cRIO noisier than cFP

[Sholyoake]

We recently purchased a cRIO system for data logging in our laboratory facility, primarily to log 4-20mA devices, strain gauges and 0-10V output devices, we are using a 9203 card for current input, a 9237 for strain gauges and a 9205 for voltage measurement

We already have 3 Compact FieldPoint systems which have been running for the past 5-10 years

After commissioning our new laboratory facility, I installed the compact RIO and connected up a test rig we had moved from one of the old labs. My first observation is that the data quality is much noisier and less precise than we are used to with fieldpoint, working with an alliance member partner we managed to improve the data quality somewhat by enabling differential signaling on the voltage channels and increasing the excitation of the strain gauges, this got those channels to the point of being useful, however the data quality is not remotely similar to that of the cFP based systems

We have tried shielding the mA cabling which improved things slightly but caused ground loops and interference with other components in the system

We are looking to loan out a 9207 card in the hope that the inbuilt mains filter will improve data quality, however this does not explain the higher noise levels experienced with the strain and voltage cards

We have moved the entire rig into an older lab alongside a cFP system so the conditions are the same and the data is no less noisy

we spent a large amount of money on this system and were hoping to begin replacing our fieldpoint-based systems in future, however as it stands we cannot justify the outlay for a system which will produce data of poorer quality

has anyone else experienced this when transitioning? is there anything else anyone can suggest? the cFP has never required cable shielding and has been a robust workhorse for the last 10 years!

[Henrik_Volkers]

OK, I scry with my crystall ball on the hornet's nest....

What I have seen in the past: Different analog input filters,  ADC technologies  (both: newer usually lead to higher bandwidth) , input impedances and sample rates  lead to 'noisier' signals.

Mostly the noise is always there, but was just filtered ....

If you take 1 sample a second , the actual sample time could be anything between 1000ms (dualslope) and 10ns (fast, maybe multiplexed  ADC) wich could (and usually will) lead to huge differences in the noise.

And different grounding schemes cause different ground loop currents. For this case I like to do tests with a isolation 'amplifier' you get for the industrial 4-20mA /10V/RTD/straingauge.  

[Sholyoake]

I would be more than happy for the noise to be filtered.... it just isn't for some reason!

When I've spoken to NI tech support they just sort of say "I can't understand it, RIO should be better"

the problem is that working with precision measurements, it's often impossible to discern an actual change because the data is swamped by the noise.... I understand new style hardware may be causing the issues, but it's just not fit for our purposes as it stands!

[Sholyoake]

your point about sampling time is a very good one, I've queried it with the alliance partner to see if they can shed any light on the sampling

[Logan_K]

Do you have any detail on the equivalent cFP modules that were used for the same cRIO measurements? It would be worth confirming that the grounding is the same between the solutions. On the cRIO side, the voltage modules are isolated, which could be a problem if the original cFP was not isolated, and your source signals were relying on a ground reference to be supplied by the device. However if the the cRIO devices are a similar architecture to the cFP, then I would agree with Henrik's assessment that the issue is likely because most of the cFP modules were low bandwidth which typically had at least 3dB of attenuation on 50hz/60hz signals. Since the cRIO devices you're using don't have this filtering, in order to get comparable results from cRIO you would need to oversample and average the readings to minimize the impact of the noise.

[Sholyoake]

Hi Logan,

we used the AI-111 (4-20mA), AI-112 (0-10V) and SG-140 (full bridge strain gauge) cFP modules

I'm definitely thinking there may be an issue with a filtering effect on the older cards, the main problem we have is that to be able to retain the 'plug and play' functionality of the cFP on our cRIO system we're using the scan engine so bandwidth is limited for the purposes of oversampling, we also have an alliance partner writing our software, they are investigating this very issue at present

I'm just setting up a 9207 card for testing, it has an inbuilt 50/60Hz mains filter so we're hoping for improved results

[JarleEkanger]

Just a quick tip (that might be kicking in open doors). We tried, for convenience and economy, to use a single power supply for both a cDAQ chassis and a NI9203 module, and found that this introduced significant additional noise. By just supplying the NI 9203 (or more specifically the sensors connected to it) with a separate power supply, noise levels were reduced to about 30% of the case with shared power supply.

[Sholyoake]

thanks for the reply

we've been testing an ni-9207 card in the system for the past couple of weeks now and the data quality has massively improved, it appears that many of the modules have 'high speed' and 'high accuracy' setting, with the high accuracy setting, sampling times are significantly increased, leading to a better quality signal

we're looking at exchanging our 9203 cards for 9207/8 cards in the near future, we've also found that exciting the strain gauges at 10V tends to give a better S/N ratio and improve the signal quality, although unless your bridge resistance is very high, you will probably need an external supply to achieve the full 10V across all 4 channels

we're getting there, thanks to everyone who has responded, your input has all been very useful!

[Henrik_Volkers]

---

@Sholyoake wrote:

.., we've also found that exciting the strain gauges at 10V tends to give a better S/N ratio and improve the signal quality, although unless your bridge resistance is very high, you will probably need an external supply to achieve the full 10V across all 4 channels ...

---

If you use piezoresistive bridges check the valid (and calibrated)  supply voltage range!