Power Supply Ground Question

In another forum, fat16lib brought up that he specifically bought a grounded power supply from Phihong to minimize the measurement error /noise he was observing on his 16bit ADC. He confirmed that the ground is a pass-through design, so now his circuit is grounded to the mains ground. He observed much less noise on the board, resulting in better measurements.

I too am putting together an ADC solution with 16 bit resolution though so far my on-board switchmode power supply the VOF6 from CUI, is installed as delivered, i.e. isolated. It only features two input pins, Line and Neutral, and two output pins, 5VDC and GND.

I don’t know if there is a potential for my board to develop a ground loop going forward, but I thought it would be a good design practice to ground the PCB to earth in general. Thus, what are the downsides to grounding a PCB to earth? If one were to retrofit such a ground, would the best approach be to use a 2200pF, 2kV capacitor as suggested at Wikipedia? Would you fuse such a connection in case the outlet is hopelessly miswired or the AC ground develops a high potential for some other reason? Or perhaps use a very thin trace coupled with a 0805 fusible resistor with a low resistance for said connection?

For that matter, is it good design practice to always tie the purported AC ground and Neutral lines together on a PCB or to leave them separate? While I should be able to do this (assuming the outlet is wired correctly), experience in the field has taught me otherwise.

Never ever tie neutral to earth, that's a grave error. Treat the neutral line as live if you want to remain live yourself!

Hi MarkT and thanks for the response.

As far as the Neutral to Ground question is concerned, I agree. For one, I could see how tying the two together could be really fun if someone mis-wired a outlet w/o a ground (i.e. reverse line and neutral) followed by the homeowner using one of those adapter plugs that "converts" a non-grounded outlet into a grounded one - naturally w/o first attaching the ground ring on the adapter to ground first. Any metal surfaces would be "hot" then.

Never mind the fun that would ensue if a dead short is inserted into an mis-wired but grounded outlet if the appliance had the ground and neutral line tied together. So I can see why one would not want to do that.

But coming back to the original question: Is it good design practice to ground a PCB to the AC mains ground or not? Sure, some components might be required to do it safely! But isn't it a good practice, especially if anything attached to the Arduino is grounded as well? From what I recall reading, quite a few power supplies create 'floating grounds' relative to the earth ground, potentials that can cause unhappiness if the arduino is exposed to earth ground (analog input pin, for example).

Additionally, fat16lib noted significant performance improvement re: his ADC after using an earth ground. I presume this has three reasons: 1) He's using a switch-mode power supply which likely has a bunch of harmonics/noise/etc. that older supplies based on linear regulators and transformers would not, if implemented properly. 2) His ADC is very sensitive. 3) His Arduino board power was floating relative to earth ground.

4) Not enough/the right bypass caps?

I've seen in the ATMEL docs where people shut down parts of the chip to get cleaner ADC. I don't remember where but....

I’ve seen in the ATMEL docs where people shut down parts of the chip to get cleaner ADC. I don’t remember where but…

Was that related to sleep? I seem to recall the ADC having some sort of mode related to better precision after the chip waking up… That said, I doubt this is terribly relevant to an Arduino (noise that is) unless you use an op-amp to boost small signals. As is, the best resolution one can hope for from an Atmega 328P, 1284, 2560, etc. is about 1V/1024 or about 1mV. Under normal circumstances (i.e. AREF=5V), that rises to about 5mV.

16 bit ADCs offer 64 times the resolution, so there is much more opportunity for noise to be measured with them (i.e. 0.07mV resolution with a 5V range). Having read some more literature re: noise, etc. I now understand better why differential measurements are as popular as they are.

Unshielded wires of any real length to sensor could be a noise source as well. Make a big enough loop and the wire could be the sensor. That could work in physics class along with a spark gap to teach about waves and wavelength.