ripple/noise and faulty corrections acquiring with PXIE-4496 from impact hammer

[LAPadron]

Hello,

Reading an impact hammer with our PXIE-4496 in a 1073 chassis, and with LabView, I am getting things like these ones (that are obviously wrong). I attach a simple project file. I know this card cannot read DC, but I think it should be able to read this impact correctly. Shouldn't it?  Any idea of which could be the problem.

Thank you in advance, and happy new year!

[johnsold]

You say these are "obviously wrong" but you do not say what you think is wrong with them.Please be as specific as possible and defining what you think this should be doing.

One I do not see any excess noise there. If you look at the very ends of the traces, you see very little noise.  Several things that I do see are: 1. Your sampling rate is probably too low.  The reason I think this is that you do not have smooth variations, only very sharp jagged ones. 2. The apparent DC offset in the first trace may simply be due to not waiting long enough for the signal to return to baseline. It looks like there may be a slight upward slope in the trailing edge part of the signal.

Lynn

[LAPadron]

Thank you very much for your reply,

The reason I say is obviously wrong is because the negative values cannot be real, as it is an impact hammer impulse. It should be a very clear peak. It was taken at 1000Hz, and I thought it should be more than enough, but I will try with a higher sampling rate as you say. Even though, if it were a problem of the sampling rate, I still wouldn't understand the reason of the negative values.

(I will post the tests with higher sampling rates, but any idea to also test will be very welcomed)

Thanks

[johnsold]

Is the transducer connected to the hammer or to the object being struck by the hammer?  In either case the mechanical behavior of the object to which the sensor is attached will be like a damped spring-mass system: It will oscillate under the excitation of an impact.

Consider a pond. Drop a stone into the pond.  The waves which propagate outward from the point of imapct have components which are above the mean level of the water and components which are below that level.  The same thing happens in solids.

The required sample rate may depend on the speed of sound in the objects being measured and the resonsnt frequencies of the objects and of the transducer.  Based on the images you posted, you need to sample at least ten times faster.  It is quite likely that the peak amplitudes are significantly higher than the highest points on your graphs.

Lynn

[Brad_K]

The ripple/ringing is explained by this KB: Why Does my Dynamic Signal Acquisition (DSA) Device Not Handle Impulses, Square Waves, Triangle Wave...

Brad

[LAPadron]

Thank you very much. I will take a deep look into it, but I still have one doubt that I don't know if could have something to do also with the DC filtering or not:

I repeated the experiment with a much higher sampling rate (50000Hz) as suggested by Lynn, and I guess a much nicer response now (complete signal on the left, zoom of the trail on the right):

As you can see, after the impact and the not very nice trail (probably due to the Gibbs phenomenon), the signals does not go back to zero, but gets stuck in -0.15, going back to zero very slowly afterwards. Is that related to the DC filtering? Is there anything one can do about it? (I have tried with different sampling rates (up to 50.000 Hz) and the problem is always there)

Thank you very much in advance,

Luis

[johnsold]

Luis,

If you look closely at the left image (5 ms x-axis), you will see a slight peak at about 3.6 ms.  I suspect that there is something happening in your mechanical system that has a time scale of a few milliseconds.  Can you acquire 20 - 50 ms of data after the impact and see what it looks like at the end of that time?

You applied a peak force of 140 N and a short time later you have a slowly varying component that is 0.1% of that force. The hysteresis in the load cell in the hammer could be that big.  Is this a real problem or ar you just trying to understand what is going on?

Lynn

[Henrik_Volkers]

No return to zero:

As you stated in your first post: You already know you can't measure DC  😄

Most IEPE signal conditioner have a build in high-pass filter (a CR and buffer)
So after a single pole pulse (impact) you will always get an offset. The amout depends on pulse length and strength (integral of exitation)  and the HP frequency.

[Henrik_Volkers]

---

@LAPadron wrote:

Is that related to the DC filtering? Is there anything one can do about it?(I have tried with different sampling rates (up to 50.000 Hz) and the problem is always there)

Luis

---

Yes, math 🙂 If you know / measure the frequency response of your signal conditioner ( or even better including the FRF of the sensor) you can match a compensation filter.

Or use other equipment that don't have this effect.

BTW: Happy new year Lynn 🙂

[LAPadron]

Thank you very much to both,

First, I attach the image Lynn asked. We can see that there is a tendency towards zero, but after 0.2 s, it hasn't reached it yet.

And, as this is the reading of the impact hammer, I guess that, after the impact, it is not measuring the system itself, but it is only noise.

Henrik, one question. When you say measure the requency response of my signal conditioner (or even better including the FRF of the sensor), what do you mean exactly? measure the frf of the DAQ system or of the impact hammer?

On the other hand, if I know that the signal read from the impact hammer after the impact is "noise", Could I just make it zero, or would I be introducing larger errors that way?

Happy new year everybody!

Luis