symbol rate accuracy for PSK demodulation

[SunJab]

Hi all,

I have a question related to PSK demodulation vi that is the part of NI Modulation toolkit. I know that the vi performs both carrier recovery and symbol clock recovery. And I also know that before feeding the signal to the demodulator's input, it must be resampled at a rate that is an even integer multiple of the symbol rate. My question is that how accurate the symbol rate needs to be inorder to demodulate the signal correctly?

Thanks

Sandy

[Yardov]

Hi Sandy,

This question does not have an exact answer because the received signal on the analyzer might have other impairments that the demodulator also needs to compensate for.

You are correct on your post that the signal must be resample to an even integer multiple of the symbol rate. As you notice on the VI, there is no input for "symbol rate". However, it is infer by the input "samples per symbol". This way, the demodulator will take each X samples and treat them as a symbol to estimate the exact symbol rate.

If we are clear on this, then the only way to test symbol rate accuracy is to test samples per symbol accuracy or better yet, non-integer number of samples per symbol. For this, you can resample your signal to a different sample rate and look at the results you care.

For example, I have below a graph where I modify the symbol rate of the signal and look at EVM. As you can see, you have to be VERY close to the exact number (in this case 100k) to get the best EVM.

 

I have included the code in case you want to play more with different values.

Hope this point you in the right direction to find what you are looking for.

[SunJab]

Hi,

Thanks for the help. I am also having problem understanding the PSK demodulator's VI that computes the optimal sample point. If you look into its block diagram, it computes the square of the incomming signal magnitude, Applies a triangular window and then computes the one point DFT before finding the sampling offset. I want to know the mathematical meaning of one point DFT? what does a one point DFT represents?

Thanks

Sandy

[Yardov]

Hello Sandy,

Basically the one point DFT is finding the phase of the symbol rate (i.e. the optimum sample point). This is similar to perform the eye diagram and choose the point of highest power.

There are several text books about these algorithms. "Digital Communications" from Proakis chapter 6 talks about this.

If you have more questions, please include some information of the context such as your application or what exactly are you concern with.

[SunJab]

I have to demodulate some PSK signals that I have recorded through niRFSA. These signals are at baseband and are stored in a file with IQ 16bit storage format. I read the I16 values from the file, type cast them to CDBL and feed them to the demodulator (after resampling to samples per symbol * symbol rate). But I am not able to demodulate these signals.

I have tested the demodulator by generating PSK signals using the modulation toolkit VIs and its working fine on those signals even if I introduce a large frequency offset.

I want to know what are the things that I need to care for before feeding the signal to the demodulator? Is there any requirement for symbol rate that it needs to be known accurately? Any requirement for the matched filter length etc ?

Thanks,

Sandy

[Yardov]

Hello Sandy,

Are you scaling your data correctly? When recorded, there is an output from the niRFSA Fetch that is called "Waveform information". You should use the gain value to scale your I16 data correctly. I have attached a VI to do this if you know the gain.

Could you upload a small section of this data with some of the demodulation parameters. That will help the audience to understand better what's going on. 

There are some possibility that  the data was recorded with a very poor dynamic range and therefore is not your frequency offset that is causing the problem. 

[SunJab]

Yes I am scaling the I16 data correctly using the gain value from 'waveform info' cluster.

[Yardov]

Hi Sandy, 

What about dynamic range? Do you remember your saturation level when recording? 

This is the time when uploading a piece of your recorded data will help the audience give you a more clear answer of what you are looking for.

Regards,