Go Down

Topic: ATTINY85 EMI emissions (Read 316 times) previous topic - next topic

dustin02rsx

Suppose one was to design a very simple PCB using an attiny85 running on the internal oscillator driving a few 15mA leds. How much is really involved in passing FCC emissions testing? Would you have to purposely design an antenna to radiate enough EMI to fail or is it the contrary and the design would have to be done very carefully by an engineer very familiar with RF to pass? Does the ATTINY85 have any rise/fall time limiting to help with EMI emissions?


Side question, I see in the data sheet the ATTINY85 can have a clock speed from 0- 20MHz, what would be involved in clocking one at less than 9kHz? Could one still generate a PWM signal, and what would be the maximum PWM frequency? Pardon the trivial question, it has been several years since I have manually generated PWM by directly accessing the timers.

ron_sutherland

EMI is complex.

If I were to boil it down to the simplest idea I can think of... well that would be how well matched is the impedance of the circuit that will radiate to the electromagnetic (EM) impedance of free space.  When the two sides are matched up fairly well then the power flows right out as EMI, but if there is a big difference then the power is reflected.

I am guessing that was no help, and I may even be confused about this stuff. The real answer is to study EM until it starts to make sense (and the math to help it make sense). I am not saying you need to spend monies to study by the way.

avr_fred

You've reached the point of analysis paralysis...  Why do you think this type of device might need FCC certification? Does it intentionally emit radio frequency energy?

If all the thing does is control LED's, the answers are no and no.

dustin02rsx

Ron:
Yeah I understand about unmatched impedance transmission lines reflecting (I would assume this would be when EMI would be at its worst but I've never really worried about EMI before) but the only signal part of the circuit where something like this would need to be considered is in the oscillator itself as the 490Hz pwm signal is so slow it shouldn't matter.


avr_fred
Yes you could say that Im in the analysis paralysis phase lol. It was my interpretation that any product being sold in the states cannot emit more than the FCC standard even for a no intentional emitter. The only way to know for sure is to get it tested.

For my purporses it's not feasible to pay to have the device tested and I really can't see what I could possibly do to make a microcontroller blinking some LEDs emit a significant amount of EMI. So I was just curious if there was any major pitfalls just from attaching a microcontroller to a PCB other than don't make a trace shaped like a WiFi antenna lol.


My secondary question basically would make my device excluded from the FCC stuff so I was contemplating doing that just so I wouldn't have to wonder what the EMI actually was.

avr_fred

Serious question: Do you think Sparkfun and Adafruit have every new processor board they design certified for FCC compliance?

westfw


ron_sutherland

#6
Jan 11, 2018, 04:22 am Last Edit: Jan 11, 2018, 04:35 am by ron_sutherland
I guess think of free space as a transmission line, but its impedance changes depending on how much is in contact with the circuit node. Those 500 Hz PWM signals have edges, so they can have some high-frequency content.

The real question is who do you want for a customer? If you want Centrylink or Comcast then they will need the FCC certification for your blinky. Perhaps Walmart can offer it, but I'm not sure on that. If you want to offer it the way Sparkfun and Adafruit do (as development boards) then there is no need to do FCC at least until someone tells you to get it certified or they won't buy it (maybe Centrylink does want it after all).

As far as keeping those high-frequency edges from radiating EMI all you need to do is to slow them down e.g. perhaps drive the LED with a constant 15mA current source so that the long wires (which have capacitance) running to the LED are not slammed with a quick current flow that ends up radiating a lot of EM.

hansibull

I wouldn't worry too much about the microcontroller itself...

Two weeks ago I spent some days at a certification company to test a few products for the company I work for. These products was analog audio equipment controlled by AVR microcontrollers running at 8 MHz. The PCBs are four-layer boards with ground plane on top and bottom

We tested for CE approval and for maritime use. CE approval way quite simple to pass, as it does not require to measure the noise on the input power (conductive noise). The radiated emmision (radiated out from the product) was measured 15m and 3m away IIRC.


