Dear Frieds, I am facing a problem while finding SNR of a adc output. I am giving sinusoidal wave at the input amd getting sinusoidal wave with harmonics and noise at output. Now, to calculate power for my fundamental frequency and noise, I am using DFT. But as I am changing the number of samples, I am getting different-different SNR. Here, I am also considering spectrum leakage in FFT. So, as my fundamental frequence is 122070312.5Hz (122.07MHz) and I am sampling it with 25GHz, I am taking number of samples 1024 and 2048. I am getting 5db SNR with 1024 samples and 14db SNR with 2048 samples. Please help me find where am I going wrong ? Thanks
Problem in SNR
Started by ●February 8, 2008
Reply by ●February 8, 20082008-02-08
On Feb 8, 6:17 am, "sunil_ic" <sunil...@gmail.com> wrote:> Dear Frieds, > I am facing a problem while finding SNR of a adc output. I am giving > sinusoidal wave at the input amd getting sinusoidal wave with harmonics > and noise at output. Now, to calculate power for my fundamental frequency > and noise, I am using DFT. But as I am changing the number of samples, I > am getting different-different SNR. Here, I am also considering spectrum > leakage in FFT. So, as my fundamental frequence is 122070312.5Hz > (122.07MHz) and I am sampling it with 25GHz, I am taking number of samples > 1024 and 2048. I am getting 5db SNR with 1024 samples and 14db SNR with > 2048 samples. Please help me find where am I going wrong ? > Thanks25 GHz ???????
Reply by ●February 8, 20082008-02-08
On Fri, 8 Feb 2008 03:50:32 -0800 (PST), John <sampson164@gmail.com> wrote:>On Feb 8, 6:17 am, "sunil_ic" <sunil...@gmail.com> wrote: >> Dear Frieds, >> I am facing a problem while finding SNR of a adc output. I am giving >> sinusoidal wave at the input amd getting sinusoidal wave with harmonics >> and noise at output. Now, to calculate power for my fundamental frequency >> and noise, I am using DFT. But as I am changing the number of samples, I >> am getting different-different SNR. Here, I am also considering spectrum >> leakage in FFT. So, as my fundamental frequence is 122070312.5Hz >> (122.07MHz) and I am sampling it with 25GHz, I am taking number of samples >> 1024 and 2048. I am getting 5db SNR with 1024 samples and 14db SNR with >> 2048 samples. Please help me find where am I going wrong ? >> Thanks > >25 GHz ???????Not infeasable on a newer oscilloscope. Regards, Allan
Reply by ●February 12, 20082008-02-12
>On Fri, 8 Feb 2008 03:50:32 -0800 (PST), John <sampson164@gmail.com> >wrote: > >>On Feb 8, 6:17 am, "sunil_ic" <sunil...@gmail.com> wrote: >>> Dear Frieds, >>> I am facing a problem while finding SNR of a adc output. I am giving >>> sinusoidal wave at the input amd getting sinusoidal wave withharmonics>>> and noise at output. Now, to calculate power for my fundamentalfrequency>>> and noise, I am using DFT. But as I am changing the number of samples,I>>> am getting different-different SNR. Here, I am also consideringspectrum>>> leakage in FFT. So, as my fundamental frequence is 122070312.5Hz >>> (122.07MHz) and I am sampling it with 25GHz, I am taking number ofsamples>>> 1024 and 2048. I am getting 5db SNR with 1024 samples and 14db SNRwith>>> 2048 samples. Please help me find where am I going wrong ? >>> Thanks >> >>25 GHz ??????? > >Not infeasable on a newer oscilloscope. > >Regards, >Allan >Its not Oscilloscope setup. I am doing matlab simulation.
