07-19-2005 01:55 PM
Hello schi,
I have done measurements of the ground acceleration with 2 same accelerometers. These accelerometers have a frequency range from 0.01Hz to 50Hz.
I have done an acquisition using the PXI4472B with IEPE Constant Current Source : my two sensors were connected respectively to the channels 0 and 1. I found a good coherence only in the frequency range : 10Hz to 50Hz.
Now, I have done the same acquisition with these accelerometers. The only difference was that the accelerometers were connected respectively to the channels 5 and 6. The parameters were the same for each channel. I found a good coherence in the frequency range : 0.1Hz to 50 Hz with just a fall of the coherence at 10Hz.
I think the problem is that the IEPE powered supply is not enough constant :
- if I perform the coherence between channels 5 and 6 without any sensors plugged in and with the IEPE configuration, I have a good coherence between 0,1Hz to 1 Hz.. If I do the same manipulation without the IEPE configuration, I have a bad coherence between channels 5 and 6 for all frequencies. So the noise seems to be the current source which is not perfectly constant, but which have low frequencies components.
- if I perform the coherence between channels 0 and 1 without any sensors plugged in and with the IEPE configuration, I have a bad coherence for all frequencies. But, without any sensors plugged in, if we use IEPE powered sensors or not, the Power Spectrum Density of acceleration has approximatively the same amplitude for channels 0, 1, 5, 6. This Power Spectrum Density of acceleration is increasing from high to low frequencies.
The problem is that we seem to have much noise inside each channel at low frequencies, which give us inacurate results if we want to measure ground acceleration below 10 Hz with accelerometers.Do you think it is because of the IEPE powered sensors? I have done these measurements during 320 seconds. The length of the NFFT was 16384 with an acquisition frequency of 1024 Hz. So, the PSD and the coherence were performed during 16 seconds and I have done 20 averages.
Please answer me,
ben
07-20-2005 02:41 AM
Hello ben,
please let me first point out that I never used a 4472B for myself, so I am not that familiar with the practical details of this device.
First, I think you should get clear about the IEPE concept. The base idea is to have some constant current (voltage) applied to the sensor so that the sensor itself can have an integrated (charge) amplifier. So you do not need an extra charge amplifier and have no sensitive cabling. To suppress this constant supply, you have to do some AC coupling, as otherwise the IEPE voltage will saturate the ADC with its constant offset. It does not really matter, whether you do this by an external IEPE supply (which will include AC coupling) or at an ADC card. A typical -3dB corner frequency for this filters is 3 Hz (e.g., 4472, 4462) or 1 Hz (some other companies devices). If yoou need to measure low freq. vibration, you can take more special devices as the 4472B, which is specified with a -3dB point at 0.5 Hz (if I remember correctly). Regarding your idea to do the high-pass filter digitally on the system, that does typically not work as you have the high DC voltage on the ADC. IEPE supply voltage is specified to be 12 V (or so), so it will even overload the 4472 input. You have to keep away the DC level (offset) from the ADC.
Regarding the setteling time: It is the high pass filter of the AC coupling. You should, as a rule of thumb, allow the filter at least a settling time of 2 to 3 times its inverse corner frequency. For 0.1 Hz, this is 20 to 30 seconds! And this is also the time scale you need to have a real constant (in terms of muV) voltage / current of the ICP supply not to see any additional noise in your results - you see the challenge ...
Another major problem is with your sensor side: You want to measure velocity or deflection, but you do measure acceleration. So, what you have to do in your data evaluation is to integrate the acceleration. Integrating acceleration at very low fequencies is hard to do as you often are integrating noise only. The accelleration levels are very very low at these frequencies and you really need highly sensitive accellerometers to measure this. Please check carefully, where the expected noise floor of the sensor (including its build-in amplifier) and the input card (including the transfer characteristics of the high pass filter for the sensor side, but without for the ADC side) really is in terms of your finally measured data and the required data range.
Outside the NI product range, there are vibration measurement systems which will fit your needs. I will not recommend you any specific right here, as I might be biased.
Please think about the overall measurement chain for your problem, not about an ADC card only. You have to fulfil a complete measurement task and you have to make the whole chain fit and deliver good and valid results. That is much more than the DSA card.
Hope, it helps... good luck, anyway.
schi
07-20-2005 03:53 AM