4-20mA Signal noise

[Gadgit]

Hi All,

I am hoping you might have some ideas on a problem I've been having with signal noise. I've been searching around for ideas, and have tried quite a few things, but am a bit stumped. I keep reading that 4-20mA is supposed to be less susceptible to noise, which is only confusing me further.

To give you a bit of background, I am the control engineer for a company who are developing a new type of engine, and have recently build a new engine test cell, which is causing me noise headaches. I built our first test cell with what limited knowledge of best practices I had and attempted to separate signals and power, crossing at 90 degrees and using Belden twisted pair shielded cables (only grounded at one end). This first cell has very low signal noise and I'm happy with it. We use a NI CompactRIO for data acquisition and use a mix of thermocouples, 0-10V and 4-20 mA signals. This cell has an ABB 15kW VSD 'dyno' which interferes very little with signals.

The next cell I've built, I went to even more efforts to reduce noise - as it has a 75kW dyno, I was more concerned. I separated high power lines completely, so there are no crossing or parallel runs, I've used enclosed trunking from sensors all the way into an enclosure for the DAQ system, again, all shielded cables etc. However, this setup is giving me significantly more noise!

The noise is not from the VSD, as it's present even with the drive powered off (but gets slightly worse with it on). I see noise on all my signals, but what surprises me most is that the 4-20mA seem to be the worst!

The main differences between my two cells are: A metal enclosure for the DAQ system, this contains two 24V power supplies and one 5V supply powered by 240V AC which also enters the panel (at the top, signals in bottom). Signal cables are run in enclosed metal trunking. The 4-20mA signals are in a single shielded multi core cable (rather than twisted pair shielded belden). The 4-20mA signals run in wall trunking with 240 V (although the noise looks more like white noise rather than AC coupled).

I've tried quite a few things to find the source - mainly going back to basics i.e. running CRIO from external power supply with all modules removed and a fixed 5V (for 0-10V) signals into analogue inputs. I still seem to get more noise than I would expect - and more than my other system. My feeling is that the noise is coming from the environment or power supplies.

Sorry for the long post, any ideas would be greatly appreciated! I've got some screenshots etc. which I can upload if anyone would find them useful.

Many thanks

 

[ScottyUK]

Linear or switchmode power supplies? Switchers produce a lot more noise than linears. That said, switchers are used in industrial instrumentation systems all over the world and they are generally acceptable. Have you got something horribly messed up with the earthing (grounding)?

 

[IRstuff]

I've found that grounding is almost never that straightforward. Invariably, the magical "single-point" ground often just doesn't work, and you just have to play around with the grounding and shielding.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

Of course I can. I can do anything. I can do absolutely anything. I'm an expert!

 

[Gadgit]

Thanks for the responses.

They are switched mode supplies (I believe) one from NI and the other pheonix contact. Both of which are used in both setups, although different power ratings, and not enclosed in the first lab. I have also tried powering from a 3rd supply located outside of the enclosure.

I originally set it up with basically a single point grounding scheme, with power supply grounds, shields, sensor grounds and measurement grounds all linked with commoning blocks. I then experimented with separating each, and additionally linking in the enclose and trunking, and the building earths into the power supplies. None of these schemes appeared to have much effect at improving the noise. But disconnecting the shield does make the noise worse, and I think theres some coupling going on as I have watched a voltage build on the shield with it disconnected.

I get about 10mV p-p on the power supply lines (24V) does that sound reasonable? All my sensors are amplified outputs, and I'm typically getting about 50mV p-p noise on my 0-5V sensor outputs (RS pressure transducers, same in both labs) and about 1mA on the current signals. Which I'm sure is excessive.

Can enclosures or box trunking cause any sort of funnel/noise trap? my understanding was that they should capture and dissipate noise provided they are grounded (probably wrong terminology)

Thanks again for any ideas

 

[danw2]

>single point grounding scheme, with power supply grounds, shields, sensor grounds and measurement grounds

Do you consider the analog input (-) as a ground? Are the AI's single ended or differential?

