5660 accuracy for hopping signal

[adnan_fazil]

Hi,

I am doing my MS thesis in which I am working on real-time detection of hopping frequency. I am working in frequency band of 20 MHz (220-240 MHz).The available hardware is:

• PXI Chassis 1044

• Embedded Controller: NI PXI 8106

• NI PXI-5660 (PXI-5600 and PXI-5620)

I have some queries listed below:

1. What is the maximum number of hops per second that can be detected with available hardware?

2. What is the minimum time required to detect an active hoped frequency with available hardware?

3. What is the frequency detection accuracy with available hardware?

4. What is the maximum FFT computation rate with available hardware?

REGARDS

[Eric_K]

adnan fazil,

The number of hops per second that you will be able to detect is going to depend greatly on how you set up your acquisition.  The biggest factor of this is the resolution bandwidth that you choose, which is going to affect how long it takes to acquire the information across the full 20 Mhz.  Of coarse, the resolution bandwidth will also determine how small of changes you will be able to detect in the signal.  If the changes you are looking to detect are farther apart, you will be able to scan the 20 Mhz band for hops much more quickly than if your hops are small and you need to see a lot of detail in the signal.

The FFT rate that you car run at is going to be very software dependent as well.  Since the math will all be done in software, it will depend on how your code is written, and what else is running on the system while you are calculating the FFT.

Hope this helps,

Eric K

Applications Engineering

National Instruments

[adnan_fazil]

Hi ERIC

        thanks for replying..i am analysing 20Mhz BW(220-240MHz) with resolution bandwith equal to 78Mhz, which gives me FFt size of 2048 points.My question is with this RBW what is the maximum hoping rate that can b detected by 5660 with out missing any hop.

Another query is the number of averages (an input to the block "ni5660 read average power spectrum").what is the effect of this number on the above question....and what will b the answer for the first question if the number of averages is set to 10.???

THANKS IN ADVANCE

ADNAN

[Chris_W.]

Adnan,

There are a couple of specs that you mention that should probably gets squared away before you can identify exactly what kind of times you're looking at. First, and most important, if you're only looking at a signal that is going to be hopping around in a 20MHz span, then the speed at which you can effectively see the frequency hops is going to be totally determined by your FFT size, which will determine how many points you acquire and therefore, the total time of acquisition per acquired frequency.

This being said, Eric is right that the resolution bandwidth will also be a limiting factor that is going to determine the size of your FFT, however, the resolution bandwidth only needs to be small enough to define the smallest jumps in frequency that the signal you're monitoring is going to be making. So, if the smallest resolution of jumps is 1kHz, then your RBW shouldn't need to be too much smaller than that.

In your last post you say that your RBW is 78MHz? That is larger than the max capable RBW for this device, which is 10MHz. Did you mean 78Hz? if this is the case, then I don't think that the FFT size would be 2048. You should double check these numbers. Once you do figure out how large of an FFT you're using, you should be able to calculate how long it takes the 5660 to acquire that by dividing the number of samples in your FFT by 64M Samples/s, which is the sample rate of the 5620 digitizer.

So, if your FFT size was 2048, then it would take the calculation would be:

2048/64M = 32milliseconds

I hope this helps.

Chris

64 MS/s

[adnan_fazil]

Thanks CHRIS

your ans is very useful for me....and sorry for mistake...the RBW i m using is 78kHz. with 10 points average....and 2040 points FFT...

REGARDZ

ADNAN FAZIl

[adnan_fazil]

Thanks Chris

 you have written"So, if your FFT size was 2048, then it would take the calculation would be: 2048/64M = 32milliseconds".

But i think
2048/64M = 32microseconds
kindly comment on this.

Thanks in advance .

REGARDS
ADNAN FAZIL

[Chris_W.]

Oops, you're correct. That was a typo. 32µs was the correct calculation for my example.

I hope that information helps clarify things, regardless of my typo.

Chris

[adnan_fazil]

Thanks chris.

First i thought that the mistak is on the calculation side and the result i.e 32ms is right, becz my vi is almost running at this speed.

But i m surprized to know that the acquisition time is 32us because my vi is running at about 40-50 times(almost 20ms for each loop) a second with 2040 FFT points (and RBW=78KHz).

i m using ""NI PXI chassis 1044"" with embedded controller "NI PXI-8106", my analyser is PXI-5660 (PXI-5600 and PXI-5620), and i m generating a sweeping or u can say a hopping signal with "NI PXI-5652" with LABVIEW 8.2.

I have to detect the frequency(FFT) and then save it in a file to make the record and start recieving the next frequency(FFT) ,but the limitation is the time for this process as target in my thises is to do this in less than 250us and then start acquisition for the next frequency(FFT).

