4-20mA Distance/Resolution Figures

[hyprfrco]

I have experience in 4-20mA loops -both active and passive- for fields instrumentation in process plants sensors - level, pressure, temperature, general devices, etc.- Normally the acceptable error and sampling time here are low... about 1% of error and 1 Hz sampling time.

For a structural monitoring application, i am requiring a single DAQ with 4-20mA loops of about 200m distance each. They will measure accelerometers and displacement sensors, mainly. The standard on this application is having high sampling rates (about 250-1000Hz) and very low errors (less than 0.01%). Because of the diminished size of structural/mechanical excitation (from earthquakes/seisms or environmental perturbances), the standard is to have a minimal (or known) signal to noise ratio/error/precision. Known built in systems exist (i.e. Kinemetrics,

http://www.kinemetrics.com),

but they are unnecessarily high cost for buildings (i.e. you don´t need NEMA 4 in a building).

My question is, which are the typical signal to noise ratio/error/precision figures you could obtain from a properly 4-20mA loop setup?. How you can estimate it?.

As per in the standard resistance cable on the 0-10V loops, here the key factor is the capacitance of the cable, which imposes a concrete limit -discarding factors as ground looping, power source unstabilities, among others...

Thanks in advance.

 

[djs]

The biggest issue would be the receiver or ADC. Precision and speed usually do not come together or if they do it is very expensive. Lowering the signal to noise ratio is just being careful with shielding and grounding. Current loops are low impedance so have a high immunity to noise. On the other hand because they are current driven, wire inductance and capacitance will affect the signal response time.

Typical high end PLC 4-20ma AI modules have a 0.1% accuracy. However max sample rates are about 100Hz.

 

[danw2]

My last 25 years have been in the process field with 1-10Hz sampling rates, so my experience from the early '80's with DAQ at higher speeds is way out of date.

Given the range of AI's out there - 10 to 16 bit A/D, single ended to fully isolated, I don't know how a "typical" S:N could be stated.

My thought is that to get a system signal to noise value, you'll have to measure it. Put a solid mA signal in at the field end (can the transducers be put into a held static state at a fixed output?) and see how much jitter you get at the receiver end. The problem with using a battery powered floating calibrator or a battery and a resistor is that neither shows the inherent instability of the transducer and its power supply.

I have no idea whether it is typical or not, but the unit I was working with last week has a factory spec sheet 4-20mA accuracy spec of 0.2%FS. With a calibrator attached at the input, 4.4000mA input gave a 4.4002 or a 4.4003 reading. When I connected the calibrator at the field end, the reading was identical. There is a cal routine to zero any error out (probably the dropping resistor), but given the application, I didn't bother. Rock steady, no noise. Belden 8760. The low pressure DP sensor would jitter about plus or minus 0.0002mA at zero DP.

Without looking up all the capacitance values, I'll speculate that shielded twisted pair like Belden 8760 (18g) or 8762 (22g) will not have any trouble with 1Khz sample rate; similar cable is used for serious audio all the time (ignoring the market segment that values every dB of deliberate distortion . . .)

Common mode (and the normal mode noise that comes with it) might be more of an issue.

Are your transducers 2 wire, deriving power from the loop, or 3 or 4 wire which use a separate wires for DC power? If 3/4 wire then there's a reason to use 0-20mA, if available on the transducers in order to get an additional 20% range at no sacrifice (unless you need the 'live zero' for long term diagnostics/troubleshooting).

Are the transducers and their power supply 'floating' from ground? My antiquated experience was that higher speed AI's were like scope inputs - one side grounded, even the multiple inputs. That could be a problem over 200 yards with common mode between the grounds of each end. Or maybe higher speed AI's are now isolated.

I'm curious about what your analog receiver is with a 1KHz sample rate; that's not a typical PLC AI. 0.01% is one part in 10,000, so you'll need a minimum 14 bit (1:16,000) A/D for that kind of resolution. But it'll probably be a 16 bit A/D in the end given that the leading bit is probably a sign bit and the LSB is a 'toggler'.

National Instruments has a variety of PC based AI cards at a range resolutions and sampling rates.

Dan

 

[itsmoked]

For the fidelity you're after I'd go with smart sensors that are at the monitoring spots and talk over RS485 to the logging system.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[hyprfrco]

First of all, thanks for your replies. I was seeking in Electrical Stack Exchange, but i dont know why i cant seek for proper feedback in there.

This is my first question in this forum, so i hope to start collaborating in a more active way in this an other sections.

The biggest issue would be the receiver or ADC. Precision and speed usually do not come together or if they do it is very expensive.

I am agreed with that. Nevertheless, the standard specs for the NI 9203 +- 20mA 10-bit AI Module DAQ are 0.04%/0.02% Gain/Offset errors and 24k Samples/s per channel. Which is quite good for me for $541 (plus $820 for the chassis).

Lowering the signal to noise ratio is just being careful with shielding and grounding. Current loops are low impedance so have a high immunity to noise. On the other hand because they are current driven, wire inductance and capacitance will affect the signal response time.

I agree too. I could suppose here that i am following all clean recomendations.

Current loops are low impedance so have a high immunity to noise. On the other hand because they are current driven, wire inductance and capacitance will affect the signal response time.

I am indeed concerned on this point. Do you know how could i calculate error figures from these parameters?. I am really interested in knowing how to obtain an estimate from cable specs, so i can calculate a priori how much and which type of cable should i use for an specific sensor.

My last 25 years have been in the process field with 1-10Hz sampling rates, so my experience from the early '80's with DAQ at higher speeds is way out of date.

I dont think so... Industries are really filled with 70-80 machinery ...

