AWG sampling rate and analog bandwidth

[sabe]

 I was going through an implementation of OFDM on National instruments by University of Texas  In it a 64 tone OFDM was taken it was split into half and then 336 zeros were inserted in between and a 400 point FFT was taken. Then the last 100 complex values were inserted in the beginning of every symbol as cyclic prefix. The result was a 500 complex values . With 336 zeroes,64 data values and 100 cyclix prefix values. then many OFDM symbols were grouped together and transmitted

After sampling at a rate of 100 MS/s at the Arbitrary Waveform generator(baseband to IF conversion) the signal of 16 MHz bandwidth would be generated. Then at the receiver the signal was sampled at a rate of 64 MS/s with 16 Mhz filter bandwidth and then decimated by 4 to get 80 complex values per OFDM symbol,64 of data and 16 of cyclic prefix.

I dont understand how the signal would carry 16 MHz bandwidth and how the reception and then decimation would yield 80 complex values with 64 data values and 16 cyclic prefix length..

[Marcus_M]

Hi,

Would you please give me a little more information about your setup, specifically the hardware you are using and the type of test that you are running.

Thank you.

Regards,

Marcus 

[sabe]

I am using PXI 5670 as RFSG and PXI 5660 as RFSA. I am trying to implement OFDM  system .

[Vimal_F]

Hello sabe,

Are you trying to replicate the exact setup that University of Texas had? If not, what are you trying to do? Can you please describe your application in greater detail. If you are trying to build an OFDM system here are some examples to help you get started:

http://zone.ni.com/devzone/cda/epd/p/id/3456

http://zone.ni.com/devzone/cda/tut/p/id/2831

Note that you need the Modulation Toolkit to use these examples: http://sine.ni.com/nips/cds/view/p/lang/en/nid/12855

Thank you,

Vimal F

[sabe]

Yes for now i will replicate the exact setup of University of Texas . I have gone through this example of OFDM and I also have the modulation toolkit .I have understood the OFDM well,  I just had difficulty in understanding how the sampliing rate is choosen and how the whole system has bandwidth of 16 MHz.

This is their implementation http://www.profheath.org/wp-content/uploads/2010/06/SISO-SysPrototype.zip

This is what is written in their project software desription

Our arbitrary waveform generator (ARB) operates at 100MSamples/s.  As a result, our 16MHz signal must be correctly upsampled in order to be generated correctly.  We achieve this by zero padding our signal in the frequency domain before we take the inverse fast fourier transform of our OFDM symbol.  We zero pad by adding 336 samples to our frequency domain representation so that the total number of tones representing our OFDM symbol is really 400, where 336 of these tones are zeroed out.  This gives us the correct 16MHz bandwidth for our ARB which transmits at 100MSamples/s.

Now what I seem to understand from this is that their ARB could not operate at 16 MS/s  and they had to transmit at 100 MS/s . So to transmit their signal which was 64(OFDM symbols)  by an ARB running at 100 MS/s they had to add some extra samples in it which was zero padding. But I am still unsure I hope someone can clarify the point.

[Vimal_F]

Hi,

Yes you are correct, they had to pad the wavefrom with zeros so they could sample at 100 MS/s and still achieve the the 16 MHz bandwidth. 

Also, have you seen the file I have attached? 

Thank you,

Vimal F

[sabe]

Thanks a lot for your reply.

Yes I have read this document. So if I wanted to use 512 tone OFDM then I ll have to pad it so that it becomes 512 x (100/16) = 3200 complex values. Which means i ll have to pad 3200 - 512 =2688 zeros in it to sample it at 100MS/s and still have 16 MHz bandwidth.

Or can I just specify the arb I/Q rate as 16MS/s and dont have to worry about any zero passing .

[Vimal_F]

Hello,

The first method seems to be consistent with U of T's setup. However, I would try both setups to see which one is more efficient. U of T's paper was dated 2004, and I feel a lot has changed since then.

Thank you,

Vimal F

[sabe]

but is  the second method (i.e choosing sampling rate of 16 MS/s without zero padding) applicable regardless of efficiency ??

[Vimal_F]

Please note that 16MS/s is different than 16 MHz that they were trying to transmit at. To get from MS/s to MHz you must divide MS/s by S/cycle. The PXI-5670 provides true 16-bit resolution arbitrary waveform generation at 100 MS/s and 20 MHz real-time bandwidth. It also can generate frequencies from 250kHz to 2.7 GHz. So, you cannot set a sampling rate of 16 MS/s.