short term spikes in the output 2

[hydrae]

Got a problem I haven't seen before,
got these spikes in the output of a
submersible pressure transducer

These spikes occur at random times about once a day or so
This is measuring level in a water tank
Tank is 24 ft tall
The reading ignoring the spikes is cycling between 18 and 20 ft as expected
Range of the transducer is 0 to 33 ft
Any Ideas?

Hydrae

 

[IRstuff]

> are they real, i.e., has any other sensor been used to see if they show up>
> how long has this sensor been in use? It's possible that there is intermittent failure on digital output
> are the spikes the same value?
> pressure transducer might be excited by physical shocks or accelerations of the fluid

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[hacksaw]

Submersible pressure sensor?

You'll have to describe your fluid especially if solids are present, the design of the vessel,etc.

 

[hydrae]

Thanks for your quick replies

Spikes would not be real
It is a ground level steel water tank for domestic use
350,000 gallons
The fluid is filtered water

Spike height is different value every time
Scada box transmits the reading every 2.5 minutes, the window you see is a few hours
the spikes are only there for one reading

Sensor is 10 to 15 years old, it is a chore to replace though so trying to avoid that

Other analog inputs do not show any abnormalities

Can I put a capacitor across the 250Ω reader to dampen the spikes until we can change out the sensor?

Hydrae

 

[TugboatEng]

Domestic water tank? Why not use a top mount sensor (ultrasonic or radar) and abandon the old one? I'll often install them on manhole covers or caps to avoid the issues related to hot work on an in service tank.

 

[IRstuff]

Do you use only analog data, or do you use the digital data? If the former, then possibly, but any such filter will likely clobber the real transients, such as the putative drain/refill "V" features shown in your data. If the latter, there are lots of digital filters that can remove spikes; a simple median filter or clipping function would wipe the spike.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[danw2]

A submersible pressure transmitter is likely to be a 4-20mA 2-wire loop powered transmitter, which means it's inherently isolated signal source. Usually 2-wire transmitters can tolerate and operate over a wide range of supply voltages, so even a 24Vdc power supply that drifts to 30V momentarily is not likely to produce a bad output value from the pressure transmitter.

It's not a real level change in 350,000 gallon tank, but it could a noise artifact. The field wiring might pick up an intermittent ground loop offset from some device ground-faulting in the panel or from a motor with a ground fault that runs for 2 minutes and then stops.

If it's a noise artifact, then the source has to be "ON" for a reasonable period of a minute or two, to be captured by the 2.5 minute sampling interval. The spike acts as a time stamp. Is there any piece of equipment that energizes for 1-2 minutes at every spike occurance?

 

[itsmoked]

So this spike is lasting minutes??

Keith Cress
kcress -

http://www.flaminsystems.com

 

[TugboatEng]

It's a sample once every 2.5 minutes. The spike is possibly happening often but at a short interval so the sampling picks it up infrequently. The software simply holds the high value until the next sample.

 

[3DDave]

With no information on the sort of sensor it is or how it's monitored, the problem could be anywhere in the system.

 

[TugboatEng]

That's why I suggested to install a new sensor of a completely different type. Bypass all of the current problems and start over.

 

[thebard3]

From the description you give, it sounds like the problem is as likely to be caused by 'Scada box' as the transmitter itself. Need more information to troubleshoot here.

Brad Waybright

The more you know, the more you know you don't know.

 

[hydrae]

Thanks Dan
Yes it is a 2 wire loop powered device, there are no motors in the area or any other
source of noise, it is a tank on top of a hill, there is a farmhouse nearby but no heavy machinery. the times of the spike appear to be random. I do not have a tight correlation on the source being wind in the area but it is windy now and it is more often now. There is no outside influence that I can see other than possibly wind, which points toward the ground loop fault. there are also some spikes in the downward direction.

I am also inclined to believe it is a noise artifact. I will recommend to the client the cable get replaced with a twisted shielded pair cable. It might be only a twisted cable inside a PVC conduit.

It is analog signal to a analog input to a scada box that reads the analog data and sends it via cell phone text message to the servers.
The power supply is a 24vdc with crowbar and powers another analog device that does not show any abnormal behavior. The loop is +24 from power supply to sub pressure transducer to scada to 0vdc.

Thanks thebard3, since the other signal from the scada box is reasonable, I am not looking at the box at this time, though the oem is looking in the programming to suppress this alarm.

Thanks tugboat, the installation of a different type of sensor is more costly, would put another penetration in the tank. We will stick with the sub pressure transducers.

Here is the past 2 days of graphing.

The different top off points are a normal variation in the control that shuts the valve which is subject to velocity depression in its sensor line. (it is a bug that is lived with). The fact that the top is a curve and not a point is due to the surge reduction method used on this valve as the fill line is miles long and goes through several pressure reduction valves along the way.

 

[IRstuff]

Do you want to remove all the spikes, or only the really large ones? Below shows the level and 3-point limnar regression slopes, so looks like ± 10 hr^-1 slope passband would eliminate most of the spikes, even some of the smaller ones. Your second image implies that it's the sensor just glitching

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[BrianE22]

Loose wire? Bad ground connection somewhere?

 

[hydrae]

Thanks for all your comments
We got the SCADA settings to ignore the spikes and will schedule a replacement sensor and cable
as soon as time and budget allow

Hydrae

 

[AlmostRetired]

I'm with BrianE22, it looks very much like a ground fault to me. I'd be checking every wire connection to make sure they are tight and have good contact. Noise is also possible if the instrument wires are close to an intermittent high voltage source, such as motor cables, if the instrument wires aren't a shielded twisted pair.

 

[djs]

I had a submersible level transmitter which would give wildly inaccurate readings when ever one of the pumps in the wetwell started. It turned out that the transmitter had an internal short to the exterior metal casing. Replace the transmitter and end of problem.

 

[TugboatEng]

Not to plug a specific company or product but do consider moving away from the pressure transducer level measurement. Ultrasonic is well suited for large tank applications like this.

https://www.omega.com/en-us/level-m...el-sensors/ultrasonic-sensors/p/LVU700-Series

The power demands are higher so you may need to add a power supply for excitation. I would probably try to run without shielded cable. Have you seen the price of that stuff lately? It would be cheaper to run metal conduit if you had to.