I recently helped out debugging the Wifi on another AR2317 router project. This is the same SoC chip as the Mesh Potato, so I was revisiting the same Wifi debug territory as 18 months ago.
Turns out this board had the same major bug as we found in the Mesh Potato – the reference clock crystal was off by a few kHz, detuning the radio. Easy fixed by changing the master 40MHz crystal.
The AR2317 reference design includes some short (5mm) antenna feed lines that are meant to be 50 ohms impedance. These are the tracks leading to J4 and J5. The diversity (receive only) antenna is the coat hanger shaped structure on the right above J5. The impedance of these tracks is set by the width and height above ground plane.
Here are the antenna feed tracks on the MP01:
and for comparison the new design:
On the new design, I noticed that the antenna feed line tracks were not the same width as the reference design, in fact they looked kind of skinny. This means they are not 50 ohm lines. The new design was also a 4 layer PCB, so the track distance above ground is different to the 2 layer AR2317 reference design. However this new design passes calibration just fine, which means it meets the AR2317 reference design specifications. So those skinny tracks, over these short distances at least, have a zero or trivial effect on performance.
This puzzled me, as I thought that careful feed line performance was really important at these frequencies. Then I remembered when we were designing and debugging the MP, our RF guru Jeff Wojtiuk said that close in to the chip, impedances can be treated as “lumped”, and transmission line effects don’t come into play. Turns out he was right. Makes me really appreciate Jeffs advice on the MP RF design and debugging. He had the experience to understand what matters and what doesn’t when it comes to microwave design. Without experience it’s very easy to over-engineer the trivial, and miss the important but subtle bugs. The learning experience is one reason I really like working with experts.
Open source and blogging is also like that – the expert knowledge gets propagated. You get to learn from experts.
However the line between lumped and transmission line behaviour still confuses me. Perhaps it’s related to the wavelength, about 120mm in air at 2.4GHz. Perhaps the presence of a good ground plane has something to do with it. Maybe I have to think in terms of Magnetic or Electric fields.
In some cases microwave circuits can be very sensitive to small changes in physical dimensions. For example the PCB antennas we designed had measurable differences between 17 and 20mm versions of the same antenna. A few mm mattered. When I incorrectly terminated a PCB antenna the effect was very obvious.
If anyone can shed more light on why these skinny tracks work OK in this case and when to treat circuits as lumped or transmission-line I would love to hear from you.