Re: [casper] pfb/fft confusion
Great! That's what it seemed like it was doing, but I couldn't quite convince myself. Thanks for the help. Date: Wed, 28 May 2014 15:16:06 -0400 Subject: Re: [casper] pfb/fft confusion From: glenn.calt...@gmail.com To: jay_br...@live.com CC: casper@lists.berkeley.edu Yes, i.e. if you set it to 2^2 then it means the PFB processes x0,x1,x2,x3 on the first FPGA clock, then x4,x5,x6,x7 on the second FPGA clock, etc... On Wed, May 28, 2014 at 3:15 PM, Jay Brady wrote: So does simultaneous inputs refer to simultaneous samples (within one fpga clock) of a single physical input channel? Date: Wed, 28 May 2014 14:43:24 -0400 Subject: Re: [casper] pfb/fft confusion From: glenn.calt...@gmail.com To: jay_br...@live.com CC: casper@lists.berkeley.edu You want The PFB to have 2^0 simultaneous inputs, then make one PFB for each stream. On Wed, May 28, 2014 at 2:41 PM, Jay Brady wrote: Hello all, With all of the different pfb/fft blocks, I've gotten a bit confused. Some of the terminology used on the wiki is a bit unclear to me (simultaneous streams vs simultaneous inputs vs parallel time samples) which has led to this confusion Anyway, I'm trying to build a spectrometer (a la tutorial 3) for the adc16x250, with the adc running 16 channel at 200 Msamp/s (so no demuxing). This gives me 16 inputs that are fed into a pfb/fft block combo. I'm concerned that I am somehow getting those 16 inputs mixed up in the pfb or fft. Here are the blocks and parameters I'm using right now:pfb_fir_real with size: 2^10, Taps: 4, Simultaneous Inputs: 2^2, and the Make Biplex box is not checked.fft_biplex_real_4x with size: 2^10 and Simultaneous Inputs 4^1. This gives me 4 inputs/outputs on each of the blocks which are just wired straight through. With this, the fft outputs seem ok-ish. I get a peak at the right frequency channel, but the sidelobe behavior is pretty terrible. I get high sidelobes that sometimes extend out 20-30 bins on either side of the main peak. Any advice you can give would be great. Jay Brady
Re: [casper] pfb/fft confusion
Yes, i.e. if you set it to 2^2 then it means the PFB processes x0,x1,x2,x3 on the first FPGA clock, then x4,x5,x6,x7 on the second FPGA clock, etc... On Wed, May 28, 2014 at 3:15 PM, Jay Brady wrote: > > So does simultaneous inputs refer to simultaneous samples (within one fpga > clock) of a single physical input channel? > > -- > Date: Wed, 28 May 2014 14:43:24 -0400 > Subject: Re: [casper] pfb/fft confusion > From: glenn.calt...@gmail.com > To: jay_br...@live.com > CC: casper@lists.berkeley.edu > > > You want The PFB to have 2^0 simultaneous inputs, then make one PFB for > each stream. > > > > On Wed, May 28, 2014 at 2:41 PM, Jay Brady wrote: > > Hello all, > > With all of the different pfb/fft blocks, I've gotten a bit confused. Some > of the terminology used on the wiki is a bit unclear to me (simultaneous > streams vs simultaneous inputs vs parallel time samples) which has led to > this confusion > > Anyway, I'm trying to build a spectrometer (a la tutorial 3) for the > adc16x250, with the adc running 16 channel at 200 Msamp/s (so no demuxing). > This gives me 16 inputs that are fed into a pfb/fft block combo. I'm > concerned that I am somehow getting those 16 inputs mixed up in the pfb or > fft. > > Here are the blocks and parameters I'm using right now: > pfb_fir_real with size: 2^10, Taps: 4, Simultaneous Inputs: 2^2, and the > Make Biplex box is not checked. > fft_biplex_real_4x with size: 2^10 and Simultaneous Inputs 4^1. > This gives me 4 inputs/outputs on each of the blocks which are just wired > straight through. > > > With this, the fft outputs seem ok-ish. I get a peak at the right > frequency channel, but the sidelobe behavior is pretty terrible. I get high > sidelobes that sometimes extend out 20-30 bins on either side of the main > peak. > > Any advice you can give would be great. > > Jay Brady > > >
Re: [casper] pfb/fft confusion
