PXI-5660 and f0 - difference

[KR4K]

Using PXI-5660 to read spectrum. Using ni5660 Configure for Spectrum.vi, I set start frequency at 2000000 Hz and stop frequency 30000000 Hz, resolution bandwidth 3000 Hz. I then read the spectrum using ni5660 Read Averaged Power Spectrum.vi.  I get the output values:
- spectrum
- f0
- df
Why does f0 not agree with what I input, and why does it change? For example
Input start Freq    Input stop Freq    ResBandwidth    f0                      df
2000000              30000000            3000                   2000187.421     856.783
2000000              30000000            10000                 2000000.0         2857.14     
2000000              15000000            3000                   2000000.0         855.798
20000000            60000000            3000                   20000187.42     856.783
20000000            35000000            3000                   19999572.1       855.798
200000000          350000000          3000                   200000187.4     856.783
5555555              26666666            3000                   5556373.665     856.783
17777777            48888888            3000                   177778280.54   856.783
17777777            48888888            1000                   17778000.0       285.714
17777777            48888888            10000                 17780000.0       2857.14

f0 should be exactly what you enter as the start frequency, per your manuals. These differences are unacceptable, and they open questioning to all the results.

73/gus

[Yardov]

Hi Gus,

Thank you for using NI Forums. I am looking into the driver(s) and the VIs and we will post back as soon as we have updated information. If you have any new information feel welcome to add it as well.

[Andy_Hinde]

Hello Gus,
The ni5660 Configure for Spectrum VI asks for a center frequency and span, so I assume you are converting an input for start and stop frequencies to center frequency and span on the block diagram. I just wanted to state that for the purposes of posterity on the Discussion Forums.

After looking into this and speaking with some of our SW engineers, the reason you are seeing this behavior is due to coercion of the RBW and some basic assumptions. One assumption is that the center frequency requested should show up exactly on a spectral line frequency. In your first case listed above (Fstart = 2MHz, Fstop = 30 MHz, RBW = 3kHz), there will be a spectral line positioned exactly at 16 MHz. The remaining spectral lines fill out the spectrum at intervals of 856.783 Hz to either side of the center, and usually will not end up with an integer number of lines to fall exactly at the start and stop frequencies requested.

An underlying assumption to the one above is what we term in our software as 'RBW Definition'. When specifying an RBW, an assumption is needed as to whether the bandwidth specified in RBW is the 3 dB BW of the RBW filter, the 6 dB BW of the RBW filter, the Equivaqlent Noise Bandwidth of the RBW filter, or the Bin Width of the RBW filter. The default in all cases is set to 3 dB Bandwidth.

Your use case implies the use of Bin Width as the RBW definition and changing this will allow you to achieve the behavior you're looking for and allow you to specify the bin width in the RBW parameter.

There is an 'Advanced Settings' input for the ni5660 Configure for Spectrum VI. This is where you can change the RBW Definition, as well as specify a different spectral window or specify the exact number of spectral lines you want. Create a control for this input and change the RBW Definition to Bin Width. Alternatively, you could achieve the same result by entering '-1' into the RBW control and specifying a specific number of spectral lines.

In the case of using RBW Definition of Bin Width and a value for RBW that does not equate to an integer number of spectral lines for the requested span, a coercion will still be made and the assumption will still be to provide a spectral line at the requested center frequency, with adjustments made to the start and stop points. This is the case for your first use case above, where a 28 MHz span and 3 kHz RBW result in 9333.33 spectral lines.

I'm attaching a ni5660 example with this change configured for your second test case above. A 28 MHz span and 10 kHz RBW (Bin Width) will result in 2800 spectral lines and allow the f0 to fall exactly at 2 MHz and df to be exactly 10 kHz.

Try it out and let me know if you have any questions.

Regards,
Andy Hinde
National Instruments

Message Edited by Andy Hinde on 03-14-2007 12:17 PM

Message Edited by Andy Hinde on 03-14-2007 12:18 PM