I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones, DVB-T sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHz sample rate (I arrange the IFFT input with 1K tones (worth 4 MHz) having QPSK data, then zero out all the central tones, and put remaining QPSK data on the last 1 K tones ( worth the remaining 4 MHz BW).So I get a base-band OFDM spectrum to begin with. I'm then up-sampling the output by 20 , and then filtering with a low-pass interpolation filter. What I do not understand is,I only get the desired output if my LPF filter has cut-off frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not get the output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on each side), why can't I filter at a frequency of 5 or 6 MHZ.
OFDM transmitter
Started by ●August 18, 2010
Reply by ●August 18, 20102010-08-18
On 08/18/2010 08:56 AM, Zeph80 wrote:> I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones, DVB-T > sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHz sample > rate (I arrange the IFFT input with 1K tones (worth 4 MHz) having QPSK > data, then zero out all the central tones, and put remaining QPSK data on > the last 1 K tones ( worth the remaining 4 MHz BW).So I get a base-band > OFDM spectrum to begin with. I'm then up-sampling the output by 20 , and > then filtering with a low-pass interpolation filter. What I do not > understand is,I only get the desired output if my LPF filter has cut-off > frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not get the > output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on each side), > why can't I filter at a frequency of 5 or 6 MHZ.You haven't answered your question already? "If I take my 8MHz wide signal and chop off everything except the middle (or lower) 5 to 6MHz, why do I lose information?". Well gee, I don't know -- why does whacking off over 1/4 of your signal degrade it? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●August 18, 20102010-08-18
>On 08/18/2010 08:56 AM, Zeph80 wrote: >> I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones, DVB-T >> sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHzsample>> rate (I arrange the IFFT input with 1K tones (worth 4 MHz) having QPSK >> data, then zero out all the central tones, and put remaining QPSK dataon>> the last 1 K tones ( worth the remaining 4 MHz BW).So I get abase-band>> OFDM spectrum to begin with. I'm then up-sampling the output by 20 ,and>> then filtering with a low-pass interpolation filter. What I do not >> understand is,I only get the desired output if my LPF filter hascut-off>> frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not get the >> output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on eachside),>> why can't I filter at a frequency of 5 or 6 MHZ. > >You haven't answered your question already? > >"If I take my 8MHz wide signal and chop off everything except the middle >(or lower) 5 to 6MHz, why do I lose information?". > >Well gee, I don't know -- why does whacking off over 1/4 of your signal >degrade it? > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Do you need to implement control loops in software? >"Applied Control Theory for Embedded Systems" was written for you. >See details at http://www.wescottdesign.com/actfes/actfes.html >I'm sorry, I don't understand what you mean. I think the way I'm feeding the inputs to the IFFT results in the 8 MHz being centered around 0. So on each side I have only 4 MHz. So why can't I apply a LPF with cut-off of 5 MHz.
Reply by ●August 18, 20102010-08-18
>>On 08/18/2010 08:56 AM, Zeph80 wrote: >>> I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones, DVB-T >>> sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHz >sample >>> rate (I arrange the IFFT input with 1K tones (worth 4 MHz) havingQPSK>>> data, then zero out all the central tones, and put remaining QPSK data >on >>> the last 1 K tones ( worth the remaining 4 MHz BW).So I get a >base-band >>> OFDM spectrum to begin with. I'm then up-sampling the output by 20 , >and >>> then filtering with a low-pass interpolation filter. What I do not >>> understand is,I only get the desired output if my LPF filter has >cut-off >>> frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not getthe>>> output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on each >side), >>> why can't I filter at a frequency of 5 or 6 MHZ. >> >>You haven't answered your question already? >> >>"If I take my 8MHz wide signal and chop off everything except the middle>>(or lower) 5 to 6MHz, why do I lose information?". >> >>Well gee, I don't know -- why does whacking off over 1/4 of your signal >>degrade it? >> >>-- >> >>Tim Wescott >>Wescott Design Services >>http://www.wescottdesign.com >> >>Do you need to implement control loops in software? >>"Applied Control Theory for Embedded Systems" was written for you. >>See details at http://www.wescottdesign.com/actfes/actfes.html >> >I'm sorry, I don't understand what you mean. I think the way I'm feeding >the inputs to the IFFT results in the 8 MHz being centered around 0. Soon>each side I have only 4 MHz. So why can't I apply a LPF with cut-off of 5 >MHz. >Page 10 of this tutorial http://folk.ntnu.no/htorp/Undervisning/TTK10/IQdemodulation.pdf talks about how the interpolation filter's cut-off frequency for a complex signal should be greater than 0.5fs. I did not understand why, but I guess this related to my question.
