Another (different) pressure transducer question

[Noway2]

One of the key function of my application is monitoring the system pressure and so I am naturally going to need a pressure transducer. My question is: is it better to select a current output of 4ma to 20ma or use a voltage output such as a 1v to 5v range. When I say better in this case, I am thinking mostly in terms of noise immunity. While I don't have any facts to back it up, my inclination is that current source would be less susceptible to interference from and AC or other induced sources.

From an interface standpoint, a 5v maxium from the transducer would exceed the natural input range of my ADDC. My thinking is that with a current source output, I could use a small, precision burden resistor, of say for example 150 Ohms, and a buffer amplifier. The 4ma would then give me a signal voltage of 60mV and the 20ma would give me 3.0V, which is within range of the ADC and well above lower (noise) limit. Since the value of burden resistor is about 1/100th that required to achieve an attenuation with the amplifer I would use it seems to me that the potential for error would be less this way too.

If anybody here has some experience with these issue would you please help me to understand the positives and negatives assoicated with these two approaches?

 

[itsmoked]

Hi Noway2.

Normally 4-20 outputs require some hassle to "create" so they cost more than voltage outputs. Of course they are more complex to do so they probably are somewhat less reliable, hardware wise, though it would probably be hard to pin that down.

If your sensor is close to where you are "reading" it by all means go with Voltage Output. As you will save on cost.

If the transducer is inside your machine, 4-20 would generally be a waste of money.

If the transducer is going to be more than oh, um, something like 20 feet away from your reading hardware then 4-20 starts being more logical.

If your product is a crap-shoot in that, gee customer X might want the transducer a 100ft away then YES go with 4-20mA.

Other points to consider:
Remember resistors have a noise factor too.

Often depending on your "reader" voltage devices may be easier to reduce voltage offsets and drifts by having your system supply "referenced" excitation.

Using voltage units you can often cancel out the excitation voltage altogether using ratio-metric setups.

Many 4-20 output devices use a micro to actually create the 4-20 so there can be some lag that effects your process control stability.

 

[Ashereng]

How about converting twisted pair to fibre optic, and run the FO back? FO will be more immune to noise.

 

[Noway2]

Thank you for the information and the suggestions. Fortunately, in this case, I am in control of the transducer location as it is inside the cabinet of my (OEM) system. The reader (probably an AD628 or similar difference amp) should only be within a few feet of the transducer. Speed requirements are slow. I know that is relative, but if I am able to respond within about a second I will be fine. I consider a one second time frame to be very slow, relative to the 120 MHz processor that is running the application. To that effect, since everything I am measring is DC or slowly changing signals, I am filtering all of the analog inputs fairly heavilly with a low cutoff frequency (about 100 Hz). I was able to use this to my advantage, relative to the ADC to get away with a simple 2 pole anti-aliasing filter. Combined with some software calibration, I have been able to achieve a stable and accurate result of within 1% with the on processor ADC when alot of people seem to consider the ADC to be about worthless.

I definately see the point about in a distance type application 4-20ma beeing a wise choice. If the control system were at a significant distance, like itsmoked suggests, I don't think there would be much question.

Regarding Ashereng's suggestion of fiber optic, the use of FO cable to communicate between the display unit and the main computer was part of the original game plan because of noise considerations. While it would have been real overkill, I was still stinging smartly from the after effects of someone elses "poor noise" design. The design architecture has evolved considerably from that point and I didn't need to go that route. I realize that the evolving spec and architecture can result from a lack of upfront specification, but I have been attempting to hit a moving target without a marketing department to tell me what the customer needs . itsmoked, I wish I had your "flushing" icon for that comment :-D

 

[felixc]

If the sensor is within your system, then you can also think about sensors that just contain a measurement bridge. There are ADCs like the AD7710 from Analog (and some more recent models from ADI too), that contain all the excitation circuitry and integrated variable-gain amplifiers for such applications.

 

[buzzp]

Generally, transducers are more easily found with 4-20mA outputs. These can always be converted to voltage signals if you wish. I realize some use this language loosely so are you developing your own pressure transmitter or are you buying an off the shelf unit with an output? If buying or making I would still use 4-20 since voltage can always be converted from the current signal. This would allow it to be used more universally (if this matters).
Even if the distance is short, the current loop is the way to go unless this is an end product that might cost $10-20US then you can start counting your pennies to keep the cost low. You can make a 4-20mA signal without adding more than a couple of dollars in hardware (in quantity).

