2.4GHz module to SMA connector

[DiBosco]

pcb lambda ghz

Folks, I am designing a low volume product that I need to make (point to multipoint) wireless. I have found some excellent little 2.4GHz modules from Radiocrafts that are surface mount, which I have put on a four layer PCB.

The PCB will be inside a metal box and will therefore need to have something like a whip antenna. From doing lots of searching on Google and a wee apps note from a disti it would appear that I need to make my track from the module 50 ohm. I was pointed to Agilent's appcad which to my mind seems to suggest I need a 10 thou track on the top side of the PCB with a ground plane on the bottom. This seems in direct contradiction to other guides I have found which suggest the track width should be 1.8 times the thickness of the PCB. This would suggest a 100 thou track, which seems crazy when you consider the pad on the module is only 40thou!

The way the PCB is at the moment, the track from pad on the module to the SMA is only around 500 thou and has four forty-five degree bends in it. When Atmel release a smaller version of the processor to free up space on my board I think that could come down to 250 thou and be in a straight line.

So, could someone clarify?

1. Whether I might be misunderstanding the appcad or whether 10thou seems lie a suitable track width (I will create a ground plane on the bottom of the PCB and have the two middle layers empty for this part of the PCB if that seems right).

2. Are those forty-five degree bends an absolute no-no or whether they would just reduce range for the prototype or whether they make no difference?!

3. Am I just unduly worring and for such a short distance from module to SMA connector it makes no odds?!

I am a very experienced design engineer, but haven't done any RF before.

Many thanks indeed for any help.

Rob

 

[dipnirvana]

45 deg is no good , avoid it . Also length if small, it doesn't matter much. thumb of rule is you don't have to worry about it if length is maller than (lambda/20).where lambda is wavelength of RF you are doing.

 

[DiBosco]

Many thanks for that, I will change the PCB to takw out the bends. as it's a prototype I have some leeway on the form factor.

I'm still not clear about track width though. Does that follow the same rules as length in that if it's shorter than lambda/20 it also doesn't mater too much?

Thanks again. Rob

 

[kspalla]

You are right on
length in that if it's shorter than lambda/20 it also doesn't mater too much

Track width calculation by appcad should be fine.
How ever if RF Path is having any vias then it is properly modeled using 3D simulators.

 

[smiling84]

there is an alternative to 45 degree bend -> you can go for an arc bend which is efficient compare to 45 degree bend particularly in RF pcb design.

Remaining questions i am not sure.