How to protect the shield of the 447X?

[Ian_Ren]

In the 447X manual (page 2-10), it is stated "The shield does not have overvoltage protection. Do not make a non-ground connection to the shield. Also do not connect the shield to a ground that varies more than +/- 2.5V from the ground of the NI 447X".

Most dynamic signals (eg. accelerometers and microphones)after signal conditioning can be safely connected to 447X, however, some dynamic signal sensors do have a DC offset from the ground, and some may have a varying ground (eg, some proximity sensors). Unfortunatly, not every engineer understands everything about sensors and DAQ system he is using everytime, and everyone likes a fit-for-all "fool prove" system.

Here are my questions

1) What is the best practical way to pr
otect the shield of the 447X?

2) Will it help to protect the 447X if I connect the shield of 447X to the chassis (earth) ground, and effectively makes 447X Referenced Single Ended (RSE). In most of our applications, we can probably tolerate certain levels of noise, but frying a 447X every so often is not an option.

Thanks.

Ian

[duncan carter]

This sounds like a continuation of the "Difference between unbalanced-differential and single ended" thread. What is the output impedance of the sensors that have negative voltages on their "grounds"? What happens to the sensor megative outputs together if they are not floating? It really doen't sound like the 447x series board is a good candidate for this without some external conditioning. Do you need the characteristics of the 447x series or would it be better to use something else like an E-series board?

[Ian_Ren]

Hi, Duncan

I work for an NVH consulting company. We have developed a 40-channel dynamic signal acquisition and analysis system based on PCI 447X boards, mainly for NVH (Noise and vibration)projects. We are able to acquire 40channels of data and write to hard disk at 102.4kHz sampling rate(if necessary)continuously.

The characteristics of 447X, such as ICP power supply and anti-aliasing filtering, and its compactness fit our need really well. We would have to use additional signal conditioning (filtering, sample and hold, etc)if we were going to use E-series board.

Most of the time, we use accelerometers and microphones. 447X works great with these signals. However, sometimes, our customer ask us to measure something from different sensors. In o
ne recent case, we were asked to measure a signal and were told that the signal range was only a few hundred mV. After some setbacks, we found out that the negative of the source was about 6V above the ground (the sensor effectively uses a wheatstone bridge).

So far, we have not damage any 447X cards. However, I am worried that one day, a fresh young engineer may connect an improper signal to a 4472 and damage the card. I cannot really blame him if he tells me that the output range of the sensor in the specifications is just +/- 1V.

What I am trying to do is to have some ways to protect the 4472 cards. In that sense, even an inaccurate/wrong measurement is tolerable as long as we can tell that the measurement is not right.

You suggestions are most appreciated.

Regards

Ian

[duncan carter]

If you really don't need 24 bit resolution, one approach for the general, unknown, case might be to use something like a balanced version of an oscilloscope probe with a 10 to 1 or 100:1 attenuator with the series resistance divided between the two input legs so that there's a fairly high resistance in series with both the "ground" input of the 447x board and the normal signal input plus fuses, at least in the "ground" lead. My company is in the machinery diagnostics business (see vibrotek.com). Mostly, we use ICP accelerometers so the input of a 447x is just fine but sometimes there's a need for others like eddy current probes; in this case, if we need to make a dc coupled measurement, an attenuated measurement is just fine and usually the
supply ground isn't a problem. One application that can be a problem for us is some railway shop testing is done with the surface that the accelerometer is mounted on is at something like 70 volts above ground, in this case, our most foolproof low-tech solution is to use a battery powered portable data collector with the operator trained to not allow himself to be a path to ground. This works in Russia but probably wouldn't in the USA ;-)).

I'm an old geezer and I learned long ago to be cautious about connecting things to other things. I think I'd make a rule that any sensor connections need to first be mulimetered first. This isn't always foolproof; I was found myself standing behind a client who multi-metered two phases of a unexpectedly live 4400 volt 3 phase power line. He survived with burns but his Fluke dvm didn't.

We have software drivers for both 447x boards and E-series boards. Having one E-series board with a balanced, attenuated input available might be the s
implest solution for unplanned situations.

[duncan carter]

One variation of the balanced attenuator might be to make a differential preamp probe. Another variation might be to use two 447x channels, one input connected to one either bridge terminal plus a divider between the two for the 447x ground.

[duncan carter]

Or maybe there is a case ground to connect to the 447x "ground" to use with the two 447x input channels.

[Ian_Ren]

Hi, Duncan

Thanks for the advice.

We are making several differential amplifiers to address the non-ground signal issue. In the mean time, We also need a probe to check in the first place if a "unknown" signal can be measured suitably by 447X.

I have attached a file containing the sketch of such probe (which makes use of two 447X channels). Do you think that they will work? I am not an old timer but just know enough that my "brilliant" ideas do not always work.

Thanks.

Ian

[duncan carter]

I don't think the single 20 K resistor does anything for you. I was thinking more of having 2 channels with 10:1 dividers for checking and having a VI that displays the individual inputs and the difference in the inputs, multiplied by 10 and, if the mazimum magnitude of both inputs are less than 10 volts, go to direct connections for both channels. For these odd signals, do you need to follow them with complex VI's or could you make a few special (channel 1 - channel 0) VIs?

[duncan carter]

If the sensor actually floats, then you do probably needs some added ground return to make it not float.

[Ian_Ren]

Adding a divider is probably a more prudent way, though signals we have dealt with have never come close to 42.4V (the positive protection limit of 447X).

One version of our DAQ system is DC powered, and the sensors may also float (DC powered). That's the 20k resister for. Does 20kohm sound like a resonable number to you?

Thanks

Ian