radar pulse generation with interference after the pulse

[babycow]

We are using a RFSG 5670 and a RFSA 5661 to modulate, transmit, receive, and demodulate an IQ signal pulse. To test the equipment, we connect the 5670 to the 5661 directly. But we are finding interferences or artifacts that appear even when the signal pulse has ended. Because our application needs our attention within a microsecond of the generated pulse we need the pulse to be short and clean. What is the possible cause for this kind of behavior? Is there some way to get rid of the artifacts? 

In previous posts Jordan_L provided a Pulse Generating Demo. This demo provided a vi that generated a windowed pulse that had smooth rise and fall edges. There is also a vi that is used to receive and analyze the pulse that the generator created. The link is on the bottom of the post. Using this VI with the settings below, the resulting signal is shown in the attatched picture.

http://forums.ni.com/ni/board/message?board.id=290&message.id=812&query.id=1105901#M812

The generator settings:   pulse width to 0.2us

                                      power level -30dBm

                                      center frequency 440MHz

                                      rise/fall time 50ns

                                      pulse period 20us

                                      IQ rate 100MS/s

                                      samples 1048576 

The analzer settings  :     center frequency 440MHz 

                                      reference power -30dBm

                                      IQ rate 25MS/s

                                      samples 80000

Please help

Thank you 

Darren

[JerryAL]

Hi Darren

I downloaded the VIs from the original posting and tried them out on a NI PXI-5670.  For the analyzer side, I used a NI PXIe-5663.  I did not see the interference signals that you have in your image.

I just realized that the generation VI is not using the correct Power Level Type setting for this type of application.  By default, niRFSG uses average power.  niRFSG looks at the total waveform average power, and then makes hardware settings to get the average power that is set. In your case -30 dBm.  For a pulse, the average power is much less than the peak power.  So, the peak power at the output for the pulse is a lot higher than the -30 dBm set.  It may even be over driving the output of the NI 5670, and definitely overdriving the input to the NI 5661 with a configured reference level of -30 dBm.  In your image, the peak pulse power shows up at -15 dBm.

Can you make a change to the generation VI to set the Power Level Type attribute to Peak Power?  It is one of the RF attributes.  You can add it to the bottom of the property node on the VI’s diagram.

Then please try your test again.  If necessary, ensure that the NI 5661’s reference level is set higher than what it is measuring and see if the interference signals disappear.

Jerry

[babycow]

Hi Jerry thanks for your reply, I've listened to you and changed the average power to peak power. I've also tried to set the reference power level higher. The interferences do disappear. But when the interferences disappear the signals that we want to test also disappears. As in the ppt below. Is there a way to lower the reference level without the interferences or is that the machine limit? I will be needing to test signals within the interference range. Please help. Thank you

- Darren

[JerryAL]

Hi Darren

We have a few issues here to work around.  The first one is dynamic range.  The NI 5661 has about 80 dB of dynamic range.  But this is also influenced by the configured reference level as well as sampling rate.  It’s easier seen when looking at spectrums and FFTs where you configure the Resolution Bandwidth (RB) of the spectrum, and the higher the RB, the higher the noise floor.  There is a relationship of 10 dB of change of noise floor to 10x change in the RB.  Averaging also helps see the actual noise floor of an instrument.

We have some of the same issues when working with IQ data.  Different IQ rates have different associated bandwidths associated with them due to the filters implemented in the digital downconverters creating the IQ data.  A rough calculation says that for every power of ten increase in the acquired bandwidth, the higher the noise floor by a power of 10 dB.  So, for a 25 MSps sampling rate, that corresponds to almost a 70 dB change in the noise floor.  Now, since the reference level is set to about 0 dBm, we also have to factor in the actual device noise floor of about -140 dBm/Hz, as well as the 80 dB dynamic range of the system.  But it is leaving us with about 55 dB dynamic range in your images.

So, you have to optimize the input for your transmitted incident original signal, and still hopefully get your response signal at a much lower power.  If the incident pulse is over driving the input of the VSA, you are going to see artifacts of that from the filters, etc.  To truly optimize the level, you may want to look at the raw IQ data.  During the pulse, you should see good sine waves.  When you overdrive the VSA input, they should start to become distorted.  You can try to optimize the VSA input by setting the lowest Ref. Level that gives you a good sine wave in the pulse part of the acquisition.  Another issue here is the NI 5661 sets its reference level in 10 dB increments.  I’ll refer to the NI 5663 Help file for more information on this.

Now, the reflected signal may still be an issue due to the low power that it is coming back at. 

In typical applications that I have seen that do this type of measurement, they use a circulator to direct the signal energy.  The incident pulse goes in port 1, and most of the power comes out port 2, with an attenuated version out of port 3, which is connected to the VSA and still acquired/recorded.  The reflected pulse comes back in port 2, and usually amplified as well, goes out port 3 to the VSA for acquisition, and an attenuated version out port 1.  But this puts the two signals on the same relative power lever for easy acquisition for analysis.

Jerry

Message Edited by Jerry_L on 08-19-2009 09:29 AM

[JerryAL]

Hi Darren

One extra note, if your signal is repetitive, you could try averaging the zero span spectrum.  That will reduce the noise floor slightly, and should make your reflected signal a little easier to see.

J