Reply by ●August 18, 20102010-08-18
On 08/18/2010 12:02 PM, Zeph80 wrote:>>> On 08/18/2010 08:56 AM, Zeph80 wrote: >>>> I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones, DVB-T >>>> sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHz >> sample >>>> rate (I arrange the IFFT input with 1K tones (worth 4 MHz) having > QPSK >>>> data, then zero out all the central tones, and put remaining QPSK data >> on >>>> the last 1 K tones ( worth the remaining 4 MHz BW).So I get a >> base-band >>>> OFDM spectrum to begin with. I'm then up-sampling the output by 20 , >> and >>>> then filtering with a low-pass interpolation filter. What I do not >>>> understand is,I only get the desired output if my LPF filter has >> cut-off >>>> frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not get > the >>>> output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on each >> side), >>>> why can't I filter at a frequency of 5 or 6 MHZ. >>> >>> You haven't answered your question already? >>> >>> "If I take my 8MHz wide signal and chop off everything except the middle > >>> (or lower) 5 to 6MHz, why do I lose information?". >>> >>> Well gee, I don't know -- why does whacking off over 1/4 of your signal >>> degrade it? >>> >>> -- >>> >>> Tim Wescott >>> Wescott Design Services >>> http://www.wescottdesign.com >>> >>> Do you need to implement control loops in software? >>> "Applied Control Theory for Embedded Systems" was written for you. >>> See details at http://www.wescottdesign.com/actfes/actfes.html >>> >> I'm sorry, I don't understand what you mean. I think the way I'm feeding >> the inputs to the IFFT results in the 8 MHz being centered around 0. So > on >> each side I have only 4 MHz. So why can't I apply a LPF with cut-off of 5 >> MHz. >> > > > > Page 10 of this tutorial > http://folk.ntnu.no/htorp/Undervisning/TTK10/IQdemodulation.pdf talks > about how the interpolation filter's cut-off frequency for a complex signal > should be greater than 0.5fs. I did not understand why, but I guess this > related to my question.That 0.5 Fs is the Nyquist limit, and if you don't know what its about then you have a lot to learn. This may help. It's aimed at someone who knows that there's a Nyquist limit but has misconceptions. You're fortunate, in that you don't seem to have misconceptions to dispell! http://www.wescottdesign.com/articles/Sampling/sampling.html -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●August 18, 20102010-08-18
On 08/18/2010 11:36 AM, Zeph80 wrote:>> On 08/18/2010 08:56 AM, Zeph80 wrote: >>> I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones, DVB-T >>> sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHz > sample >>> rate (I arrange the IFFT input with 1K tones (worth 4 MHz) having QPSK >>> data, then zero out all the central tones, and put remaining QPSK data > on >>> the last 1 K tones ( worth the remaining 4 MHz BW).So I get a > base-band >>> OFDM spectrum to begin with. I'm then up-sampling the output by 20 , > and >>> then filtering with a low-pass interpolation filter. What I do not >>> understand is,I only get the desired output if my LPF filter has > cut-off >>> frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not get the >>> output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on each > side), >>> why can't I filter at a frequency of 5 or 6 MHZ. >> >> You haven't answered your question already? >> >> "If I take my 8MHz wide signal and chop off everything except the middle >> (or lower) 5 to 6MHz, why do I lose information?". >> >> Well gee, I don't know -- why does whacking off over 1/4 of your signal >> degrade it? >> >> -- >> >> Tim Wescott >> Wescott Design Services >> http://www.wescottdesign.com >> >> Do you need to implement control loops in software? >> "Applied Control Theory for Embedded Systems" was written for you. >> See details at http://www.wescottdesign.com/actfes/actfes.html >> > I'm sorry, I don't understand what you mean. I think the way I'm feeding > the inputs to the IFFT results in the 8 MHz being centered around 0.You _think_? Have you _checked_?> So on > each side I have only 4 MHz. So why can't I apply a LPF with cut-off of 5 > MHz.Are you getting a complex signal out of the IFFT? If you're only using the real part then it is impossible for you to have unique spectra on each side of 0Hz. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●August 18, 20102010-08-18
