Analog Input Fluctuations

[Steve Etter]

Hi all,

I am using a GE Fanuc IC693ALG442 Analog (4 In/2 Out) I/O module with a 90-30 363 processor. In my application, I have a two descrete 5K ohm potentiometers with a single, externally supplied 10VDC source for two input points on this card (%AI2 and %AI4). I also have a third descrete 5K pot with a separate 10VDC source as a third input point (%AI4). The two sources are completely isolated from each other and the rest of the machine as well. The last input is a 4-20mA signal (%AI1)..also completely isolated from the rest of the machine. My programming software is VersaPro.

My problem: When monitoring the %AI2 value using my laptop or the HMI (a Mitsubishi E910 touchscreen), it is fluctuating by as much as +/- 10% while the input voltage remained steady (measured using a Fluke 23 Multimeter).

I tried removing the second pot from the circuit but this had no affect (I am not surprised since the input voltage is constant already anyway).

The input is only called in one point in the applicatation program and I have verified this.

Aside from being able to tell it that the input is a voltage or current type, there appear to be vertually no other calibration factors available for this card that I have found.

Does anyone have any clue what could be causing this problem or how to resolve it? I have not yet contacted GE...thought I might ask y'all first.

Steve

[PLucas]

Steve

We use analogue blocks on a genius bus system to carry the speed signal from our joytsicks, they are +8, -8 Vdc inputs. Our signals fluctuate slightly, but only when the joystick is at zero, to overcome this we have programmed in controlling logic to ignored this signal until the joystick is away from the off position. I am aware that this might not be much help to you, but we put this fluctuation down to noise on the genius bus. As this signal does not effect the running of our cranes we ignore it. Could you do the same or is it vital that the signal is spot on?

If you do get in contact with GE and they give you an answer maybe you could post their answer here, I would be interested to hear what they have to say about it.

Paul

[kalle]

For what it is worth: In some systems where we have had "noisy" analog input from sensors with own power, the trick has been to connect the 0 VDC.

I don't know if it correspond to your hardware.

[Tom Jenkins]

The last time I had that problem with a GE analog input, it turned out to be improper wiring technique. Make sure that the power supply is properly hooked up, and the common (-) tied back from the PLC input card to the power supply and from the device to the power supply. I find that drawing a very detailed loop diagram, including the internal commons and internal I/O card resistors and such helps me make sure I've got it right.

Next, make sure that you have shielded cable on the analog signal wires, and that the shields are properly grounded in one and only one end. (This was the problem on my last experience with this - the panel had green/yellow striped terminal blocks, but they were floating!) Use an ohmmeter to measure the impedance to a known good ground.

[Steve Bailey]

A couple of things come to mind. First, what kind of filtering is the Fluke doing on the analog signal? There might be some noise that the module sees that the fluke doesn't. Have you looked at the signal with an oscilloscope? The IC693ALG442 module updates all four input channels sequentially approximately every 8 milliseconds, meaning its internal controller spends about 2 milliseconds per channel.

Second what are the raw data values in the %AI002 register? The module is 12-bit resolution, but instead of a 0 - 4095 range, the bit pattern is shifted 3 places left so as to make the MSB the sign bit. The result of doing that is to make the step size of the data 8 counts, corresponding to a voltage increment of 2.5 millivolts.

Have you tried moving the wires from %AI002 to a different channel to see if the problem follows the field device or is confined to channel 2 of the module?

Did you check your configuration to make sure that there is no overlap of %AI addresses between different modules?

[Steve Etter]

First of all, thank you for your responses. I would have responded sooner, but my ISP had an outage this morning and I just barely got back on-line.

Paul - Yes, it is pretty important that my input is steady…not necessarily "spot on", but I can't tolerate a 15% swing either. Unfortunately, my system fluctuations are across the board rather than at a single point like the one you described. I have called GE since posting, but have yet to hear back from them…I will certainly post whatever solutions they might offer.

Kalle - I assume you are suggesting that I connect the 0VDC common to a system common or even ground. Unfortunately, my system cannot tolerate this, either. The source of the 10VDC is an analog DC drive whose control circuitry cannot withstand having its 0V tied to the same one as the PLC is (I already had a design that way and burned out several drive control boards before learning from the manufacturer not to do this).

Tom - Improper wiring, while still a possibility, is one of the first things for which I checked. I already have a detailed (and simple) diagram that I am pretty confident is correct for the application. I will go back and double-check it again, though. With that in mind, what all do you suppose qualifies as "wired right"? I have my 10VDC common connected each of the two respective PLC input "0V" and the pot wipers connected to the respective inputs. The +10VDC goes solely to the pots. Is there anything else I really should do, too? I do know for a fact that my shielded cables are all only grounded on one end. This is a very well known and practiced technique in our shop (only three of us do any wiring) and I double-checked it on this machine to boot. This brings up an on-going question, though; should shields be tied to ground or DC common? The question arises from the fact that transient voltages and such are known to come in on ground circuits. Can these negatively affect the protection otherwise provided by shielding?

Steve - No, I have not been able to analyze this using an o-scope. My company does not own one and the customer's site is far away, so I have not been able to borrow one yet and go see them. While I did not move the signal wire from one input to another, per se, I did replace the module and had the same problem. This leads me to believe the problem is likely somewhere in the channel or wiring…the thing that is puzzling though is that all three potentiometers are mounted side-by-side and the wire routing is virtually identical. It seems that if noise is the problem I am fighting, then all three inputs would be acting similarly. I did check for I/O overlap and there does not appear to be any…all %AI's have a length of 1 and %AI2 is only used once in the program. There are no words or double words accessing this either (I am not even sure if there could be, but I looked anyway).

