I’m trying to make a very small UAV board with all my neccesary componets on a single compact board that includes micro sd card, gps, some im sensors, and a radio. I read a few appnotes regarding my ADC (mcp3208) and it recommends the use of a ground plane and suggest that I create two seperate planes of analog and digital circuits. Currently all my analog componets lie underneath a xbee xcs 900 mhz radio :? and near a digital pressure sensor , will this have a large affect on my measurments? I really don’t want change my layout completly but it seems like I may have to.
Will a ground plane be sufficent enough to stop any interference and do I really need to use two ground planes?
And for each (or one) ground plane are there certain rules (I believe one is not to make many small pieces)?
I attached my board. The top left is a 3.3 LDO regulator. In the middle-left are the sensors, an adxl335, one bmp085 (these two are on the bottom plane) and two ixz500s (these two are on the top). Right above the ixz500s is the xbee module. To the right of the sensors is my ADC. Top right is the micro sd socket and the bottom rows are for servo inputs/outputs.
First off, that’s an absolutely beautiful board, very nice work!
EMI / EMC (Electro-Magnetic Interference / Electro-Magnetic Compatibility) is a very interesting (and often frustrating) topic. People who are good at it are accused of witchcraft by their peers and rewarded handsomely by their employers. You’re doing the right thing by going through the application notes and picking up hints and tricks to minimize problems. And these problems are always easiest to minimize at the design stage; trying to fix a problem in a completed board is much more difficult.
There are a lot of general techniques that can be used to minimize EMI/EMC issues, but they’re not always practical, especially when working on a small board. But here’s a short (and I’m sure incomplete) list of things to do if possible:
- Note the difference between a ground *pour* and a ground *plane*. A pour is putting copper in the unused areas of a signal layer. A plane is a whole layer (usually internal e.g. a 4-layer board) dedicated to power or ground.
Do use ground pours (good), and if you can afford it, internal power and ground planes (best). EMI/EMC is conducted in and out of circuitry that acts like an antenna. Think about the “loop” that power makes as it travels through the traces to a component, and then back to the source as ground. The bigger the area this “loop” encircles, the more it will couple EM energy in and out. Ground pours and planes reduce this by letting the ground current find it’s own way to travel directly under the power current, making the loop as small as possible.
- Speaking of antennas, trace lengths that are exact multiples or fractions of the wavelength of your transmitter, will couple some RF energy into your circuitry. But this type of analysis is more NASA-level than most people normally want to deal with.
- If you can, put your analog stuff on one half of your board, and digital on the other half, with separate ground pours/planes for each. If you can't do this, having one ground pour/plane for both analog and digital is better than no ground pour/plane.
- Put a grounded shield (Faraday cage) over especially noise-sensitive (or noisy) circuitry. You can even do this on a completed board with copper foil, taking care not to short to underlying conductors (this may be a good idea for your analog components under the transceiver).
- Make liberal use of decoupling capacitors. Each chip should have an 0.1uF cap as close to it as possible, and there should be a 10uF cap where power enters the board.
- If you don't have a power plane, route power as much in a "star" manner as you can, rather than daisy-chaining. Do this especially when mixing analog and digital circuitry (at least route separate lines from the supply and big decoupling capacitor to the analog and digital sections).
- Switching DCDC converters are very efficient, but create electrical and magnetic noise. Linear converters are less efficient, but run very clean. When reducing analog noise is critical, some people use a separate linear converter just for the analog section.
- When connecting multiple boards together, "ground loops" are a common topic, with different opinions on how to deal with them. My philosophy is to star discrete power and ground lines to each board, avoid grounding each board to the chassis, and run ONE ground line to the chassis to ground it. This hopefully avoids the possibility of ground current finding its way back through the chassis (which makes a very large antenna loop). Antenna lines are an exception; the shield portion of an antenna line should always be well-grounded where it exits the chassis.
Other people condone grounding to the chassis at as many places as possible, but to me this increases the risk of a proper ground wire breaking, but things still looking like they work, only with a huge ground loop in the process.
That’s my quick list, any other tips from the SparkFun elite?
As for your board, I don’t personally have enough experience with the Xbees to say whether you’ll have problems or not. Do Google for other people’s experiences with Xbee and nearby analog components. Personally I’d go through the above list and if something is easy to implement on your board I’d do it (like some copper tape under the Xbee) but I wouldn’t do a full redesign unless I read someone’s horror story. Good luck and let us know how it goes!
I’m trying to make a very small UAV board with all my neccesary componets on a single compact board that includes micro sd card, gps, some im sensors, and a radio. I read a few appnotes regarding my ADC (mcp3208) and it recommends the use of a ground plane and suggest that I create two seperate planes of analog and digital circuits. Currently all my analog componets lie underneath a xbee xcs 900 mhz radio :? and near a digital pressure sensor , will this have a large affect on my measurments? I really don’t want change my layout completly but it seems like I may have to.
Will a ground plane be sufficent enough to stop any interference and do I really need to use two ground planes?
And for each (or one) ground plane are there certain rules (I believe one is not to make many small pieces)?
I attached my board. The top left is a 3.3 LDO regulator. In the middle-left are the sensors, an adxl335, one bmp085 (these two are on the bottom plane) and two ixz500s (these two are on the top). Right above the ixz500s is the xbee module. To the right of the sensors is my ADC. Top right is the micro sd socket and the bottom rows are for servo inputs/outputs.
Comments appreciated.
Can't look at your board, it's some proprietary format (Eagle I'm guessing). Attach it as a JPG if you actually want people to be able to look at it. The separate planes have little to do with stopping interference from external sources. The purpose of the planes is to get rid of digital noise on the power supply for the ADC. This is an issue for high resolution ADCs, less so for something like a 12b ADC like yours, but still good to keep in mind. The problem comes from the SPI interface (or any serial interface). When you're talking to your ADC over SPI at, say, 5MHz, you're injecting 5MHz into the ground lines for the chip. PSRR at 5MHz is typically not good at all. Thus that noise can be a significant problem and show up in your conversions. If you provide a separate path for the digital ground current you can improve this significantly.
, will this have a large affect on my measurments? I really don’t want change my layout completly but it seems like I may have to.