudging it a little?
Here's my train of thought:
First I was thinking "What the hell is going on with your math? There's no clear factor to turn base 2 into base 10 what the hell. How can what you say be true? How does this work?"
My next thought (based on your incorrect statement) was, oh, they didn't choose base 2: they chose every 3 to be another factor of 2 - so let's see why that works, why +10 is the same as * 10 if every +3 is * 2. Well, you can get to 10 by going 3 + 3 + 3 + 1 and you can also get 10 by going 2 * 2 * 2 * (1.25) = 10.
Okay, so if every +3 converts to * 2 then why exactly does the last term, +1 convert to * 1.25?
I thought, and thought about it. I couldn't make it work, based on your rules. So I checked. And the answer is it doesn't: 2^(1/3) isn't 1.25 as we would expect, it's 1.2599. That might seem "close enough" but I think it's not exactly how you say and your statements are misleading.
Thus 23 dB isn't 200 times stronger as you state (23 = 10 + 10 + 3), it's only approximately 200 times stronger. 200x stronger is 23.0103 dB, and 23 dB is 199.52 times stronger. [1]
While it's useful, and the error is pretty small, it doesn't help for those of us used to thinking in terms of bitfields or something that converts quite exactly.
It's definitely a very useful mental estimation trick though!
[1] which I checked with an online calculator here - https://www.rapidtables.com/electric/decibel.html (first I entered a level of 200 and clicked the top "convert" button, then I entered a dB of 23 and clicked the second "convert" button)
Well these are practical rules that are more or less used by people working with dBs. These people are usually don't think in term of bitfields - when you have a 43 dB gain antenna you don't care if it amplifies 20 000 times or 19 952 :) Also it's a nice way to show-off to people that do not know this rule!
In any case, I never said that you get 100% accurate results; if you want accuracy then you should use your calculator (or your logarithmic ruler); but why use a calculator when you roughly want to understand how much a 15 dB gain would be?
Finally, there's a nice way to find out how much 1 dB is with the mentioned rule: Notice that 1 = 10-3-3-3 thus it's 10/2/2/2 = 1.25 so 1 dB is approximately 1.25 gain as you said :)
Yeah, it's not exact, but in RF/microwave work you often get on order of a dB of loss through the cable or connectors anyway. Plus signal sources and spectrum analyzers are not spec'ed to be as accurate as you might expect. So you end up with error bars in your head.
The logarithmic scale is appropriate to represent things that behaves in a logarithmic way. A linear scale would be hard to manage since you'd often end up with extremely small and extremely large values next to each other which would be painful to represent and think about. Take a typical filter response graph for instance: https://iowegian.com/wp-content/uploads/2015/04/rcfreq.gif
Try graphing that with a linear scale, there won't be a lot to see I wager. It's true that it does take a little time to familiarize oneself with non-linear units at first (especially if one wants to add or subtract them for instance) but they have their uses. For measuring a signal they're clearly appropriate.
Out of the first 10 or so results for "wifi routers" on Amazon, atleast half of them mention antenna dBi values. What is Joe Buyer supposed to infer if one router says 5 dBi and another says 6 dBi? Is it 5 times more powerful? Is it 10 times more powerful? I've read through the Wikipedia article on antenna gain and I still can't tell for sure. Negative values are even more confusing for people. Ultimately, I think the loss is entirely that of router manufacturers because their preferred units don't help potential buyers judge which is better and by how much. They can use dBi for research and development, but customers really need something more user-friendly. It's a matter of usability.
The gain in dBi is 10 * log10(linear gain), so the linear gain is 10^((gain in dBi) / 10). So 5 dBi is a gain of 3.16, and 6 dBi is 3.98. Log rules mean you can just subtract, so the difference is 1 dBi, meaning a 6 dBi antenna is 1.25 times more powerful than 5 dBi.
I'm not sure what alternative units you could use, because log units are more natural. Joe consumer can understand "6 dBi is 1dBi better than 5 dBi," he doesn't need to know the details any more than he needs to know exactly how much faster his car will go with 10 extra horsepower. It's the relative comparison that matters (more horsepower == more speed) and using dBi makes those relative comparisons simpler.
They use dBi (dB gain related to an isotropic radiator) on commercial devices because it's more accurate by referring to a theoretical antenna having gain of exactly 0. Antennas can't amplify signals; to have some gain in one direction they have to lose it in others, so that the isotropic antenna has a radiation pattern of a perfect sphere and its gain is 0, a dipole which loses gain in directions parallel to its axis shows some gain perpendicularly, a patch antenna uses a reflector behind it so that they get more gain in one direction, and a Yagi one also uses directors to narrow that direction even more, thus getting more gain in that direction.
The downside to using dBi is that being a number related to a theoretical antenna it is higher and welcomed by advertisers: if you take out your router external antenna having a gain of say 2dBi to swap with a higher one that gains 9dBi, you don't get a gain of 9dB, but vendors still can write 9dBi on its box.
For home devices, usually smaller dipoles or longer collinears are used because they are the best choice in a scenario where the user needs to cover his/her house and not the neighbors (just keep the antennas pointed up or down, not sideways).
There are many other kinds such as slotted, grid, helical, sector, patch, Yagi, etc. each one with its best use case. Building them is also fun and cheap.
There are too many things to be an expert in these days to be an expert in all of them.
Understanding log scales is easy. Understanding how a decibel rating on an antenna relates to your network-connected lawn ornament's expected range in meters is a whole different kettle of worms, and bigger numbers are better when they're range but not when they're cost.
http://www.proant.se/en/news/2017/03/01/raspberry-pi-zero-w-...