 

[Comcokid]

Let see - it's inside your cabinet, slow response (1 sec), 1% accuracy, and digitized by a uP A/D. Everything your are indicating leads me to believe a 1-5V output pressure sensor is just fine for the application. Simple resistor divider or op-amp with gain<1 to scale the 5 Vmax to your 3.0 V input is all you need.

Just remember - Some A/Ds may have an overrange bit, but it sometimes takes a lot of overrange to set the bit. And while that A/D channel is overdriven, it will tend to offset the values on other A/D channels that you may be using for other things.

 

[buzzp]

I guess whats in the panel is irrelavent in selecting which source to use. We don't need to know that.

If it is all on PCB then fine, a voltage signal may be alright. If hes running wire, then good luck and have fun troubleshooting. Not saying a problem would be inevitable with a voltage signal just way more probable. You need to consider closely where this signal will pass in your panel to determine which is the best. Sure the signal has to be converted to voltage for the micro to read but what happens between the transducer and the micro is anybodys guess at this point. More info is needed to determine this. Based on what we know, the current loop is the way to go until more info is provided.

As per the original question, current loops are way more immune to noise than voltage.

 

[itsmoked]

Like I said "I" would stick with voltage in this case unless your machine has a few VFD's in it and a plasma generator and EMP generator.

Just run twisted pairs to it use an adequate instrumentation amp as they are dirt cheap these days. Use some caps to short out the high freqs.

Now you mentioned something that bothered me a bit.. quite a bit.

You said you used a 100Hz filter but are only really needing once a second or so. THE biggest noise hassle in this type of stuff is often the line(mains) current which permeates everything in a line powered system.

I would really try to filter to a lower frequency to knock off the 60 or 50 Hz so I would use a 20 or 30 Hz low pass on the front end.

Barring the ability to do that I would sample locked to line frequency so the noise picked up from the mains would always be the same and hence you could ignore it.

I could easily see this as your system's biggest noise issue causing subsequent [flush] followed by

Hey just click on emoticons/smileys above the {Preview Post} button and at the top of the window that pops up hit "Animated" and you too can

 

[Warpspeed]

Another way to do it would be to mount a voltage output pressure transducer right onto the main circuit board. That will practically guarantee electrical noise immunity.

Then run a small bore pipe to whatever pressure point you wish to monitor.

 

[itsmoked]

Or rubber tubing! Depending.

 

[Noway2]

All, thank you for the replies.

In response to the requests for more information, per regulatory agency requirements I have a limited number of vendors that can supply an approved transducer, unless I am willing to spend $10k and 6 months to test another model. The existing product, which I am designing a replacement for, has been using a transducer by Gems Sensors for a number of years now. The Gems device has proven be extremely reliable. For political reasons concerning what drove us to switch to the Gems transducer I am very reluctant to use anything else. The transucer uses an ASIC and some form of configurable (by them) output amplifier to select the pressure range, output type (0 - 5, 1-6, 4-20ma, etc) all from the same device. The transducer will always be located in or attached to my cabinet. The signal wires, which are attached to the transducer by the manufacturer, are not twisted but there is a shield wire, which I am hestiant to connect because the body of the transducer is connected to this shield and that would be grounding the shield at two points. The length of the run can be upto 6 feet in length.

ITSMOKED, as far as noise generation sources, I have been at one installation where there were some VFDs to drive the potable water pumps and this caused our equipment to malfunction, though not the pressure transducer portion). Also in certain modles there are multiple pieces of fairly large (upto about 1000 Amps at 480Vac) switch gear IN the cabinet, which is known to cause interference issues. Also, the machine is designed to operate when the facility is burning to the ground and I can just imagine what sorts of noise sources will be present and active then. I see your point about it being desirable to filter out the mains frequency. In the sister application to this one, which will use the same hardware, it is necessary to monitor the mains power, so I don't want to use a hardware filter to remove this. I did implement a fourth order butterworth filter as a digital IIR filter and I can selectively run this on any of the channels (at different cutoff frequencies if I desire).

In an earlier post of mine, we had discussed placement of the ADC relative to the signal to be converted. Based upon what I learned from that discussion, I am planning on putting an ADC on the board where teh transducer wires wires will be terminated.

 

[buzzp]

Hmmm, maybe I was right and the 4-20mA is the way to go with all the potential radiators in the same cabinet. Save yourself the trouble and design a 4-20mA. Since the mains can't be filtered either, this is a current loop application.

You also need twisted pair, shielded and not just shielded wire. The twisting helps eliminate any common mode noise present on the signal.

 

[Ashereng]

Sorry to sound like an old record (skipping), I would still go with fibre optic (FO) - what with all that potential interference.

 

[itsmoked]

Noway2:

What pressure are you measuring?