Here's a few design rules that will make you probably pass a test like this:
* Use decoupling capacitors for all ICs on the board. 100n should work fine.
* Place the decoupling capacitors as close to the power pin as possible and use a lot of ground vias to reduce the trace inductance.
* Keep your ground plane(s) whole to reduce the impedance
* Switched regulators are the worst, so be careful when designing with these. We use the L5973D in most product, and the switching frequency + harmonics were dominating, even when properly decoupled see attachment). Add an inductor in series to the Vin pin if you're going to measure conductive noise (noise on the power input).  this reduced the noise on one of our products by 20dBuV (dB microvolt)
MightyCore -  ATmega1284, mega644, mega324, mega164, mega32, mega16, mega8535
Github.com/MCUdude/MightyCore

MiniCore - ATmega8, mega48, mega88, mega168, mega328
Github.com/MCUdude/MiniCore

dustin02rsx

Serious question: Do you think Sparkfun and Adafruit have every new processor board they design certified for FCC compliance?

I would imagine their intentional transmitters would have to be. But all of their products fall under engineering tools/ development tools so they should be exempt.


I could possibly in a round about way twist the marketing of my device to call it an engineering/development tool, but thats not really the market Im going for.


I guess think of free space as a transmission line, but its impedance changes depending on how much is in contact with the circuit node. Those 500 Hz PWM signals have edges, so they can have some high-frequency content.

The real question is who do you want for a customer? If you want Centrylink or Comcast then they will need the FCC certification for your blinky. Perhaps Walmart can offer it, but I'm not sure on that. If you want to offer it the way Sparkfun and Adafruit do (as development boards) then there is no need to do FCC at least until someone tells you to get it certified or they won't buy it (maybe Centrylink does want it after all).

As far as keeping those high-frequency edges from radiating EMI all you need to do is to slow them down e.g. perhaps drive the LED with a constant 15mA current source so that the long wires (which have capacitance) running to the LED are not slammed with a quick current flow that ends up radiating a lot of EM.
Yeah, I would be distrubiting through my own means so I wouldnt need to worry about the FCC cert just to get it on shelves so thats good.

Honestly theres a .01% chance this would ever be an issue, but theres always murphy's law.


I may look into rounding the edges slightly with passive components, but adding a constant current source will add quite a bit to the parts cost (its hard to beat a 3 cent mosfet) so I dont really want to do that either.

I may just place the micro as close to the fet/leds as possible to reduce surface area of the pwm trace.

I wouldn't worry too much about the microcontroller itself...

Two weeks ago I spent some days at a certification company to test a few products for the company I work for. These products was analog audio equipment controlled by AVR microcontrollers running at 8 MHz. The PCBs are four-layer boards with ground plane on top and bottom

We tested for CE approval and for maritime use. CE approval way quite simple to pass, as it does not require to measure the noise on the input power (conductive noise). The radiated emmision (radiated out from the product) was measured 15m and 3m away IIRC.


Here's a few design rules that will make you probably pass a test like this:
* Use decoupling capacitors for all ICs on the board. 100n should work fine.
* Place the decoupling capacitors as close to the power pin as possible and use a lot of ground vias to reduce the trace inductance.
* Keep your ground plane(s) whole to reduce the impedance
* Switched regulators are the worst, so be careful when designing with these. We use the L5973D in most product, and the switching frequency + harmonics were dominating, even when properly decoupled see attachment). Add an inductor in series to the Vin pin if you're going to measure conductive noise (noise on the power input).  this reduced the noise on one of our products by 20dBuV (dB microvolt)
Thanks, this is what i was looking for.

My board will only be a 2 layer board, and it will be using power from a usb source so I don't have to worry about any switched regulators.

Is there a specific type of capacitor that is best for IC decoupling?

hansibull

I'm only using 100n ceramic ones. The pictures attached uses an ATmega324P TQFP44 with 0805 sized 100n decoupling. They're placed really close to the MCU, and all ground pads has a via close to it.
MightyCore -  ATmega1284, mega644, mega324, mega164, mega32, mega16, mega8535
Github.com/MCUdude/MightyCore

MiniCore - ATmega8, mega48, mega88, mega168, mega328
Github.com/MCUdude/MiniCore

Go Up