PCIe-6353 - Why so many AI GND?

[MrMattPayne]

We use many 6353/6363 cards to read as many as all 32 RSE analog inputs, but we generally only wire up one AI GND per connector. Why are there so many? Should we be using more of them? It seems there are 16 total, 8 per connector. We don't generally have a noise problem except for one setup that appears to be experiencing some unexplained intermittent noise on some channels, so I'm wondering if using more AI GND's would help those channels and we've just been doing it wrong all along.

Side note: can I seriously no longer ask NI this question without a "service plan"?! I am extremely displeased with how NI is doing business lately. I am their last defender in this company...or I was...

[ZYOng]

In our lab, we convert pairs of AI and GND from a terminal block into BNC connectors. Each pair of BNC use up an individual AI GND so that we don't have to clump all 16 wires into a single AI GND pin.

If you have a noise issue, I would recommend you refer to Field Wiring and Noise Considerations for Analog Signals

[MrMattPayne]

---

@ZYOng wrote:

In our lab, we convert pairs of AI and GND from a terminal block into BNC connectors. Each pair of BNC use up an individual AI GND so that we don't have to clump all 16 wires into a single AI GND pin.

---

So you use 1 AI GND for each pair of AI lines?

We have our sensors all grounded to a terminal block which itself is connected to the - side of the DC power supply. One AI GND from each connector/cable of the DAQ card is also connected to this terminal block (so we are using 2 AI GND total for a 32-channel card). Our thinking is apparently that all AI GND are ganged together, so all the sensors and the card are fully grounded by simply connecting 1 AI GND terminal.

Is our thinking correct? In other words, are all the AI GND terminals connected and redundant? Or is there a reason there are so many and we should be using one for each pair of AI channels?

The trend I'm seeing in this one setup is that the 6 AI channels in use on Connector 0 do not experience the noise that the 14 AI channels on Connector 1 are seeing. When looking at the wiring, Connector 0 is wired with 1 AI GND, 1 AO GND, and 1 DI GND all grounded to the common ground/- terminal block. Connector 1 only has a single AI GND wired. I understand that all the different GND's are somehow connected internally on the card, so it almost seems that connecting more GND terminals (even though they are not all AI GND's) may be reducing the noise in Connector 0 while Connector 1 remains susceptible due to only having 1 GND total connected.

[santo_13 · Accepted solution]

There are so many AI GND for noise immunity. To get the best performance, you must use all AI GND properly and not use it for electrical grounding through just one of the pins. Use all the AI GND for proper electrical shielding.

Electrical connectivity perspective, all GNDs, AI, AO, DI, DO are all internally connected. But those are kept separate to keep the noise at bay. To learn more about keeping GND separate, you can look up analog and digital grounding considerations for PCB layout.

[MrMattPayne]

Thank you! That was exactly the information I was looking for. I will commission a re-wire and re-test to see if this solves, or at least reduces our noise problem.

[MrMattPayne]

After wiring all AI GND's to the negative side of the DC power supply, all channels now seem to be all but immune to the noise we are experiencing. We did still have one channel experience problematic noise once out of some 2,000,000 AI reads, but that's worlds better than most channels failing dozens of times over a few thousand reads. We're still not sure what the source of the noise is or how it's even getting picked up by the DAQ system (we suspect it's the environmental chamber's refrigeration components, even though it's on a totally different building circuit), but the moral of this story is that properly grounding the NI hardware is important.

[crossrulz]

---

@MrMattPayne wrote:

we suspect it's the environmental chamber's refrigeration components, even though it's on a totally different building circuit

---

A compressor will put out a lot of electromagnetic noise, especially when it turns on.  Any amount of shielding you can add to your wiring will help.

This is also a good read: Field Wiring and Noise Considerations for Analog Signals 

[MrMattPayne]

---

@crossrulz wrote:

---

---

A compressor will put out a lot of electromagnetic noise, especially when it turns on. 

---

True, but the odd thing is we've never experienced any on this chamber or any other with identical DAQ systems. I haven't seen it in action myself, but the reports are that it occurs only when cooling and it definitely coincides with the operation of a solenoid somewhere. They hear the click and can see the noise on the screen (LabVIEW program showing live data). I suspect something is going bad as this is an older chamber. I've had a problematic fridge at home trip GFCI outlets on the same circuit when it would turn on. Still seems weird to me that it would only happen in the cooling cycle since the same components are used in the heating cycle, just with reversed coolant flow, no? I wonder what the difference is.