Arduino spuriously resets with power draw

Question

I have a simple circuit, pictured below, that uses an Arduino to control two large bidirectional motors via 12V relays. Normally this works a treat, but sometimes the Arduino ends up resetting itself.

I think what's happening is that the motors eat up a lot of power when starting, causing the arduino to have a momentary loss of power wherein it resets. To remedy this, I added a 1000uF electrolytic capacitor, and a 22uF ceramic capacitor. Additionally, I "walled off" the arduino's power/ground planes with 1N4001 diodes so that the motors aren't able to steal charge from the arduino during a power drop.

Looking at the VIN/GND to the arduino, I still see the occasional power drop, but with the diodes/capacitors the spikes are very short and sharp. What's surprising is that these power drops don't always precede a restart as far as I can tell.

My current working theory is that the power drops are corrupting memory or just precipitating instability in the arduino, which is why power drops don't generally immediately cause a restart.

I've spent a few days struggling with this problem, and I'm still completely lost as to what I should do to fix it. I don't understand why there would be sudden loss of power on the arduino when it's walled off with diodes. Any and all help would be appreciated.

Edit 1 Here's the schematic for the motor control. Ambiently relays are on GND, only one is on during a movement. Looking at the scope, I do see some spikes on the gate. I added 10uF ceramic capacitors to the gate lines, but that did not completely fix the restart issue.

And here is the schematic for the power going to the Arduino. I have a diode on the high side and on the low side. The high side diode is to prevent the motors from stealing charge from the capacitors, and the low side capacitor is to prevent the motors from moving the ground plane. The issue persists with or without the low side diode in place.

I don't think that the issue is from the power side, I can unplug the power and the arduino continues to run for roughly two seconds. I've spent multiple days on this, and I have not found anything that makes this circuit reliable. I'm at a loss for what to try.

Edit 2 I think that the issue is that the relays sometimes arc, which causes EMI that interferes with the arduino. I need an RC snubber to eliminate that, but how do I properly connect such a snubber on my circuit?

Answer 1

Your N FETs are upside down.

Should be:

^{simulate this circuit – Schematic created using CircuitLab}

Answer 2

With the schematic I see you have diodes in both power and ground.

The one in the ground side is not necessary and is probably complicating how you measure any disturbances because your scope will also be connected to ground. Also it would be better to use a Shottky diode as it has less voltage drop.

It would be useful to have a scope trace showing both the motor power and the Arduino power simultaneously. expand the trace so we can see the width of the pulses to determine the likely cause. (it is better if possible to do a screen shot using the scope rather than taking a photo, it will be clearer).

If the width of the dropout is consistent with the time constant of the large capacitor and the drain of the processor then it is being caused by a power drop-out, if it is very short (nanoseconds to microseconds) then it is likely interference in some other way.

It may be worthwhile setting the brown-out-detector(BOD) voltage to lower than normal (say 2.7V) to give more margin if the voltage does drop.

A number of years ago I designed a similar system. In order to avoid the voltage drops caused by the heavy starting motor load I stepped up the battery voltage to 6V with an LT1949 boost converter, stored energy in a large 2200uF cap) then regulated it down to the 5V needed by the processor (ATMega128). The extra 1V stored in the capacitor allowed it to drop by up to one volt without affecting the supply to the processor, and another 500mV or so before getting to the minimum voltage requirement for the processor. This is three times the margin you get by just putting a large capacitor on the supply.

That system allowed the battery voltage to drop down to about 3V without trouble and could survive 10s of milliseconds with the voltage even lower.

Answer 3

I added an RC snubber around each relay Vcc / GND side, so 4 RC snubbers in all. This seemed to fix the issue.

I believe what was happening was that there was an arc within the relay, which caused EMI that would reset the IC.