Electronics > Projects, Designs, and Technical Stuff

EMF pickup from amplifier in I2C line causing glitches. (Now with scope trace!)

<< < (23/31) > >>

T3sl4co1l:
Like I said -- it was pretty bad indoors, but the building was probably good enough shielding.

Never underestimate the capabilities of a pile of steel?  :-//

Tim

Starlord:
You can all stop freaking out over interference; I ordered a new prototype with filtering on the amplifier today.  I went with 2.2uH indicators and .1uF caps.  Should attenuate the signal enough that I won't even see it on the scope. 

Wasn't able to do anything about the I2C since nobody could answer me as to whether the ADC could handle higher currents on the I2C lines, not even Microchip.  But since the system worked fine when sitting on my bench and I expect my filter will fix the noise issue, the slow rise times should not be a problem. 

Btw, Microchip's support system is crap.  They require you to register and enter a part number before you can submit a question but their system would not accept any variation on the part number I entered.  I had to tell it I was having an issue with the website instead in order to get it to allow me to submit anything without a part number and aside from a confirmation email four days ago I haven't heard from them.  I should have gone for the slightly more expensive TI part.  In the future, I'll avoid them.

CJay:

--- Quote from: Starlord on July 14, 2016, 03:54:46 am ---You can all stop freaking out over interference; I ordered a new prototype with filtering on the amplifier today.  I went with 2.2uH indicators and .1uF caps.  Should attenuate the signal enough that I won't even see it on the scope. 

Wasn't able to do anything about the I2C since nobody could answer me as to whether the ADC could handle higher currents on the I2C lines, not even Microchip.  But since the system worked fine when sitting on my bench and I expect my filter will fix the noise issue, the slow rise times should not be a problem. 

Btw, Microchip's support system is crap.  They require you to register and enter a part number before you can submit a question but their system would not accept any variation on the part number I entered.  I had to tell it I was having an issue with the website instead in order to get it to allow me to submit anything without a part number and aside from a confirmation email four days ago I haven't heard from them.  I should have gone for the slightly more expensive TI part.  In the future, I'll avoid them.

--- End quote ---

I can answer that, the ADC won't give a damn about the current on the I2C lines, it's not sinking any appreciable current and never will do.

You could make the bus driver capable of sourcing 500 amps and the ADC would still tolerate it as long as the voltage doesn't overshoot.

From your 'scope shots of the signals it looks like whatever is driving the bus isn't capable of providing enough current to charge whatever capacitance there is on the bus fast enough to provide a nice square edge at the ADC bus input.

T3sl4co1l:
AFAIK, I2C pin drivers are good for 74HC-sized currents and voltages.  They pull down only, with a bit stronger strength than regular HC.  Almost as if the complementary P and N channel output transistors were replaced by one N channel transistor of the same total die area -- thus, capable of sinking more current for the same voltage drop and thermal limit.

The logic tests (V_OL @ I_OL) are done at something like 4mA, to guarantee < 0.5V, which is well within the V_IL threshold of 0.8V or 0.3*VDD or whatever it is.  The pin itself will be capable of much more, at half- or short-circuit conditions (probably up to ~50mA), which is where line driving capability comes into play -- it only needs to deliver high current for a fraction of a microsecond, as the line discharges.

So, they don't guarantee that you can operate with more DC current, and no, strictly speaking, it's not something you should extrapolate.  The Rds(on) of the transistor is not a flat resistance, so at a higher current (say 10mA), it may drop more than the expected voltage (i.e., over 1.2V, say).  Which may violate V_IH and then you get dubious thresholds and much more sensitivity to noise...

By the way, about that capacitance -- the line capacitance consists of two parts.  One wire is "ground", and the other is SDA or SCL, right?  The greater part of capacitance is between these two wires.  The lesser part is between SCL/SDA and free space -- each time the output pin turns on and discharges that capacitance, it's delivering current directly into free space!  This is the first part of how the line can radiate EMI.  (I2C isn't terribly fast, so this current isn't very high, but it's there.  The second part is by inductance of the line, and the last part is by more complicated resonant effects that depend on the length and position of the cable.)  By the same method, noise in the environment couples into the line, so that it only takes a few volts of incident noise to start throwing your logic thresholds into question.

Tim

janoc:

--- Quote from: T3sl4co1l on July 14, 2016, 09:51:35 am ---By the way, about that capacitance -- the line capacitance consists of two parts.  One wire is "ground", and the other is SDA or SCL, right?  The greater part of capacitance is between these two wires.  The lesser part is between SCL/SDA and free space -- each time the output pin turns on and discharges that capacitance, it's delivering current directly into free space!  This is the first part of how the line can radiate EMI.  (I2C isn't terribly fast, so this current isn't very high, but it's there.  The second part is by inductance of the line, and the last part is by more complicated resonant effects that depend on the length and position of the cable.)  By the same method, noise in the environment couples into the line, so that it only takes a few volts of incident noise to start throwing your logic thresholds into question.

Tim

--- End quote ---

I think he is only going to see this type of EMI. Right now his edges are so slow, that any interference caused by I2C lines is not going to have a ton of high frequency harmonics, with the parasitic wire capacitance acting as a low pass filter. However, the moment he fixes the drivers on the line and gets nice square edges, the amount of high frequency harmonics will shoot up, with the long cable acting as a nice antenna. Even 100-400kHz I2C can produce harmonics in HF/VHF bands. I hope the cable is going to be shielded.

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod