4-20mA accuracy - Is it even possible?

Hi Everyone,
The situation: 4-20mA Druck pressure transducer (5 psi, 138.5inches) in a lake. 1/4 mile from Aromat PLC reading in a scale of 0-8000.
Challenge: To read in the lake level to the nearest .1 inch.
Problem: Incoming signal is varying by 50 points. (57 points is 1 inch)
Questions:
Is it even possible to get the desired accuracy with this set up?
I am having trouble getting an extremely accurate milliamp reading as I only have a meter not an oscilloscope. Is this variation of .1 milliamp "normal"?
I am scanning at 3 milliseconds and taking a reading every .1 seconds to average 10 readings. Does this sound like a good averaging rate?
Any other advise on this would be greatly appreciated. I've been arguing with it on and off for 6 months.
Thanks everyone,
Sarah

Let me get this straight. Your trying to measure the level of a lake to a 1/10 of an inch? Now your seeing a fluctuation of less than an inch? Just a slight breeze on a lake, I would think would make it vary by much more than that!

Sarah,


Where is the pressure sensor mounted/ what does it’s diaphragm see? The fluctuations of almost an inch sounds like "waves", you may need to filter the sensing location to eliminate this noise. Think of an open ended cylinder that will allow the noise to occur outside the cylinder and the inside where the transducer is located should be much more stable. Height of the cylinder depends on the maximum level variation you are expecting.


MikeW

I guess I am curious as to how you are translating the pressure signal into a linear inch measurement? Is the device itself giving you an inch signal or is it giving you a pressure signal that you are mathing out to inches? Are you trying to do this by simply reading the pressure at the bottom of the lake?

OK, I see what it is now. Did you heed some of the warnings in the manual?

enclosure or a Campbell Scientific DES2 desiccant case.

Dislodging Bubbles​

While submersing the sensor, air bubbles may become trapped
between the pressure plate and the water surface, causing small
offset errors until the bubbles dissolve. Dislodge these bubbles by​

gently shaking the pressure transducer while it’s under water.

Temperature Fluctuations​

Temperature fluctuations can be minimized by using a minimum
cable burial depth of six inches and a sensor submersion depth of
one foot. Also, if your site may experience extreme temperature
fluctuations, the transducer must be measured using the six-wire​

configuration.

I would think that waves/wind would change your reading in addition to changes in atmospheric pressure. Pretty tricky to get a level reading on something that's always in motion in one way or another.

There are dataloggers and instrumentation specifically geared to taking the kind of readings on water level and quality you are atttempting. I suggest looking at USA Blue Book or buyer's guides for Pollution Equipment News or Water Online magazine.

Druck has a reputation for high quality. I doubt the fluctions are a false reading from the pressure transmitter, I suspect it's true data - waves producing alternating head pressures on the sensor. The transmitter's temperature spec and its influence on accuracy is worth looking into, to see how much error winter to summer water temperature change can influence the head pressure reading.

I doubt the fluctuations are from your analog input or from the vent tube.

The problems with vent tubes is that they can pinch closed, but more oftent the vent tube collects condensation because the ambient water is colder than the dewpoint of the atmospheric air in the tube, so the air's humidity condenses, drops down the tube, collects at the bottom and then the zero shifts because the vent tube pressure is the gauge pressure reference leg. But that would appear as a slowly changing offset, not rapid fluctuations.

I can't imagine that the fluctuations are not real wave action.

I'm not sure what the folks do that Tom Jenkins recommends, but I suspect that some heavy damping or filtering, rolling average type thing, would be used to average out wave action and be more along the lines of what you're looking for, not instantaneous changes from second to second. After all, how quickly can the nominal lake level change?

I'd first start by looking at your A/D conversion. What type of convertor is used in the Aromat PLC? How many bits? What is the quality of the power supply for the PLC and transmitter?

I don't know the instrumentation used, but the USGS keeps an accurate track of the level of the lake I live on down to .01 feet elevation. There are some random anomalies, but they appear as a sudden, and usually single bad reading. For what it is worth, they only have one recorded measurement per day, which is probably an average of a 24 hour period.

Mischief55 said:

I am scanning at 3 milliseconds and taking a reading every .1 seconds to average 10 readings. Does this sound like a good averaging rate?Sarah

Click to expand...

There are A/D's that integrate over a whole or multiple AC cycles in order to integrate out the noise (positive integral of random noise summed with the negative integral of random noise equals zero).

An A/D sampling at 3mS is not likely an integrating A/D, but a flash A/D, taking a momentary 'snapshot'.

Some instruments require a setting for line frequency, 50 or 60 herz, because their analog sampling needs to synch to the line frequency to avoid aliasing noise into the end result signal.

Your idea of averaging is good, but it's far too few samples, giving you a 30mS window. Every bit of jitter is being captured and any line frequency noise, as well.

I strongly suspect that RD Rast's 'average over 24 hours' is what is done for these types of measurements.

Think about how fast your lake lavel can physically change based on the incoming flow of all feeder stream CFS. This than can lead you to a better time period for your filtering.
My guess is that you are not taking nearly enough samples, your sample rate is too high, and your averaging window is way too short.

Thanks Everyone,
We have the pressure transducer in a stilling well and the vent tube and wiring is at the top in a lock box. I will check the condensation at the vent tube and put in something to absorb extra moisture. The waves are a small part of the problem, but I think I agree that my sampling rate needs to be adjusted. Just as a little more information: this is a hydro site in the Hudson Valley. We are being required by the DEC to hold the lake level to within 2 inches of crest.
Thanks again for all the help and I'll keep you posted on the results.
Sarah

My guess is that you are not taking nearly enough samples, your sample rate is too high, and your averaging window is way too short.

Click to expand...

Sarah, I second that thought. I would change the averaging time period to at least 1 hour. I do not think you are concerned about measuing every boat wake and wind wave, but need an indicator to show whether the lake level is rising or falling over some time period.

Averaging real-world analog signals over time does wonders to filter out unneeded noise. It takes a while to realize that the real world is not set up to produce nice clean PLC signals, but instead produces dirty, varying, fluctuating, bouncing signals.

We are being required by the DEC to hold the lake level to within 2 inches of crest.

Click to expand...

You should request a clarification on this requirement. On a 5-acre lake, I know from experience that wind-born waves can easily make the level fluctuate more than 2 inches. On a larger lake, it could be much more. For this small 2" variation in level, and reading the level with a pressure gage, there are many other factors: gravitational pull of the moon (tides), changes in atmospheric pressure, temperature changes that cause the water to expand or contract, amount of dissovled gases in the water, condensation, electrical noise from unshielded points in the terminal blocks, and so on.

Would a 1-hour average level be acceptable to determine the +/- 2 inches of crest? If not, then what averaging time period would be acceptable? Bureaucrats do not think in terms of what is practical, but are pie-in-the-sky thinkers, so you have to pin them down to a measurement method that is doable.

Personally, I would first determine what averaging time period WILL give the required resolution, then ask the DEC if THAT averaging period would be acceptable. Like a lawyer, phrase the question so that a positive answer will give you what you need to make it work.

As others have stated, there are many variables to consider when using a pressure sensor for this application. With any sensor, I think you would need to average the reading over some amount of time, to account for waves and disturbance in the water surface. With a pressure sensor, you also have wind / pressure variances that will effect it, as well as currents in the lake.

I always liked the pressure sensor, for measureing level, but if you are looking for 0.1" accuracy, then another sensor may be better for this application (if for nothing else, to take a few variables out of the equation.) I'm thinking of a laser/radar type level sensor looking into a tube (to eliminate waves) Pressure/current fluxuations will still vary the level in the tube, but it should be minimual. The questions is always what is really needed for the type of control/data logging you are doing.

I'm thinking of a laser/radar type level sensor looking into a tube (to eliminate waves)

Click to expand...

This is hard to explain, but a stilling tube eliminates the crests of the waves, but does not eliminate the effect of wind blowing water against a land masss. This results in waves, but also slightly increases the depth of water on the down-wind side, and decreases it on the upwind side of the lake.