12 replies to this topic

[Robert L.]

I have been working on an external watchdog circuit using the MAX6751 chip. For several frustrating days, I could not get the chip to work, it constantly held the /reset line low. Naturally I suspected my circuit design, then I suspected a bad chip, but eventually I figured out the problem was a very noisy 5 volt power line coming from the N+ board. This line is so noisy that the MAX chip thinks the voltage has dropped so low a reset is needed. I also noticed that the noise increases substantially when the N+ makes connections over the Ethernet cable, or even receives messages over that port. I have added 40uF of filtering and that "almost" fixes it. So my message is, if you plan to built an add on board, and you plan to use either the 3.3V or 5V supplies from the N+, plan on adding filtering. I also wonder if this noise affects the ADC circuits within the ARM. I have heard whispers that the conversions suffer accuracy issues when using the Ethernet.

[Chris Seto]

It is standard practice to follow the app circuit in the chip datasheet for information on Vdd decoupling. I have not observed an excessive amount of noise on the Netduino's linear supplies.

[Chris Seto]

It is standard practice to follow the app circuit in the chip datasheet for information on Vdd decoupling. I have not observed an excessive amount of noise on the Netduino's linear supplies.

[Mario Vernari]

Good idea to analyze the noise on the supply lines. Once at home, I'll check them out. Cheers

[Robert L.]

It is standard practice to follow the app circuit in the chip datasheet for information on Vdd decoupling. I have not observed an excessive amount of noise on the Netduino's linear supplies.

Using a Tektronics MSO 2012 (100 MHz) scope, I observed over 1 volt of noise on the 5 volt supply, all of it negative going. I am assuming the actual noise is greater than what I observed due to the limitations of my scope's bandwidth. Also the MAX chip should ignore the noise unless it is greater than that. I observed that the noise disappears if the reset button is held down, indicating to me that the CPU and/or the Ethernet/PHY chip is creating the noise by requiring different current draws as it runs.

I am assuming the 10uF caps on the Netduino Plus's linear supplies are not large enough to handle the short term spikes in current demand. What noise levels have you seen?

[Stefan W.]

Larger is not necessarily better. Try wiring a ceramic capacitor (100nF or so) in parallel to your electrolytic cap.

[Chris Seto]

Over 1V of noise(!!) ? If you mean that the amplitude of the noise exceeds 1V, something is very very wrong.

[Stefan W.]

The mobile version of the forum won't let me edit my post, so I'll use a new one: the reason for this are the high ESR values of electrolytic caps at high frequencies (they are "sluggish"), while ceramic caps have low ones.

[Scott Green]

Using a Tektronics MSO 2012 (100 MHz) scope, I observed over 1 volt of noise on the 5 volt supply, all of it negative going. I am assuming the actual noise is greater than what I observed due to the limitations of my scope's bandwidth. Also the MAX chip should ignore the noise unless it is greater than that. I observed that the noise disappears if the reset button is held down, indicating to me that the CPU and/or the Ethernet/PHY chip is creating the noise by requiring different current draws as it runs.

I am assuming the 10uF caps on the Netduino Plus's linear supplies are not large enough to handle the short term spikes in current demand. What noise levels have you seen?

Robert,

I saw the same thing with both my 5V and 3.3V lines coming from my netduino+. I switched over to powering the netduino with a wall wart, and added dedicated 5V and 3.3V regulators to my circuit supplied from the VIN line on the netduino. Smoothed the supply voltage right out.

Scott...

[Mario Vernari]

I have measured right now both the +5V and the +3.3V: the Netduino Plus is USB powered only. No Ethernet cable plugged, and almost no load over the minimum hardware on board. On the +5V line, there are both a flicker noise, and a ripple sine-wave (oddly shaped). The sine runs stable on about 500Hz and 20mVpp. The flicker is much lower: hard to measure. On the +3.3V line the noise is below 1mVpp, so fully acceptable (excluding for using it as Aref). The caps suggestions are absolutely correct, but I'd believe much more the suggestion of Scott, to change power adapter. I guess the noise could be generated by the PC, and carried via USB. I don't know how much could do a cap alone. Maybe adding some chokes... Anyway, it was a good occasion to test the purity of the supply lines! Cheers

[Robert L.]

Larger is not necessarily better. Try wiring a ceramic capacitor (100nF or so) in parallel to your electrolytic cap.

Thanks Stefan, I know I did not mention it above, but I did use two caps as you suggest. The "small" one did not seem to make much difference. I guess the good news is I finally got the MAX6751 chip to work by putting a small resistor (100 ohms) in line with the 5V supply and then a cap to ground. This forms a low pass filter which suppressed the noise enough to get the MAX chip working.

Strangely, since I made this change and added the watchdog chip, the Netduino Plus seems to be much more reliable while doing IP operations. I am doing a long term test, but I am wondering if the lockup I was trying to recover from, IE socket.connect() freezing all threads might somehow be related to the power supply noise. Even with the resistor, my filter is reducing the noise on the 5Volt line back on the N+ board.

I am powering the N+ using the USB cable, but I also tried a wall wart. The wall wort did not seem to be a whole lot better. Or perhaps I should say it was not quiet enough to let my watchdog chip work properly.

[Mario Vernari]

By connecting the Ethernet cable the noise get higher, and reach 50-100mVpp. However it seems far from the 1Vpp. Sorry, Robert, where do you point the leads of the scope?

[Robert L.]

By connecting the Ethernet cable the noise get higher, and reach 50-100mVpp. However it seems far from the 1Vpp.
Sorry, Robert, where do you point the leads of the scope?

I connected the scope probe's tip to J6 pin 3 to read +5v and the probe's ground clip was on a ground connection on my add on board about 2 inches away from J6 pins 4&5. The probe was set in the 10X mode. My add on board was depopulated except for the MAX6751 to try and eliminate my board adding to the noise. If others are seeing less noise, perhaps there is just an issue with the N+ board I am using.

My overnight test of "socket" ran pretty well with the extra filtering. I was setup to connect to a web server every 60 seconds and send a single message with a POST transaction and then process the reply. My web server is recording the transaction in a log file so I can review the results. I experienced 2 failures, after 6:38 hours:minutes and 2:13 hours:minutes. My watchdog circuit rebooted the N+ in both cases, and the board was back up in around 60 seconds in each time. This is the best result I have seen so far. I plan to allow the test to continue for several more days.