Reply by ●February 12, 20082008-02-12
On Feb 12, 7:10�am, "sunil_ic" <sunil...@gmail.com> wrote:> >On Fri, 8 Feb 2008 03:50:32 -0800 (PST), John <sampson...@gmail.com> > >wrote: > > >>On Feb 8, 6:17 am, "sunil_ic" <sunil...@gmail.com> wrote: > >>> Dear Frieds, > >>> I am facing a problem while finding SNR of a adc output. I am giving > >>> sinusoidal wave at the input amd getting sinusoidal wave with > harmonics > >>> and noise at output. Now, to calculate power for my fundamental > frequency > >>> and noise, I am using DFT. But as I am changing the number of samples, > I > >>> am getting different-different SNR. Here, I am also considering > spectrum > >>> leakage in FFT. So, as my fundamental frequence is 122070312.5Hz > >>> (122.07MHz) and I am sampling it with 25GHz, I am taking number of > samples > >>> 1024 and 2048. I am getting 5db SNR with 1024 samples and 14db SNR > with > >>> 2048 samples. Please help me find where am I going wrong ? > >>> Thanks > > >>25 GHz ??????? > > >Not infeasable on a newer oscilloscope. > > >Regards, > >Allan > > Its not Oscilloscope setup. I am doing matlab simulation.- Hide quoted text - > > - Show quoted text -How are you calculating SNR? Dirk
Reply by ●February 13, 20082008-02-13
>On Feb 12, 7:10=A0am, "sunil_ic" <sunil...@gmail.com> wrote: >> >On Fri, 8 Feb 2008 03:50:32 -0800 (PST), John <sampson...@gmail.com> >> >wrote: >> >> >>On Feb 8, 6:17 am, "sunil_ic" <sunil...@gmail.com> wrote: >> >>> Dear Frieds, >> >>> I am facing a problem while finding SNR of a adc output. I amgiving>> >>> sinusoidal wave at the input amd getting sinusoidal wave with >> harmonics >> >>> and noise at output. Now, to calculate power for my fundamental >> frequency >> >>> and noise, I am using DFT. But as I am changing the number ofsamples,=> >> I >> >>> am getting different-different SNR. Here, I am also considering >> spectrum >> >>> leakage in FFT. So, as my fundamental frequence is 122070312.5Hz >> >>> (122.07MHz) and I am sampling it with 25GHz, I am taking number of >> samples >> >>> 1024 and 2048. I am getting 5db SNR with 1024 samples and 14db SNR >> with >> >>> 2048 samples. Please help me find where am I going wrong ? >> >>> Thanks >> >> >>25 GHz ??????? >> >> >Not infeasable on a newer oscilloscope. >> >> >Regards, >> >Allan >> >> Its not Oscilloscope setup. I am doing matlab simulation.- Hide quotedtex=>t - >> >> - Show quoted text - > >How are you calculating SNR? > >Dirk >Dear Dirk, At 25GHz sample rate, I am doing 1024 point FFT at output from the ADC. So my input signal (122070312.5Hz) will be at f5, 5th sample, in frequency domain (FFT of the output signal). So I calculated singal power by taking square of the 5th sample. And rest all samples in the FFT are the noise+harmonics. So I calculated noise power by integrating all samples' square. Thanks
Reply by ●February 13, 20082008-02-13
On 13 Feb, 05:05, "sunil_ic" <sunil...@gmail.com> wrote:> >On Feb 12, 7:10=A0am, "sunil_ic" <sunil...@gmail.com> wrote:> >How are you calculating SNR? > > >Dirk > > Dear Dirk, > At 25GHz sample rate, I am doing 1024 point FFT at output from the ADC. So > my input signal (122070312.5Hz) will be at f5, 5th sample, in frequency > domain (FFT of the output signal). So I calculated singal power by taking > square of the 5th sample. And rest all samples in the FFT are the > noise+harmonics.Seems reasonable.> So I calculated noise power by integrating all samples' > square.Did you make sure to leave ouut the harmonic where the signal is located? Rune
Reply by ●February 15, 20082008-02-15
>On 13 Feb, 05:05, "sunil_ic" <sunil...@gmail.com> wrote: >> >On Feb 12, 7:10=A0am, "sunil_ic" <sunil...@gmail.com> wrote: > >> >How are you calculating SNR? >> >> >Dirk >> >> Dear Dirk, >> At 25GHz sample rate, I am doing 1024 point FFT at output from the ADC.So>> my input signal (122070312.5Hz) will be at f5, 5th sample, infrequency>> domain (FFT of the output signal). So I calculated singal power bytaking>> square of the 5th sample. And rest all samples in the FFT are the >> noise+harmonics. > >Seems reasonable. > >> So I calculated noise power by integrating all samples' >> square. > >Did you make sure to leave ouut the harmonic where the signal >is located? > >Rune >I am considering harmonics also as Noise. Is it true assumption ?
Reply by ●February 15, 20082008-02-15
On 15 Feb, 09:10, "sunil_ic" <sunil...@gmail.com> wrote:> I am considering harmonics also as Noise. Is it true assumption ?Ah, I didn't read your original post carefully enough - you actually measure the SNR of an ADC, right? There are procedures for testing those sorts of things. Check your ADC's data sheet (or with the manufacturer, they might have a procedure), or read chapter 13.9.1 in Lyons: Understanding DSP, 2nd ed., 2004. Rune
Reply by ●February 15, 20082008-02-15
sunil_ic wrote:>> On 13 Feb, 05:05, "sunil_ic" <sunil...@gmail.com> wrote: >>>> On Feb 12, 7:10=A0am, "sunil_ic" <sunil...@gmail.com> wrote: >>>> How are you calculating SNR? >>>> Dirk >>> Dear Dirk, >>> At 25GHz sample rate, I am doing 1024 point FFT at output from the ADC. > So >>> my input signal (122070312.5Hz) will be at f5, 5th sample, in > frequency >>> domain (FFT of the output signal). So I calculated singal power by > taking >>> square of the 5th sample. And rest all samples in the FFT are the >>> noise+harmonics. >> Seems reasonable. >> >>> So I calculated noise power by integrating all samples' >>> square. >> Did you make sure to leave ouut the harmonic where the signal >> is located? >> >> Rune >> > I am considering harmonics also as Noise. Is it true assumption ?Only in the sense that any unwanted signal could be loosely called noise. In the same way, a tomato plant in a flower garden could be called a weed. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