Tim Wescott <tim@seemywebsite.com> wrote: (snip)> "If I take my 8MHz wide signal and chop off everything > except the middle (or lower) 5 to 6MHz, why do I lose information?".> Well gee, I don't know -- why does whacking off over 1/4 of > your signal degrade it?That description sounds much like vestigial sideband modulation, carefully designed not to lose information. -- glen
Reply by ●August 18, 20102010-08-18
>On 08/18/2010 12:02 PM, Zeph80 wrote: >>>> On 08/18/2010 08:56 AM, Zeph80 wrote: >>>>> I'm simulating a DVB-T OFDM 8 Mhz BW transmitter with 2K tones,DVB-T>>>>> sample rate is 9MHz. I'm doing this by using a 4K IFFT and 18 MHz >>> sample >>>>> rate (I arrange the IFFT input with 1K tones (worth 4 MHz) having >> QPSK >>>>> data, then zero out all the central tones, and put remaining QPSKdata>>> on >>>>> the last 1 K tones ( worth the remaining 4 MHz BW).So I get a >>> base-band >>>>> OFDM spectrum to begin with. I'm then up-sampling the output by 20 , >>> and >>>>> then filtering with a low-pass interpolation filter. What I do not >>>>> understand is,I only get the desired output if my LPF filter has >>> cut-off >>>>> frequency of 8 MHz. If I set the cut-off to 5 or 6 MHz I do not get >> the >>>>> output desired. If my 8 MHz signal is centered at 0 ( 4 MHz on each >>> side), >>>>> why can't I filter at a frequency of 5 or 6 MHZ. >>>> >>>> You haven't answered your question already? >>>> >>>> "If I take my 8MHz wide signal and chop off everything except themiddle>> >>>> (or lower) 5 to 6MHz, why do I lose information?". >>>> >>>> Well gee, I don't know -- why does whacking off over 1/4 of yoursignal>>>> degrade it? >>>> >>>> -- >>>> >>>> Tim Wescott >>>> Wescott Design Services >>>> http://www.wescottdesign.com >>>> >>>> Do you need to implement control loops in software? >>>> "Applied Control Theory for Embedded Systems" was written for you. >>>> See details at http://www.wescottdesign.com/actfes/actfes.html >>>> >>> I'm sorry, I don't understand what you mean. I think the way I'mfeeding>>> the inputs to the IFFT results in the 8 MHz being centered around 0.So>> on >>> each side I have only 4 MHz. So why can't I apply a LPF with cut-off of5>>> MHz. >>> >> >> >> >> Page 10 of this tutorial >> http://folk.ntnu.no/htorp/Undervisning/TTK10/IQdemodulation.pdf talks >> about how the interpolation filter's cut-off frequency for a complexsignal>> should be greater than 0.5fs. I did not understand why, but I guessthis>> related to my question. > >That 0.5 Fs is the Nyquist limit, and if you don't know what its about >then you have a lot to learn. > >This may help. It's aimed at someone who knows that there's a Nyquist >limit but has misconceptions. You're fortunate, in that you don't seem >to have misconceptions to dispell! > >http://www.wescottdesign.com/articles/Sampling/sampling.html > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Do you need to implement control loops in software? >"Applied Control Theory for Embedded Systems" was written for you. >See details at http://www.wescottdesign.com/actfes/actfes.html>I did not think I was violating Nyquist by using a filter cut-off frequency lower than 0.5 fs.I thought as long as my sample frequency is twice the signal bandwidth,I can filter my signal with any cut-off frequency as long as it includes the 1 sided spectrum.
Reply by ●August 18, 20102010-08-18
On 08/18/2010 12:28 PM, glen herrmannsfeldt wrote:> Tim Wescott<tim@seemywebsite.com> wrote: > (snip) > >> "If I take my 8MHz wide signal and chop off everything >> except the middle (or lower) 5 to 6MHz, why do I lose information?". > >> Well gee, I don't know -- why does whacking off over 1/4 of >> your signal degrade it? > > That description sounds much like vestigial sideband modulation, > carefully designed not to lose information.Yes, but in vestigial sideband modulation the two sidebands are redundant. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
