- bittware November 29, 2018 at 10:53 pm
I was using SA44B in real-time mode to capture 5us pulse at 10dBm with very low duty cycle of 0.01%(i.e. period of 50mS). The carrier frequency is 3GHz.
I was expecting SA44B could capture such narrow pulse event in real-time mode.
But in reality with all default setting at center 3GHz, the real-time spectrum plot showed jumping peak at -80dBm~-60dBm. Until the pulse width was increased to approx. 200uS, the plot started to show stable peak at -40dBm at carrier frequency.
I also noticed that the sweep time of real-time seems to be fixed at about 30-32mS range. This is obviously too large and atennuates too much in frequency domain to 5uS signal.
My question is: did I do something wrong that led to such result?
or if SA44B has no capability of analysing this kind of signal?
If it’s the limit of SA44B, can I hope BB60C could fix this problem?
Thanks for any comments.
Cory AllenKeymasterCory Allen November 30, 2018 at 10:22 am
The engineers that typically monitor this forum are traveling back from a trade show today, so there may be a delayed response time. I can assure you that as soon as they have access you’ll get an answer.
Just wanted to let you know.
AndrewModeratorAndrew December 3, 2018 at 8:39 am
The maximum sample rate of the SA44B at the IF is ~486kS/s, which means a 5us pulse would only appear in ~3 samples. Generally in real-time modes, the energy of a signal needs to be present in 1.25 FFTs worth of data. In fact the Spike software even tells you the signal duration required for 100% probability of intercept (for the signal to be represented on display within 3dB). At the highest RBW, the 100% POI requirement of a signal is 592us. Once signals start becoming shorter than this, the energy level displayed will decrease.
In short, the SA44B is not the best tool to measure this type of signal using real-time mode.
What you can try, is using zero-span mode to capture the pulse in the time domain. A 5us signal can be captured in this mode. I have included a picture of this configuration with my SA44B. I used a 5us pulse with 80us period to demonstrate. The spectrum plot in the zero-span mode will still not be accurate, but you can get a good idea what the pulse looks like in the time domain.
The BB60C or SM200A would be better analyzers to measure this type of signal in real-time, the BB60C has a minimum 100% POI of 4.8us. As you have noticed though, the sweep time of our real-time mode is ~30ms and will also be that for the BB and SM. We accumulate ~30ms worth of spectrum before displaying the results in real-time which means you will see the accumulation of several pulses in your config in one sweep. At this point we are unable to have a finer resolution than this.
Attachments:You must be logged in to view attached files.bittware December 3, 2018 at 6:51 pm
Thanks for the advice.
Based on your description, I’ve given up the SA44B.
Zero-span mode has little help to my application, as it only provides meaningful info. in time domain whereas my interest is the central frequency drifting performance of the carrier.
Let’s focus on the BB60C in real-time mode, I still have following concerns:
1. What is the IF sampling rate of BB60C?
2. Could you elaborate on “1.25 FFTs worth of data”? I guess it means 1.25 times the min. point of FFT which is (?) Would you do multiple overlapped FFTs within the same sweep time window?
3. When you mentioned “We accumulate ~30ms worth of spectrum before displaying the results in real-time which means you will see the accumulation of several pulses in your config in one sweep. ” I guess capturing multiple pulses within the same sweep time window is a good thing. Two seperate 5uS pulses is equivalent to one consecutive 10uS in terms of spectrum resolution providing the two pulses have the identical spectrum content. Am I correct?
Looking forward to your feedback.
Thanks a lot.
AndrewModeratorAndrew December 4, 2018 at 8:01 am
1) The sampling rate is 80MS/s IF which we convert to 40MS/s IQ.
2) Yes, we do overlapped FFTs. I apologize, 1.25 is not necessarily the exact number, it’s going to depend on the overlap amount. The idea being the same though, the energy needs to be present in a significant portion of the FFT to show up at the correct amplitude on the spectrum.
3) It depends on what you need from the measurement. This will also mean that the waterfall only has 30ms resolution, which some people might want a smaller resolution, in your case see activity in the waterfall between the pulses.
Regards,bittware December 4, 2018 at 3:23 pm
Then what is the min. amount of points that each FFT does? Is it a fixed number? Or if it could be adjusted what’s the benefit of having FFT set to a lower amount of points?
In fact, I don’t care the activity changes in between pulses, because it’s certain that the signal is totally shut off when there is no pulse generated and the pulse’s central frequency won’t drift within such short period of time. That why I was asking if multiple pulses show up in the same sweep could do any help to the central frequency discrimination in the spectrum plot.
Could you explain more on this? Thanks.
Justin Crooks December 4, 2018 at 5:29 pm
- This reply was modified 1 week, 6 days ago by Justin Crooks.
I just wanted to jump in and ask if you had considered looking at the pulse in zero span mode with a video trigger? The SA44B won’t give you a great plot due to resolution and bandwidth limitations, but the BB60C would give you a great plot of your pulse’s amplitude over time.bittware December 4, 2018 at 7:35 pm
I’ve only tried zero-span mode with SA44B with video trigger level at 0dBm. Although the pulse energy is at 10dBm, you can barely recognize the 5uS envelope.
But I’m not interested in the absolute accuracy of the amplitude of the pulse. Rather, the central freqency measurement accuracy really matters in my application.
As my signal source has high Q value of about 15000.
Even a few 10KHz deviation of the peak central frequency detection is not acceptable.
I’m evaluating if BB60C is capable of delivering such central freqency measurement accuracy for the 5uS pulse.
Justin CrooksModeratorJustin Crooks December 5, 2018 at 9:57 am
The BB60C would be the minimum for this test (SM200A would be better). Zero span amplitude vs time and frequency vs time plots, probably with a decimation of 1 for pulse shape, and decimation of 8 to 64 for frequency measurements would be ideal (higher decimation would reduce the noise). You’re basically just looking at the phase change on each I/Q sample, converted to frequency.
You must be logged in to reply to this topic.