Given the range of AI's out there - 10 to 16 bit A/D, single ended to fully isolated, I don't know how a "typical" S:N could be stated.

Neither i ....

My thought is that to get a system signal to noise value, you'll have to measure it. Put a solid mA signal in at the field end (can the transducers be put into a held static state at a fixed output?) and see how much jitter you get at the receiver end. The problem with using a battery powered floating calibrator or a battery and a resistor is that neither shows the inherent instability of the transducer and its power supply.

I have no idea whether it is typical or not, but the unit I was working with last week has a factory spec sheet 4-20mA accuracy spec of 0.2%FS. With a calibrator attached at the input, 4.4000mA input gave a 4.4002 or a 4.4003 reading. When I connected the calibrator at the field end, the reading was identical. There is a cal routine to zero any error out (probably the dropping resistor), but given the application, I didn't bother. Rock steady, no noise. Belden 8760. The low pressure DP sensor would jitter about plus or minus 0.0002mA at zero DP.

This is the hardest figure i've obtained from my research. That is less than a 0.0001% error. Can i ask you the cable distance for the cable you used?
Perhaps you are right, and i should measure and forget about calculations.

Without looking up all the capacitance values, I'll speculate that shielded twisted pair like Belden 8760 (18g) or 8762 (22g) will not have any trouble with 1Khz sample rate; similar cable is used for serious audio all the time (ignoring the market segment that values every dB of deliberate distortion . . .)
Common mode (and the normal mode noise that comes with it) might be more of an issue.

Actually the 1kHz sample rate and the cmmon mode rejection will depend solely on the DAQ side, and not on the cable.. or i am wrong?

Are your transducers 2 wire, deriving power from the loop, or 3 or 4 wire which use a separate wires for DC power? If 3/4 wire then there's a reason to use 0-20mA, if available on the transducers in order to get an additional 20% range at no sacrifice (unless you need the 'live zero' for long term diagnostics/troubleshooting).

Are the transducers and their power supply 'floating' from ground? My antiquated experience was that higher speed AI's were like scope inputs - one side grounded, even the multiple inputs. That could be a problem over 200 yards with common mode between the grounds of each end. Or maybe higher speed AI's are now isolated.

I guess it would be "safer" to have a 4 wire configuration for this case. I am from the segment that value every dB loss ...

I'm curious about what your analog receiver is with a 1KHz sample rate; that's not a typical PLC AI.

The application is for measuring sensors, like the 393B04, with 6e-5% resolution and 0.02-1700 Hz freq. range. Hope that feed your curiosity .

0.01% is one part in 10,000, so you'll need a minimum 14 bit (1:16,000) A/D for that kind of resolution. But it'll probably be a 16 bit A/D in the end given that the leading bit is probably a sign bit and the LSB is a 'toggler'.

National Instruments has a variety of PC based AI cards at a range resolutions and sampling rates.

You are completely right here.. Typical high end PLC 4-20ma AI modules have a 0.1% accuracy or less?, with typical sampling rates of 10Hz?. I am also looking for NI modules.

My open questions about how to calculate 4-20mA loops are uncountable, but could be summarized as follows:
[ul]
[li]If i have a 0.01% or better sensor, such as the above indicated, how should i read it, if they are going to be placed for monitoring, each 200m away from a DAQ?,[/li]
[li]Which is the deal breaker in distance/precision? It is 4-20mA enough?,[/li]
[li]Could the impedance of the cable degrade the sensor resolution totally?,[/li]
[li]Should i use a digital converter immediately after the sensor and move to digital buses -Fieldbus, Profibus, etc.- and forget about 4-20mA?.[/li]
[/ul]

The cost difference between 4-20mA and a digital bus scheme is simply too big for not making that prior evaluation.

Thanks for everything, fellows....

 

[djs]

The inductance and capacitance of the cable will limit the rise and fall times of the signal. This limits the effective bandwidth of the signal passing through it. An extreme but relevant example is to pass a square wave through the cable. You will see the leading and trailing edges rounded over on a scope. If the rounding is great enough, and the width of the square wave pulse is short enough, then the overall amplitude of the wave will be affected, as will any measurement of that signal.

On another note I don't think a 10 bit ADC will give your the precision and resolution you want. ten bits is 1024 counts which is about 0.1% resolution. You will need about 14 bits to achieve 0.01% resolution. Also looking at the web page you linked above, those sensors output a millivolt voltage output, not a current. Which brings up another issue which is signal conditioning. To achieve max resolution you will have to match you signal range to the range of the ADC.

 

[danw2]

1) Fieldbus/Digital protocol: Moot point for the selected transducers - there is no digital protocol.
If there were, the question would be throughput rates and the availability of the protocol and its application layer software. But there isn't one.

2) cable distance:
>Can i ask you the cable distance for the cable you used?
There were 8 points, cable distances ranged from 100 feet to 500 feet. Signal was 4-20mA.

However, my case is a moot point. Those 393B04 transducers are voltage output: volt/g output.
At 1000mV/g, and a measurement range of plus/minus 5 g pk, the signal output range is plus minus 5V.
The referenced mA value is for excitation current from the power supply, not the output signal (tells you how big a power supply you need).

Voltage signals do not have the same characteristics as loop current; some voltage drop and a higher susceptability to noise.

4) 1700 Hz - somewhere between Ab and A above high C on the piano. If it were 4-20mA signal over STP, I'd say no big deal.

But in looking at the product photo, it appears that cabling is coax. I could not find an installation guide for the product, but the site has an app note on driving long cables, which they consider > 100 Feet. So, it's probably worth your read.
[URL unfurl="true"]http://www.pcb.com/techsupport/tech_longcables[/url]