Because all of my CRIO modules have the option to connect an isolated power supply, I decided that this was good thing.
I bought 6 (switching) power supplies, in order to maintain the isolation!
I initially had some serious noise problems, caused by "grounded" signals at source (and therefore ground currents somewhere in the system).But I solved the problem by fitting some signal isolators, which gave a dramatic improvement.
However, I still I don't think I have the optimum measurement system!
When the power supply is floating, it's not obvious where to connect the cable screen for analogue signals. I eventually used the 0V line of the appropriate power supply (as NI suggest connecting RTD cable screens to the isolated COM of the NI9217 module, I followed the same principal).
But, now I have discovered that I can reduce my current input noise from 60uA to 12uA on every channel (simply by connecting the power supply 0V line to "ground" (i.e. chassis and earth). And further, when the power supply 0V line is not "grounded" most of my noise is at 50Hz, even though the power supply is floating. When the 0V line is "grounded" the 50Hz noise is effectively eliminated.
Maybe i got my isolation concept wrong from the outset?
Should I "ground" all my power supplies? (if I do "ground" them, I probably don't need so many).
Are floating power supplies good or bad?
Does anyone else have multiple floating power supplies, like me?
I have read lots of "EMC" related material.....but none of these books/documents, mention the use of floating power supplies for instrumentation. They all discuss "grounding" practise and how to connect the screen to "ground", etc.
So maybe that is actually the right thing to do?.....
...(I could go down this road, if wasn't for one tiny problem which I'll briefly mention. In our old system, our welder often damages/destroys control system electronics due to "ground" currents some how returning through the pipework and instrumentation wiring. If I connect my power supplies to ground, I'll probably need to sort this problem using something other than isolation....and to be honest I don't know where to start)
Please share your thoughts/experience on signal isolation and "grounding".
You always should be aware that the noise signals picked up by the cable shield has to be "drained" somewhere. If you connect the shield(s) to 0V of the power supply, the noise signals picked up by the cable shield(s) are not "drained" but fed into the power supply. What happens there depends on the properties of the power supply, especially its output impedance over the noise spectrum.
To avoid any problems with connecting your 0V to chassis GND (and thus connecting all 0V's of all power supplies), you should avoid using the cable shield as signal return path (or reference point for the signal). Shields and signal grounds/reference points should be treated separately. This implies that usually it is MUCH better to use differential signals whenever possible. When using differential signals there is much less risk confusing the cable shield connections and signal (and/or power) ground connections.
However, usually there is no need to use isolated power supplies for each and every signal source unless there are significant cable lengths between the source(s) and the DAQ system. Proper earthing and grounding (yes, these ARE separate issues) is much more important.
Concerning EMC issues you have to consider the whole thing from a "high frequency" point of view. A grounding scheme which works OK for DC or low frequencies may be much less satisfying when EMC comes into play, since inductances of cables come into play. Fortunately, in most cases, the signal frequency spectrum is far below the frequency ranges where EMC is concerned, and you can use a different grounding scheme for DC/LF (low frequency) and HF (high frequency). The latter can be grounded quite easily without affecting your DC and LF circuitry with small capacitors - in the right places.
Take a look at the following document when explains floating and grounded signal sources together with other important noise considerations in your measurement system.
I hope this helps.
I have read the above document many times already, but it is maybe over simplified.
I agree, that connecting the shield to chassis ground makes sense as this provides a path for the noise currents.
But I have found, when the signal is floating and the shield connected to chassis ground, a large amount of 50 Hz noise occurs on the signal.
NI don't mention this in their document, so maybe there is something else wrong in my system or the document is missing some important points.
Here are my measurements:
1. Signal floating and shield connected to power supply 0V (i.e. Signal 0V), causes approx 60uA of mostly 50Hz noise (on 4 - 20mA input)
2. Signal floating and shield connected to chassis ground, causes approx 300uA of mostly 50Hz noise.
3. Signal (power supply) 0V connected to chassis ground and shield connected to chassis ground, causes just 12uA noise with no 50Hz component.
So (3) is the best option for noise level, but I don't want to connect my power supply 0V to Earth directly (because of the welding problem and risk of CRIO damage).
I'm looking for some compromise between solutions (1) and (3).
A colleague suggested using double shielded cable (connecting the inner shield to 0V and outer shield to chassis ground) but this sounds expensive and could be overkill, I'm not sure how effective it will be.
But also I'm wondering whether to fit some passive components (e.g. capacitor or choke) between each power supply 0V and the chassis ground.
And also.. I'm wondering whether to isolate the control box back plate from ground (PE) and fit an Earth line choke (as part of a mains filter) between the back plate and the Earth line input.
