anti-aliasing filter

[MrGargantua]

Hello there,

I am in the process of purchasing the usb-9215A and plan to utilize this for an FFT application.  The signal is already normalized to 50mV/ips but is unfiltered and it represents vibration transducer data.  I have used the usb-4431 with success as I have tested it in a jet engine test cell with the zoom FFT vi.  The problem is that the 4431 is being used for our shaker table testing and to purchase another one for field testing is outside our budget.  I understand the 9215A is just a DAQ card without any antialiasing filters and AC coupling.  I plan to design a programmable 10th order low pass filter(switched capacitor type) to prevent any aliasing when I am acquiring data.  So my question(s) is:

1.  Would this hardware be adequate to create an FFT with meaningful data?

2.  Is it ok to apply a switched capacitive type low pass filter on the analog input of the 9215A?

So basically I will be adding the analog filter before the amplifier stage of the 9215A and not "right before" the ADC.  Will this be problematic?  Many thanks.

Neal

[Frank-L]

The quick answer to this question is that yes, it is possible that using a external filter to the USB-9215A will allow you to do what you need to do and will potentially give you a FFT with meaningful data

The long answer is that the USB 4431 has 24 bits of resolution to the USB 9215A's 16 bits and the 4431 will provide a more accurate representation of the data's shape. Additionally with the 4431 having all components integrated into the same circuit, we are allowing less noise, potential for problems, and a better signal to be obtained. This is why we provide an all inclusive device such as the USB 4431 to get the best measurements possible, as well as allow different combinations of signal conditioning to be on the same device, allowing more versatility to the device.

Plenty of people do signal conditioning with the USB 9215A and have had success. The biggest difference is that with the USB 4431, we have the ability to spec out the expected quality of the signal conditioning to allow you to know what values you should be expecting, whereas with a external signal conditioning you may not get what you are expecting due to impedance differences, additional noise, and signal degradation.

[MrGargantua]

Thanks for the reply Frank.  I have started to layout the programmable antialiasing circuit as I have prototyped it and tested it with a function generator.  I have also recorded the Jet engine complex signal profile when I was on the marine base so I can compare my results to actual/real vibration data.  I understand the 9215a is only 16 bit vs the 24 bit of the 4431but in relation to vibration measurements, we have a system that measures overall vibration data and the test cell operators will use that primarily to check bearing wear/health.  The FFT I am designing will complement that system while breaking down the vibration peaks within the frequency bandwidth therefore, resolving those energies down to the uV level is not that important or not completely necessary in this case.  My question is:

1.  Will a 16 bit module effect the frequency or time domain accuracy? 

2.  Would the inherent noise from an industrial enviroment/jet engine swamp the important signal information because of its 16 bit resolution?

Thanks,

Neal

[Frank-L]

The use of a 16 bit module could potentially hide important information. For example, if you are measuring 10 V with 16 bits, the resolution would be .15 mV. If you had some sinusoidal vibration under that value at a certain frequency, that vibration would not be present in the FFT because the device could not represent that small change in value. So instead of seeing a main vibration frequency, and a smaller amplitude at a second frequency; you might not see that second frequency. With the 24 bit resolution, at 10 V you would see all changes that are greater than 596 nV and could notice additional low amplitude noise at certain frequencies. 

[ChrisR.]

That is true only if there is no other noise in the system.  If there is enough noise to properly dither the system, then smaller signals can emerge from the noise floor if enough samples are taken.  The primary effect of going from a 24-bit to a 16-bit system is to raise the noise floor.  Look at the noise specs for the systems you're considering.  And if the noise is dominated by the environment, then 24 vs. 16 bits won't make any practical difference.

To answer your other question, 24 bits vs. 16 bits has nothing to do with frequency accuracy, which will be determined by the module's timing engine.

[MrGargantua]

Thanks guys I really appreciate the replys.