Are the DC power supply outputs floating or grounded?

 

[Gadgit]

I dont consider it as a ground, but since the measurement are single ended, and the sensors are referenced to ground, I tie the measurement (-) input to ground. Is this correct?

I'm not exactly sure what you mean by floating or grounded supplies. The input is 240V AC with an earth. The power supply cases are earthed. The output is two +ve terminals and two -ve terminals. The -ve terminals are tied to the measurement -input, the sensor GND and the shields. I would assume this makes them grounded rather than floating?

Cheers

 

[MikeHalloran]

I thought 4-20mA current loops were supposed to be completely isolated, i.e., with no part grounded, every component connected in series as if the +ve and -ve terminals were the terminals of a battery.
I.e., the +ve and -ve terminals of the loop supply are supposed to floating, _not_ grounded anywhere.
The primary, AC side, should of course be grounded as is customary.

Someone please correct me if I am misunderstanding this.



Mike Halloran
Pembroke Pines, FL, USA

 

[danw2]

>the sensors are referenced to ground
Probably not unless you tie them to 'ground'. They are probably referenced to the power supply (-).

If it were me, I'd disconnect all AI measurement (-) from whatever the 'ground' is and only have the signal (-) connected to each AI (-) and see what happens.

If there's still noise, I'd disconnect -ve (DC power supply (-) from the 'ground' and let it 'float'. There are vitriolic opionions pro and con about grounding one side of a DC power supply for instrumentation. If permitted, I'll float the DC output.

That would leave the AC supply side, the power supply case and the cabling shields grounded.

 

[GTstartup]

Agree, with IRstuff. Throw out the established 'norms". Try grounding at both ends, each end separately and even if you can get access to the shield elsewhere. play around with it, to get the best result.

 

[djs]

You might try grounding one side of your 24 VDC power supplies. You might be picking up common mode voltages greater than the capacity of your instrumentation. Also if both ends of the loop are isolated from ground, ground one side of the loop. This prevents build up of electrostatic charge on the signal lines.

 

[Gadgit]

Update on my testing - I have just build up a second panel in essentially the same configuration and don't have a noise issue - i.e. power supplies in same casing, with same grounding scheme, but I don't have any incoming sensor wires yet. I don't seem to have any issue with noise here, so I think I can rule out that the power supplies being in the same case are the problem.

@MikeHalloran

You might be right about this, perhaps I'm unnecessarily grounding the -ve side of the 4-20mA signals. I will try removing the ground.

@danw2

You are right, the sensor outputs are referenced to the sensor -ve terminal, this is connected to the power supply ground to give potential difference to the +ve terminal of the sensor. Therefore, I have to connect my analogue input 'COM'/gnd/(-) terminal to the power supply ground since it's where the sensor is referenced to. I don't see how I could make it work otherwise.

I have tried having my DC system floating i.e. not connected to the case, or power supply earth. Perhaps I need to connect the shields to the ground instead of the DC -ve.

@GTstartup

Have you had scenarios where grounding the shield at both ends or in the middle helps? My understanding was that this is a big no-no as it has the potential to move the grounds as referenced to each other. Are the cases where this should be allowed?

Thanks again!

 

[danw2]

> connect the shields to the ground instead of the DC -ve.

Yes.

 

[IRstuff]

"Have you had scenarios where grounding the shield at both ends or in the middle helps? My understanding was that this is a big no-no as it has the potential to move the grounds as referenced to each other. Are the cases where this should be allowed?"

Sure, particularly when SOMEONE violates the grounding rules and you're stuck with the noise that they allowed into the system. Should you do it from the get-go? NOOOOO! A properly designed system that obeys all the grounding rules should not have anything like that. That said, sh!!t happens; there may be excessive coupling from the circuitry into the grounds that was unintentional and unforeseen. I've seen otherwise well-designed systems that just have so much current or voltage running around that some of it leaks into the chassis and have to be dealt with.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

Of course I can. I can do anything. I can do absolutely anything. I'm an expert!