Are you redesigning the boards?

Is this system of yours controlling something or just monitoring?

What was the original output/signal type you were using?

 

[Noway2]

The pressure that is being monitored is the water pressure in the sense line for a (building / fire) sprinkler system. The system monitors the pressure and then starts a pump, either an electric motor or diesel engine driven one, through switch gear when the pressure falls below a programmable setpoint. As far as the original system goes, depending on the model it was either a 0 - 5 volt or a 1 - 6 volt output. I am redesigning the boards. In the product where I at least got to specify the output range, I chose the 1 - 6 reasoning that having 0 psi corresponding to an actual signal voltage would allow us to differentiate between a 0 pressure input and certain fault conditions. Consequently, the choice is probably between a 4 - 20ma and 1 - 6V output.

Regarding Ashereng's comments about using fiber: I definitely agree with using fiber and had planned it in part of the application, however, in this instance I don't have a whole lot of options as I need to connect to a transducer that comes as a signal wire referenced to common output. From the point where I am in control of the signal, it is being digitized and handle as a numerical quantity from that point.

The goals in the redesign are:
1 - dramatically reduce system cost, which is dominated by labor assembly time through smart choices of hardware
2 - decrease unit assembly defect rate by eliminating or reducing known labor failure points (wiring especially).
3 - increase controller metering accuracy and add ability to capture peaks and valleys in input signals.
4 - redo aspects to the user interface of the controller that are either poorly implemented or are lacking in our product compared to the competitors.
5 - achieve a high degree of commonality between the electric motor and diesel engine versions to limit the need for inventory, spare parts, and differing builds.

From a design perspective, I believe either a voltage output or a current outwould would work satisfactorilly with similar costs associated with either approach. What I was after was opinions on wether or not there were significant advantages or pitfalls to either method that I wasn't aware of.

 

[itsmoked]

Sounds like too high a pressure for mounting the sensor on the board and routing the media to it. Though this would probably be the cheapest by far.

Consider a small local board mounted to/near the transducer with something like an 8pin PIC that reads the transducer filters it if desired and then sends a digital value to your main board. This would allow all sorts of features or mods without effecting a main board or its software. This would also be noise immune. Would allow cheap plastic fiber if desired. (though I think that is over rated for your app)
Also if you want any kind of display or indicator lights on the panel like fire trucks, which have them all over next to hose connections, you could put this little board adjacent to the pressure sensor adding the LEDs or displays.

Otherwise use either method they will both work. But you are correct in using the 1 to 6v for error detection.

Use the shield all the way back to your main board just don't hook it up at the main board. It should still provide useful shielding.

Seems to me that your filtering schemes are far to complex for the situation at hand when a resistor or two, a few EMI beads and a cap or two would hammer all the noise without software and beating and aliasing problems, but maybe I am missing something.

 

[Noway2]

I believe that we are in agreement on this issue. A PIC is one one consideration for this board, which is really the analog interace board. I am on the fence between that and a dedicated ADC chip, that doesn't require any programming. There are few internal analog signals that need to be monitored, of which the pressure is one. They are all DC, or close to it, as they are the battery voltage, oil pressure and water temperature. Doing a board like this this keeps the analog circuitry away from the digital as much as possible and puts the converter as close to the signal source as possible.

Yes, I admit, my filtering scheme is a bit complex for this application. I used a DSP for the application, partially because I wanted to experiment with one and partially because I have future plans for it regarding some features for this controller. I didn't mention them in case any competitors read this group too and I don't want the cat leaving the bag prematurely .

I want to mention, though, I certainly do appreciate the help everyone has given me!

 

[Warpspeed]

I agree that a digital signal is always going to have a higher noise immunity than an analog signal, especially if the circuit is broken at some point with an opto isolator. The PIC sounds like the answer then.

Another slightly different approach would be a voltage to frequency converter located right at the pressure transducer. That can easily give a higher resolution and greater dynamic range than an ADC, which may for practical reasons be limited in the number of bits. For sheer simplicity it would be difficult to beat.

 

[itsmoked]

Nice point warp; V to F at the sensor.. That would work.

If you do have a couple of sensors/points then it might be a good idea to do a little general purpose PIC board using the a2d on the PICs. Then set up a comm protocol and if you run addressing you could add new features at any time by assigning them a new address. This could allow a large range of additions to be handled by SW on the main board. And if a really unique requirement comes up a special PIC board could be done without disturbing the main-board.

If your internal comm spec was 485 then you could add remote sensors hundreds of feet away. Like monitor the battery charge on a remote generator or cooling tower etc.

Anyway sounds like a fun product!