Doodles from @raphaelabrams #1
Conversation
- simplified the power supply rail to a star - strengthened the ground plane - added several decoupling capacitors
|
Is it possible this didn't commit correctly? To me the board looks like this now, lots of vias with nothing connected to them: I like the idea of using multiple sensors, would have been fun to keep the old one on the board for comparison! I don't think you can use digital sensors with XBees, otherwise I would have since the selection is much better. In sleeping mesh mode the XBees just sample periodically / can't do serial stuff, but can you ask @raphaelabrams if he knows otherwise? If we someday switch to microcontrollers on the cells then we can definitely use digital sensors. |
|
Oh, unless all those vias are what strengthen the ground plane? |
|
Yes, all the vias are there to strengthen the ground plane and reduce noise. He said they're more effective the larger they are, but in some instances there wasn't room so he made them smaller. Oh, he did other things that I didn't mention. He moved some things closer together to reduce the distance the signal had to travel through traces to reduce noise. He increased the size of the traces to reduce resistance and used the miter tool to round their edges. Apparently 90 degree angles are verboten for traces because the angle gets eaten into by the etchant during the wash and becomes narrower than you planned. In some cases it can cause the trace to short, although our traces are probably too wide for that to be a problem. He also tried to make the labels clearer by increasing the font size and positioning them in places that made it unambiguous what they were referring to. He recommended we try out tempoautomation.com for production. Apparently you just send them an eagle file and they order parts, print the boards, assemble them, and ship them in 3 days, all for not much more than Advanced charges, and they might give us a discount if we name drop him. I'm gonna call them tomorrow. I'll also email Raphael your question about using digital sensors with xbees. |
|
Oh also, he pointed out that on the data sheet it says our new sensor is extremely sensitive to light, so we'll have to make sure there's no light on the sensors when they're taking readings. Luckily those project cases are pitch black. |
|
I have quick question. To my thought, I think we are using bottom plane as well as top plane other However, the chip is sensing the temperature by analog voltage from sensor, How much do we need to reduce noise? I think we need to have target.. Thanks, 2015-12-04 0:01 GMT-05:00 William Jeffries [email protected]:
|
|
Excellent point @qelloman, we should have a target. What do you suggest? I'm not actually sure what you even measure noise as. Decibels? What level of noise do other devices have? |
|
I would suggest the degree of noise should be socially accepted since we do If so, we need to provide the temperature data which is accurate enough to Thanks, 2015-12-04 0:29 GMT-05:00 William Jeffries [email protected]:
|
|
Okay so were measuring noise in degrees, that makes perfect sense. I also can't feel a 1°F difference in heat, but the law says that 54°F at night is illegal and 55°F is not, so if we want courts to take it seriously we need to be accurate down to 1°F. Actually we'd want our error to always be <0.5°F so we can round to the nearest degree and always be right. Luckily this sensor is rated to 0.1°F so as long as we don't introduce 5x more noise we should be good, right? |
|
Cool, thanks for all the hot tips @raphaelabrams! I'll leave this unmerged until we have some time to decide on what changes we like and try things out, but it sounds like it all makes sense. I do want to ask him why getting rid of the power plane is good, though. Maybe an RF thing? |
Take these changes or leave them, they are just doodles from @raphaelabrams, the engineer behind goTenna. Raphael recommends looking at digital temperature sensors to further reduce noise. Putting a second or third type of sensor onto the board just in case the first one doesn't work out could be a good insurance policy and would allow us to experiment with multiple sensors without printing more boards.