SCXI-1125 overvoltage criterion

[coalflame]

Hello all,

I am going to use a SCXI-1125 module with SCXI-1327 terminal block to condition thermocouple signal. Estimated signal range is about 0-50mV.

The problem is, since the thermocouple bead must be attached directly to a resistance heater, on which current flows at 100Hz with certain duty cycle, it is expected that part of the heating voltage(no more than 1V) will be imposed on the thermocouple bead. We can manage in LabVIEW so that no overlap exists between data acquisition and heating.

I use an oscilloscope to measure the voltage from the thermocouple wires. Surprisingly, at the end of each current pulse, there seems to be a negative peak voltage. The duration of the peak is about 0.02ms out of every period, which is 10ms. Magnitude of the peak varies between 1V to 20V.

I look into the manual of SCXI-1125, which says when measuring voltage larger than 5V the 100:1 attenuator on the terminal block (for example, SCXI-1327) should be applied. But when I tried this way, the temperature signal suffered from large noise, maybe due to its very low nature.

If I bypass the 100:1 attenuator, is there possibility to damage SCXI-1125? I wonder whether the peak voltage or the continuous voltage should be avoided. The manual doesn’t provide parameters such as “input current during overvoltage condition”.

Any suggestion would be greatly appreciated.

Su

[johnsold]

Su,

I would certainly expect that a 20 V peak could damage the device.

Use two diodes back to back across the input to the terminal block.  Add some resistance in series to limit the current, perhaps 1000 ohms.  The diodes will limit the voltage to about 0.7 V and the reistance will limit the current to about 20 mA.  The resistance will likely have minimal impact on accuracy (unless the devices contain open thremocouple detection circuits), but you will need to run some calibration checks to verify.

If you cannot use this method due to the open thremocouple circuit, you may need a custom signal conditioning circuit designed for the application.

Attaching thremocouple beads directly to energized circuits is always asking for trouble. Is there any other way you can make the measurements you need?

Lynn

[coalflame]

Lynn,

Thank you very much for your input. You are right that connecting thermocouple bead directly to energized circuits comes with problems, but in my application the temperature changes so rapid that I have to keep close trace with it.

Our device doesn't contain open thermocouple detection circuits, so I think your suggestion is applicable.

But another question, how about if I use a Zener diode with Zener breakdown voltage well below 5V instead of the two diodes you mentioned? I feel it can be more reliable.

Best regards,

Su

[johnsold]

Su,

A 3 volt zener diode or something similar should be fine.  The protection is asymmetric but that should not be a problem.

Lynn

[coalflame]

Lynn,

Thank you for your reminder. In fact it really should be a bidirectional protection. I will post the result later on.

Best regards,

Su

[coalflame]

Lynn, I suspect that the "peak voltage" at the end of the heating cycle maybe is spurious signal. First, if it comes from electromagnetism induction, appearance at every end of heating cycle should have been expected, but in fact only at the first one or limited ones of heating cycles follows the odd signal. Second, I applied the method as you described but nothing changed. I even adopted a "custom attenuator" made of resistences but the signal remained. We can be sure that your method does work for voltage higher than Zener voltage of the diodes. So we "riskly" bypassed the 100:1 attenuator on the SCXI-1125, and we read nominally the same signal. When the task is switched to temperature reading, everything is OK and no damage occures. Can you give some comments on the "spurious signal"? Again, thank you for your help. Su

[johnsold]

Su,

Can you post an image (.png is best) or a VI showing the peak signal saved as default values on an indicator or control?  Often the waveform or timing can give a clue to the origin.

When attenuators do not attenuate the signal, it sometimes is a clue that the problem is ground bounce.  That can occur if current flows through a common conductor and the impedance of the conductor results in a voltage drop. These situations can be tricky to diagnose and to cure.

Is there anything going on in the system or nearby which has large currents flowing with the correct timing to correlate with the spurious signal?

Lynn

[coalflame]

Lynn,

Thank you for your very fast reply. In fact our heater was fed by two 12V lead-acid batteries in series and the averaged power input is ragulated by an IGBT module, and the two electrodes which stretches the heater are floating-connected rather than grounded. The thermocouple wires are also ungrounded.

During the run, I think there is no large external interference(such as a welding machine or other on/off operated device) nearby. Once I recharged the batteries beside the apparatus and interference as large as 1 V appeared. In other case no descernable random interference seems to exist.So maybe the problem is from the inner of our device?

The png file named "nonoverlap" shows the voltage signals every 50 samples for a period, during sampling no current flows. The png file named "temp" is the temperature signals corresponding to the previous voltage signals, every data is the result of the average value of the 50 samples mentioned above. The last file named "interactive" is the voltage signals every 150 samples in a period, with each last 100 samples overlapping with the heating pulse. You can see two peak signals in the beginning (the data before the first negtive peak come from a previous run).

Best regards,

Su

[johnsold]

Su,

Those certainly look like the kinds of signals often seen when something suddenly turns on or off. Transients at turn on and off do not always occur which may account for some of the randomness in when you see them.  For example your IGBT circuit is probably a switching regulator.  It is possible that if the Turn On or Turn Off command occurs when the IGBT is conducting you get a spike but if the IGBT is not conducting no spike occurs.

Since the desired signals are quite small and slowly varying (except for the random component) and the spikes are much larger, I would probably keep a running mean and standard deviation over the past 500-1000 samples. When a signal exceeds Mean +/- 5* Standard Deviation, then reject the signal as probably a spike.  If necessary a linear regression could be used to obtain the slope, but I doubt that would make much difference.

Lynn

[coalflame]

Lynn,

Yes, that our IGBT is a switching regulator, and from the trace of an oscilloscope we can see the rising edge of the heating pulse(rectangle if ideal) is relatively smooth and the falling edge is rather sharp. So to avoid interference of the heating current we manage it so that when sampling temperature no heating occurs.(even if no "spike" be generated, the voltage gradient across the thickness of a 0.002-in.thermocouple wire would then be in the order of about 12 mV, equivalent to about 300 K of input for the commonly used type-K (Chromel/ Alumel) thermocouple.) In this way the temperature reading normally is free of spike, but the noise from other sources still makes mean action as you described necessary.

By the way, it appears that, at least with SCXI-1125 and PCIe-6351 no thermocouple open-circuit detection device is available. And if something undesirable happens, such as vibration or friction to the thermocouple bead, then even if the thermocouple wires remain connected, the readings become wild (in this case the data we get should be spurious ones) and we cannot handle that. I have tried with a case structure with a rule that "any data has a deviation larger than 25K compared to the previous one should be replaced with the former one" but it doesn't work. Do you think it is also due to the nature of the "spurious data"? (For real data the rule works.)

Best wishes,

Su