I m trying to do this by detecting the frequency band of 20MHZ (from 220MHz-2240MHz) and saving the resuting peakes in the FFT in an ecxel file.

when i set the dwell time of 100ms in the generators(PXI-5652) VI the excel file show that the 5660 read each frequency 4-5 times that is about 20-25ms for each FFT but my target is 250us .I m attaching the generator VI and my analyzer VI along with an excel file for ur reference .I think i m doing some mistakes as the acuisition time for 2040 FFT points is only 32us and rest is just calculation in software...

EXPECTING some useful comments from ur side as before...:)

Even if the the target specs are impossible to echieve with PXI-5660, what is the best that can be done with PXI-5660 for samallest possible time and best ferquency accuracy. 

The files attached will also clarify the problem if i m not able to do this :)..

NOTE:I M USING THE PXI-5652 (my generator card) and PXI-5660 (my analyser) in a PXI CHASSIS WITH SINGLE PXI-8106 CONTROLLER with windows XP....

THANKS IN ADVANCE

REGARDS

ADNAN FAZIL

[adnan_fazil]

Thanks chris.

First i thought that the mistak is on the calculation side and the result i.e 32ms is right, becz my vi is almost running at this speed. But i m surprized to know that the acquisition time is 32us because my vi is running at about 40-50 times(almost 20ms for each loop) a second with 2040 FFT points (and RBW=78KHz).

 i m using ""NI PXI chassis 1044"" with embedded controller "NI PXI-8106", my analyser is PXI-5660 (PXI-5600 and PXI-5620), and i m generating a sweeping or u can say a hopping signal with "NI PXI-5652" with LABVIEW 8.2.

 I have to detect the frequency(FFT) and then save it in a file to make the record and start recieving the next frequency(FFT) ,but the limitation is the time for this process as target in my thises is to do this in less than 250us and then start acquisition for the next frequency(FFT).

 I m trying to do this by detecting the frequency band of 20MHZ (from 220MHz-2240MHz) and saving the resuting peakes in the FFT in an ecxel file. when i set the dwell time of 100ms in the generators(PXI-5652) VI the excel file show that the 5660 read each frequency 4-5 times that is about 20-25ms for each FFT but my target is 250us .I m attaching the generator VI and my analyzer VI along with an excel file for ur reference .I think i m doing some mistakes as the acuisition time for 2040 FFT points is only 32us and rest is just calculation in software...

 EXPECTING some useful comments from ur side as before...

 Even if the the target specs are impossible to echieve with PXI-5660, what is the best that can be done with PXI-5660 for samallest possible time and best ferquency accuracy.

 The files attached will also clarify the problem if i m not able to do this ..

NOTE:I M USING THE PXI-5652 (my generator card) and PXI-5660 (my analyser) in a PXI CHASSIS WITH SINGLE PXI-8106 CONTROLLER with windows XP....

THANKS IN ADVANCE

REGARDS

ADNAN FAZIL

[Chris_W.]

Hi Adnan,

The 32µs is not a miscalculation, but this does not take into account the processing time and time for buffering the data to the computer, which does the processing. 32µs is the amount of time that it takes for the 5620 to acquire 2040 points, but this data is then being buffered/sent to the computer and the read from a RAM buffer on the PC when you call the "ni5660 Read Averaged Power Spectrum.vi." This all takes some extra time. The other thing that you're doing is reading an averaged power spectrum with 10 averages, which means you're acquiring 10 sets of 2040 points, before the Read Averaged Power Spectrum VI returns it's data. 

The other thing to note, is that the code that does all of this in your while loop can only run so fast, which is going to depend on the speed of your processor. I would say that it's likely your loop rate is not going to be too much better than a few milliseconds at very best, so I'm not sure that 250µs is going to be attainable, if you have to do the processing in line with the acquisition. 

You can make some optimizations to your code that will speed up the loop rate you currently have. First, you should pull the ni5660 Configure for Spectrum.vi out of the while loop to the left of it. Currently you are redundently configuring your acquistion settings each time you take a read.

Also, instead of using the ni5660 Read Averaged Power Spectrum.vi, you should use the ni5660 Read FFT Spectrum.vi, which will take out the averaging and give you an FFT from only one 2040 point data set. Along with this, you can take out the math that you're doing to find the peaks and use the ni5660 Peak Search.vi, which will return the information that you're looking for. Another thing that will speed up your loop is not displaying the data. 

I made a modified version of your code and posted an image of it. I don't have a specific benchmark for you that you'll be able to run the code at, but I would say that your loop rate won't exceed about 1kHz maximum.

I hope this helps.

Chris