So does simultaneous inputs refer to simultaneous samples (within one fpga clock) of a single physical input channel? Date: Wed, 28 May 2014 14:43:24 -0400 Subject: Re: [casper] pfb/fft confusion From: glenn.calt...@gmail.com To: jay_br...@live.com CC: casper@lists.berkeley.edu You want The PFB to have 2^0 simultaneous inputs, then make one PFB for each stream. On Wed, May 28, 2014 at 2:41 PM, Jay Brady wrote: Hello all, With all of the different pfb/fft blocks, I've gotten a bit confused. Some of the terminology used on the wiki is a bit unclear to me (simultaneous streams vs simultaneous inputs vs parallel time samples) which has led to this confusion Anyway, I'm trying to build a spectrometer (a la tutorial 3) for the adc16x250, with the adc running 16 channel at 200 Msamp/s (so no demuxing). This gives me 16 inputs that are fed into a pfb/fft block combo. I'm concerned that I am somehow getting those 16 inputs mixed up in the pfb or fft. Here are the blocks and parameters I'm using right now:pfb_fir_real with size: 2^10, Taps: 4, Simultaneous Inputs: 2^2, and the Make Biplex box is not checked.fft_biplex_real_4x with size: 2^10 and Simultaneous Inputs 4^1. This gives me 4 inputs/outputs on each of the blocks which are just wired straight through. With this, the fft outputs seem ok-ish. I get a peak at the right frequency channel, but the sidelobe behavior is pretty terrible. I get high sidelobes that sometimes extend out 20-30 bins on either side of the main peak. Any advice you can give would be great. Jay Brady
Re: [casper] pfb/fft confusion
You want The PFB to have 2^0 simultaneous inputs, then make one PFB for each stream. On Wed, May 28, 2014 at 2:41 PM, Jay Brady wrote: > Hello all, > > With all of the different pfb/fft blocks, I've gotten a bit confused. Some > of the terminology used on the wiki is a bit unclear to me (simultaneous > streams vs simultaneous inputs vs parallel time samples) which has led to > this confusion > > Anyway, I'm trying to build a spectrometer (a la tutorial 3) for the > adc16x250, with the adc running 16 channel at 200 Msamp/s (so no demuxing). > This gives me 16 inputs that are fed into a pfb/fft block combo. I'm > concerned that I am somehow getting those 16 inputs mixed up in the pfb or > fft. > > Here are the blocks and parameters I'm using right now: > pfb_fir_real with size: 2^10, Taps: 4, Simultaneous Inputs: 2^2, and the > Make Biplex box is not checked. > fft_biplex_real_4x with size: 2^10 and Simultaneous Inputs 4^1. > This gives me 4 inputs/outputs on each of the blocks which are just wired > straight through. > > > With this, the fft outputs seem ok-ish. I get a peak at the right > frequency channel, but the sidelobe behavior is pretty terrible. I get high > sidelobes that sometimes extend out 20-30 bins on either side of the main > peak. > > Any advice you can give would be great. > > Jay Brady >
[casper] pfb/fft confusion
Hello all, With all of the different pfb/fft blocks, I've gotten a bit confused. Some of the terminology used on the wiki is a bit unclear to me (simultaneous streams vs simultaneous inputs vs parallel time samples) which has led to this confusion Anyway, I'm trying to build a spectrometer (a la tutorial 3) for the adc16x250, with the adc running 16 channel at 200 Msamp/s (so no demuxing). This gives me 16 inputs that are fed into a pfb/fft block combo. I'm concerned that I am somehow getting those 16 inputs mixed up in the pfb or fft. Here are the blocks and parameters I'm using right now:pfb_fir_real with size: 2^10, Taps: 4, Simultaneous Inputs: 2^2, and the Make Biplex box is not checked.fft_biplex_real_4x with size: 2^10 and Simultaneous Inputs 4^1.This gives me 4 inputs/outputs on each of the blocks which are just wired straight through. With this, the fft outputs seem ok-ish. I get a peak at the right frequency channel, but the sidelobe behavior is pretty terrible. I get high sidelobes that sometimes extend out 20-30 bins on either side of the main peak. Any advice you can give would be great. Jay Brady
