Different spectrum from signalexpress with oscilloscope

[myo]

Hi,

I am new to signal processing. I am facing some difficulties in measuring the signal from an acoustic emission sensor for my project. I am using PXI-6115 with 1042Q and terminal block is TB2708. I used AI0 and AI1 for main signal and trigger respectively. I acquired signal from 6115 in SignalExpress (v3.0) and convert to linear spectrum (Hanning window and RMS conversion) with RMS averaging at 200. Same settings are used in oscilloscope (LeCroy LC564DL) too. I connected both signals to oscilloscope as well for comparison. I saved the spectrum data from SignalExpress and oscilloscope and plotted. You can see the graph in attachment. It is totally different. Why is that so?

And one more thing is if I disconnect my connector to NI DAQ system, the spectrum in oscilloscope changed in magnitude at certain frequency and vice versa.

Thank you in advance.

Myo

[myo]

Hi,

I forgot to mention that rate is 2MHz and samples to read is 100k for both.

Myo

[DFGray]

It appears your X-axis units are different.  I would guess the SignalExpress one is in dB (relative log scale) and the oscilloscope one is linear.

[myo]

Thanks for reply, DFGray.

No. I set linear scale in SignalExpress with magnitude setting. I don't know what makes them different.

myo

[DFGray]

Here are a few things to look for:

1. Is the veraging being done on the spectrum or the initial signal?  In SignalExpress, RMS averaging means average the spectra, not the signal use vector averaging to average the signals).  If the LC564DL is averaging the signals, then taking a spectrum, unless you have good triggering, this will wash out the fundamental and totally eliminate higher frequencies (which is exactly what you are seeing).
2. The range of the signals is very different.  Is the LC564DL set to 50Ω input impedance?  This could easily create the difference you see, especially if your output device has a high impedance output. Check your device output.
   
3. The 6115 is a 12-bit device, the LC564DL is an 8 bit device.  This should not make a difference on the scale you are seeing, provided both devices are utilizing the full input scale.  Is the input signal almost filling the digitizing range of the device in both cases?  If the LC564DL only has a small signal on the screen, the lower level components could be disappearing in digitizing errors

If none of these is the issue, let us know and we can brainstorm more.  Good luck.

Message Edited by DFGray on 11-09-2009 08:06 AM

[myo]

Thanks for your reply, DFGray.

The averaging was done in on spectrum with RMS averaging (200). I didn't do on signal. But for oscilloscope I did FFTAvg with 200 sweeps. Does it make difference? And the coupling impedence for LC564DL was set as 1MΩ  then . It was not 50Ω. How should I check my output impedence? I am sorry if the question is quite basic.

One more thing, I tested with sine wave signal (freq 100kHz, amp 1V, offset 0V, Sampling rate 10Mhz, samples 20k)from function generator and coverted to FFT power spectrum (Power)(dB) with RMS averaging (200counts, linear), I found about -100dB(ref 1V) at 100kHz in Signal Express while LC564DL gave me about -70 dBm with FFTAvaraging (200 sweeps). Are my settings wrong or something?

Then, I created sine wave in software(SignalExp) and converted to FFT power spectrum with same settings as above, it gave -350dB(ref 1V) at 100kHz. Why is that so? It should give same results, right?

I am quite new to signal processing, so please help and advise me.

Thank you again.

Myo

[myo]

Now i know the difference of dBm and dB which have different reference electric power. so, it is the same results for daq and oscil. But I am still confused why the simulated signal in signal express can't give same result?

Please any help?

Thanks.

[DFGray · Accepted solution]

My apologies for the late reply. I have been out sick for a couple of days.

I generated a 1V amplitude (2V Pk-Pk), 100kHz, 10MS/s, 20ksample sine wave using Create Analog Signal.  I processed this with Power Spectrum using 200 linear averages with RMS algorithm.  The value at 100kHz was -3dB, as expected and as it should be.

However, at one point in the process, SignalExpress coerced my frequency from 100kHz down to 50kHz (probably due to a mismatch in frequency and number of points).  This would result in what you are seeing.  Check your project to see if this happened to you.  If it did, you would get -350dB at 100kHz (essentially the noise floor of a pure signal), and -3dB at 50kHz.

[myo]

Thanks a lot for your reply. DFG.

Yes. SignalExpress coerced my frequency from 100kHz down to 50kHz in my project and it is always showing -350dB no matter I changed back to 50 kHz. I don't know why. I tested with Labview and found out the same too. I attached my VI too.

Take care.

Myo

[DFGray]

You can also change the frequency in your sine generator back up to 100kHz.  It was originally coerced due to either too few points or too small a sampling frequency.  If you have those set to 200ksamples and 4MS/s respectively, it will allow you to set the frequency to 100kHz.