For the moment, the customer is making due but I will be visiting him pretty soon. So far, taking a look with an o-scope and swapping inputs around to identify if I am looking at a channel or wiring issue is all I have. Any other ideas?

Thanks.
Steve

[ben]

It’s a stab in the dark, but have you a flat spot on your pot. I've seen it before where the pot wears out in one place. Occurs when wiper is constantly measuring in just one spot. Very common on damper position pot run on a PID. Try changing the pot maybe?
The scope is a good idea, and changing the input to another, could be a floating voltage, already suggested though.... Sounds like a process of elimination...

Good luck...

[Allen Nelson]

Steve (Etter):

When you measured the voltage with the Fluke, did you measure it in parallel (across the terminals) or in series (pull one wire and put the Fluke in the middle).

I had this one beaten into my head once by "just a tech" when I couldn't figure out why my 4-20 mA input module was showing the equivalent of 14 mA, while the Fluke (and the OEM's generated signal) was showing 20 mA. Turned out that the OEM was only putting out 5 V (instead of 20), and the internal impedience of the Fluke allowed it to read it "properly", while the Input module didn't. (Of course, the OEM was told to get the right sized power supply).

If you put the Fluke in series, I bet you'll see the same ±10% fluxuations as in input module

Now as to the cause....

[Steve Etter]

Ben - I doubt it is anything like a flat spot because the fluctuation exists regardless of the rotational position of the pot. Anyway I should be able to more easily check this when I switch I/O points.

Allen - My measurement was parallel. This is a 0-10VDC input rather than 4-20mA. The only 4-20 is %AI001 (also verified in my hardware configuration). Are you suggesting that I go ahead and put the Fluke in series anyway and look for amperage fluctuations? I have never heard of doing this before on a 0-10VDC system, but am willing to give it a go. If so, what kind of current do you suppose I should see?

Steve

[ben]

Yes you have a very good point their Alan.

Its another stab in the dark because I can't see your schematic.

Is this input 4-20mA or 0-10vdc?

Have you a voltage going through a trim pot and then to your pot giving you a 4-20mA input? Sounds like the case?

If your voltage measured at %AI2 is steady but fluctuating in your PLC is this because you are measuring voltage on the current loop? Measuring voltage on a current loop will give you a steady voltage. Providing that your pot is still that is.

Make sure that the input is 4-20mA spilt the circuit and then put your meter on mA and in series. See what you are getting now?

As Allen said.....just diagnostics....so far....

[shahid_iftikhar]

Dear Steve,
I had been working with Ge Fanuc PLCs for more than five years, some times new modules from Ge Fanuc show the same behaviour as you described, its better to replace it and check. Usually Ge system support never reply in time.

Regards

[Kevin Hammond]

Steve,

I've run into similar problems on a number of occasions, and EVERY time it has been a common/equipment ground issue. Just a few weeks ago, I had a data acquisition system with three 0-10Vdc inputs displaying the same symptoms you describe. Solution: Replace the wire from my 10Vdc source common to equipment ground. By equipment ground I mean the frame of the machine.

Now, on all of my systems, I make sure that ALL commons get connected to equipment ground. Whenever I have noise issues, or any other unexpected machine behavior, the first things I look at are grounds and commons.

Good Luck -
Kevin Hammond

[Steve Etter]

Ok, I agree that tying common to a system ground will likely cure my problem (assuming of course that it is externally induced) but as I mentioned before, my source is an analog DC drive and it cannot accept being tied to ground.

I already swapped out PLC modules with no improvement, so I am inclined to think this is not the source of my problem.

I just got off of the phone with the DC drive manufacture and he confirms that the grounding solution cannot be done without shorting out the drive. He believes that I probably have a noise issue.

So here are the things I plan to do, in order, when I go to site again:

1) Replace any non-shielded wire in this circuit with shielded cable.
2) Install a small filter across the PLC input to filter out any noise reaching the PLC.
3) If that doesn't work, remove the filter and install a resistor across the input to load up the 0-10VDC source.
4) If that doesn't work, remove the resistor and install a new, separate 0-10VDC power supply to power the circuit.

Question: Does anyone have a formula or rule-of-thumb for sizing filters for this type of application. I have never had to do this, so I really don't have a clue what to get.

Also. Is there anything else I should put on my list?

Steve

[Tom Jenkins]

First off, When I refer to grounding shields I mean earth ground, not chassis ground (which SHOULD be the same) or power supply "-". It isn't normally required to tie power supply "-" to earth ground, but it can't hurt. Many +24 VDC power supplies already internally connect "-" to the power supply ground.

Voltage inputs are inherently more susceptible to noise than current inputs. As a last resort you can try putting a signal conditioner/isolator at the drive to convert the 0-10 VDC signal to a 4-20 mA signal. These devices are only a couple hundred dollars, and are available from M-Systems, Acromag, A-Pi, and a whole lot of other sources. Some are loop powered, but for your case I suggest a separate 24 VDC powered unit.

I try to avoid using the term "common" when referring to the "-" side of a power supply. Common can mean many things, depending on context. A 120 VAC line connection to a group of limit switches, for example, can be legitimately referred to as a common. This kind of thing has caused me lots of grief with field wiring in the past.

Steve,

We have had LVDT's do the same thing you are talking about. We had tester that was supposed to zero down to +-.001. We also had some fluctuations of upwards of .003 to .005 thousandths. What seemed to cure our problems the most was installing a choke around the input cable. While it did not completely get rid of the fluctuations, it did bring us back down to where the tester would give repeatablity of +-.001 again. These chokes are fairly cheap, and they just snap on the cable, simple to install. They might be worth the time and effort, because both are minimal.

Russ