Any opinion on this?
Could you please give some more information about your application and set up. Please give details of the environment, how distributed the system is and the hardware and sensors you are using.
Generally speaking, each of your systems should be grounded at one point, whether that be at your signal source or at the chassis is very much dependent on your application. Also, I would suggest in most cases that you bond your chassis to ground. This may also help minimize the noise in the system but very much depends on you set up. As to your questions on grounding your power supplies, this very much depends on whether you need isolated power supplies or not.
Here is some further information on my current configuration, sorry it is quite long.
Power Supply 1: Internal power for the control cabinet
Power supply 1 is not grounded
Power Supply 2: Digital Sinking Outputs
Power supply 2 is not grounded
Power Supply 3: Digital Sourcing Inputs:
Power supply 3 is not grounded.
Power Supply 4: Current Outputs:
1. 4 to 20mA output signal to pump inverter 1. The output signal from the CRIO module first connects to a 3 way signal isolator and the output of the isolator connects to the pump inverter. The output signal from the isolator is grounded at the pump inverter input. I am using screened cable between the isolator and the pump-inverter, the screen is connected to ground at the pump inverter.
2. 4 to 20mA output signal to pump inverter 2. Same arrangement as 1
3. 4 to 20mA output signal to pump inverter 3. Same arrangement as 1
4. Loop powered regulator valve. The regulator valve circuit is isolated from ground. The cable screen is connected to ground at the control cabinet.
Power supply 4 is grounded.
Power Supply 5: Current Inputs:
1. 4 to 20mA input signal from the pump inverter 1. The input signal from the pump inverter first connects to a 3 way signal isolator and the output of the isolator connects to the CRIO module. The input signal to the isolator is grounded at the pump inverter. I am using screened cable between the pump inverter and the isolator, the screen is connected to ground at the pump inverter.
2. as 1 (Pump Inverter 2)
3. as 1 (Pump Inverter 3)
4. 4 to 20mA input from flow meter, the output from the flow meter is isolated.
5. Loop powered pressure sensor (pressure sensor is not grounded). I am using screened cable. The screen is connected to ground at the control cabinet.
6. as 5
7. as 5
Power supply 5 is grounded.
Power Supply 6: RS485 Devices
Power supply 6 is not grounded.
Power Supply 7: RTD power from CRIO module
Power to the RTD temperature sensors, is supplied by the CRIO module.
The RTD sensors are supplied with a short length of screened cable. The sensor itself is not grounded but the cable screen is grounded by the metal housing of the RTD. The short RTD sensor cable is extended with another cable that goes all the way to the control cabinet. The screen of the extension cable is not connected to the screen of the RTD sensor cable. The screen of the extension cable is connected to ground at the control cabinet, where the RTD signals are also grounded. (The NI documentation for NI9217 suggests that the module should float, but I have found much better measurements by grounding the RTD signals.)
I know my arrangement is not perfect…and I have changed my mind many times already. The solution is currently a compromise because of the welding problem. When our welder sparks up, he often causes damage to nearby PC equipment. I believe this damage is caused by ground currents…where both the welding job and the PC are connected to Earth. And….because the CRIO is not cheap; I don't want to take unnecessary risk. My current solution is therfore based on the following decisions:
- Keep CRIO back-plane isolated from PE, i.e. isolate from the cabinet back plate.
- Ground power supplies only where needed to reduce noise.
- Isolate the control cabinet back-plate from the cabinet…..and then connect the system to mains (including earth) using an industrial plug and socket. The control cabinet will then (hopefully) be unplugged before welding, so that there is no current path back to earth through my system.
Your point about connecting the CRIO chassis to ground is interesting and important……as I have currently gone out of my way to avoid it. Now that you have mentioned it, I will definitely try to measure the noise difference. Do you have any further thoughts on my current arrangement? As I still have some time to make changes, before building the 17 cabinets.
Many apologies for the delay in getting back to you. This is quite a complex query that essentially has 6 systems that we need to address. I therefore suggests that we will be able to support this query in a quicker, more effective way if we are to use one of the other support channels. I recommend that you visit the following link and post your query. We will then be able to make recommendation and suggestions for optimising the noise performance of your systems in a much more direct manner.
Please let me know if you have any questions.
I decided to go with my plan as outlined in my previous post.
I tried your suggestion of "grounding" the CRIO chassis.
There was a very slight improvement with the chassis grounded.... but after considering the other points, I decided to keep it floating in my design.
My final DIN terminal layout allows any power supply to be optionally grounded by installing jumper links.
This will provide a quick fix, to any unforeseen problem caused by the floating power supplies.
Thanks for your help.
I have now finalised my hardware design.