danw
Lifetime Supporting Member
200 pressure units for a span of 150000 is 200/15000 = 1.33%.
Do the other channels use a similar difference value, percentage wise?
Do the other channels use a similar difference value, percentage wise?
200 pressure units for a span of 150000 is 200/15000 = 1.33%.
Do the other channels use a similar difference value, percentage wise?
I'm trying to determine if the AI card is wired single ended or differential, so that's why the actual terminal numbers are meaningful.
Does transmitter (+) go to IN-4 and transmitter (-) go to RTN-4,
Or does transmitter (+) go to IN-4 and transmitter (-) go to IN-5?
No, the other transmitters are configured to alarm with about a 10 difference because its temperature and pressure, and less fluctuation.
Is the DC power supply (-) grounded or floating?
How much oscillation/variation is there in the flow signal?
500 units per second ?
The wiring puts a single, 4-20mA current signal through a 250 ohm resistor. The 1-5Vdc voltage drop across that resistor is connected to each of two 1756-I16 AI cards.
The manual for the 1756-I16 card shows a single-ended wiring scheme for voltage signals.
This may, or may not be the wiring in question.
If it is, how could channel A on input card A see a different value than channel B on input card B ?
They can't. Both channels get exactly the same signal, a voltage paralleled from a common source, 4-20mA signal.
Could common mode offsets account for the Ch A vs Ch B difference?
While single ended AI cards are notorious for having common mode problems, manifested as an offset on each channel when multiple devices at different ground potentials are connected to a single ended card, that is apparently not the case here.
Common mode problems on single-ended cards affect all channels on a given card, because they all share a common ground. The complaint is typically, "the first three channels worked fine until I connected signal 4 and now all 4 channels are driven off-scale."
That is not the case here. And it is unlikely, since 2-wire transmitters are typically ungrounded (floating), so even if one side of the power supply (p/s) is grounded, the ground potential difference is limited to the difference between the p/s and the AI (RTN terminal) grounds. This assumes all the transmitters share the same power supply.
Furthermore, 3 signals, each split to dual channels, have worked without a problem for some period of time. as singled ended voltage signals.
I therefore conclude that a difference due to true differences in the signal source is unlikely.
However, a difference in signal values might be accounted for in the sampling of signal values for Channels A and B.
The signal in question is flow signal, which has a significant amount of noise/variation:
"I would say about 300 to 500 per second the flow vary."
"If I watch the PLC AIs channels live/online, I can see the difference fluctuate up to 200 quickly. Its flow reading and I know it vary, but shouldn't vary by that much."
The signals which are not problematic are apparently signals that have less variation than the flow signal:
"No, the other transmitters are configured to alarm with about a 10 difference because its temperature and pressure, and less fluctuation."
The less noisy signals do not see a deviation.
The noisy signal sees a deviation between Ch A and Ch B, even though the signals at the AI inputs are identical.
I'm an instrument guy, not an A-B guy, so I do not know how the input values are sampled in time, with respect to one another.
But I suspect that there is enough time lag between the sampling of Ch A and the sampling of Ch B that the difference is caused by the inherent noise/variation in the flow signal.
Ch A is sampled at T0 (time zero) with a value of 12444 units.
Ch B is sampled (arbitrarily) 50mS later by which time the flow signal is now 12645 units, only a 1.3% change, but enough to trigger an alarm.
An A-B guy will have to comment on how channels on two different AI cards are sampled and what parameters (scan time, scan order, whatever) affect that sampling and what a likely time skew between AI card A and AI card B samples could be.
It might be that filtering/damping/averaging the flow signal at the flowmeter so that non-simultaneous sampling will not see as great a difference as is now seen.
1). >"Since this value is steady on both channels, the sampling time is somewhat the same, so the difference is less."
No, the time skew/sampling time difference is still there but the difference is minimal because the input signal is not changing, it's stuck at whatever 'count' zero volts is interpreted as.
The disconnected flow signal is effectively 'steady state', even less noisy and steadier than the temperature and pressure signals.
2). "Whats the difference between RTN and iRTN?"
Ans: The letter I (eye, not EL) means 'current'; that terminal is used for current inputs. A 249 ohm resistor current dropping resistor is connected internally between terminal IN-n and iRTN-n. There's no resistor connected to RTN.
Totally agree.I believe the problem of the difference of signals is process flow oscillation/noise/spike. Do you agree?
1). Totally agree.
The data is not sampled synchronously on both channels. The difference in sampling is a time interval long enough that process has changed value by the time the 2nd channel is sampled. The result from a noisy data signal is a difference greater than 1.3% of span.
2). What is the flow technology? DP?
3). Options
Given that this is an SIS (safety) application, whatever changes are proposed needs to be reviewed and approved by whomever does that in your organization.
a). Live with the current situation. People will consider the alarm a nuisance alarm and ignore it.
b) Decrease the time interval between samples so that the differences are smaller to make the sampling more 'synchronous' than asynchronous.
c). Increase the difference value used for alarming due to the noisy nature of the signal.
d). Increase the time duration that the difference value must be valid before it is considered a valid alarm (on-delay timer)
e). Smooth the output of the flow meter signal. Because the algorithms are time constant based, there is an inherent time delay in an 'averaged' signal.
f). Loop tuning
I don't understand the control strategy. There are two devices controlling flow, an upstream VFD and a downstream damper/valve. I would think the PIDs fight each other.
Either controller can contribute to quick changes in flow rate, dependent on tuning.
Recording the output (demand) signals to the VFD and the valve positioner might reveal which is more aggressively tuned and more likely to cause oscillation.