Hi, I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband simulator in matlab and will eventually add in equalizer and decoder for AWGN and fading channels. I have 2 question regarding the ADC requirements: 1)What is the minimum # of ADC bits needed to represent the IQ baseband signals to demodulate correctly? 2)What is the optimum AGC setting under AWGN /Fading environment? What are the effect of quantization noise and clipping noise at the ADC output? Thanks, Pat
How to decide the number of ADC bits needed for baseband receiver?
Started by ●October 25, 2006
Reply by ●October 25, 20062006-10-25
pth10312002@yahoo.com wrote:> Hi, > I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband > simulator in matlab and will eventually add in equalizer and decoder > for AWGN and fading channels.Good for you. I have 2 question regarding the ADC> requirements: > 1)What is the minimum # of ADC bits needed to represent the IQ baseband > signals to demodulate correctly?One bit.> 2)What is the optimum AGC setting under AWGN /Fading environment?The optimum is not using the AGC at all. What> are the effect of quantization noise and clipping noise at the ADC > output?It sucks.> Thanks, > PatThe answers are the same kind as the questions are. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●October 26, 20062006-10-26
On 25 Oct 2006 10:19:38 -0700, pth10312002@yahoo.com wrote:>Hi, >I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband >simulator in matlab and will eventually add in equalizer and decoder >for AWGN and fading channels. I have 2 question regarding the ADC >requirements: >1)What is the minimum # of ADC bits needed to represent the IQ baseband >signals to demodulate correctly?That depends on how much processing gain you have between the ADC and the detector/slicer.>2)What is the optimum AGC setting under AWGN /Fading environment? What >are the effect of quantization noise and clipping noise at the ADC >output?Usually one builds a slicer that assumes the signal is gain controlled, and then the job of the AGC is to maintain that signal level. How that is done is completely implementation dependent, i.e., up to the architect. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
Reply by ●October 26, 20062006-10-26
>Hi, >I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband >simulator in matlab and will eventually add in equalizer and decoder >for AWGN and fading channels. I have 2 question regarding the ADC >requirements: >1)What is the minimum # of ADC bits needed to represent the IQ baseband >signals to demodulate correctly?You can include a quantizer in the simulation model and evaluate how the number of ADC bits affect performance.>2)What is the optimum AGC setting under AWGN /Fading environment? What >are the effect of quantization noise and clipping noise at the ADC >output?AGC should set the signal level to full scale at ADC input with some headroom, depending on the PAPR(peak to average ratio) of the signal you use.>Thanks, >Pat > >
Reply by ●October 27, 20062006-10-27
mguo wrote:> >Hi, > >I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband > >simulator in matlab and will eventually add in equalizer and decoder > >for AWGN and fading channels. I have 2 question regarding the ADC > >requirements: > >1)What is the minimum # of ADC bits needed to represent the IQ baseband > >signals to demodulate correctly? > > You can include a quantizer in the simulation model and evaluate how the > number of ADC bits affect performance. > > >2)What is the optimum AGC setting under AWGN /Fading environment? What > >are the effect of quantization noise and clipping noise at the ADC > >output? > > AGC should set the signal level to full scale at ADC input with some > headroom, depending on the PAPR(peak to average ratio) of the signal you > use. > > >Thanks, > >Pat > > > >and unless you are doing all your filtering in the analog domain, the A/D converter will also be presented with "loaded with" the other undesired signals that are in the anlog passband... in general you want to keep the PEAK of the total of all the signals in the A/D below clipping in the A/D and you want to keep the desired signal enough above 1 LSB so that the quantizing noise floor is not degrading the performance too much...
Reply by ●October 28, 20062006-10-28
On 27 Oct 2006 16:32:48 -0700, "Mark" <makolber@yahoo.com> wrote:> >mguo wrote: >> >Hi, >> >I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband >> >simulator in matlab and will eventually add in equalizer and decoder >> >for AWGN and fading channels. I have 2 question regarding the ADC >> >requirements: >> >1)What is the minimum # of ADC bits needed to represent the IQ baseband >> >signals to demodulate correctly? >> >> You can include a quantizer in the simulation model and evaluate how the >> number of ADC bits affect performance. >> >> >2)What is the optimum AGC setting under AWGN /Fading environment? What >> >are the effect of quantization noise and clipping noise at the ADC >> >output? >> >> AGC should set the signal level to full scale at ADC input with some >> headroom, depending on the PAPR(peak to average ratio) of the signal you >> use. >> >> >Thanks, >> >Pat >> > >> > > >and unless you are doing all your filtering in the analog domain, the >A/D converter will also be presented with "loaded with" the other >undesired signals that are in the anlog passband... > >in general you want to keep the PEAK of the total of all the signals in >the A/D below clipping in the A/D and you want to keep the desired >signal enough above 1 LSB so that the quantizing noise floor is not >degrading the performance too much...Theoretically the desired signal can be below 1 LSB and still be recoverable if you have enough processing gain (and enough entropy in the total composite power spectrum). Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
Reply by ●October 29, 20062006-10-29
Eric Jacobsen wrote:> On 27 Oct 2006 16:32:48 -0700, "Mark" <makolber@yahoo.com> wrote: > > >>mguo wrote: >> >>>>Hi, >>>>I am planning to build a BPSK/QPSK/8PSK/16QAM/32QAM/64QAM baseband >>>>simulator in matlab and will eventually add in equalizer and decoder >>>>for AWGN and fading channels. I have 2 question regarding the ADC >>>>requirements: >>>>1)What is the minimum # of ADC bits needed to represent the IQ baseband >>>>signals to demodulate correctly? >>> >>>You can include a quantizer in the simulation model and evaluate how the >>>number of ADC bits affect performance. >>> >>> >>>>2)What is the optimum AGC setting under AWGN /Fading environment? What >>>>are the effect of quantization noise and clipping noise at the ADC >>>>output? >>> >>>AGC should set the signal level to full scale at ADC input with some >>>headroom, depending on the PAPR(peak to average ratio) of the signal you >>>use. >>> >>> >>>>Thanks, >>>>Pat >>>> >>>> >> >>and unless you are doing all your filtering in the analog domain, the >>A/D converter will also be presented with "loaded with" the other >>undesired signals that are in the anlog passband... >> >>in general you want to keep the PEAK of the total of all the signals in >>the A/D below clipping in the A/D and you want to keep the desired >>signal enough above 1 LSB so that the quantizing noise floor is not >>degrading the performance too much... > > > Theoretically the desired signal can be below 1 LSB and still be > recoverable if you have enough processing gain (and enough entropy in > the total composite power spectrum).Oh, thank goodness. That means the radars I used to build can work after all. :-) Steve
