Hello I am sweeping a signal from 10Hz to 1MHz (in 2.5% increments) and monitoring the output signal (reference signal) as well as the signal returned from a DUT (device under test) that may down by up to 90dB from the original output signal. I am using my own dual 14bit with some logic to let me alter the sampling frequency form 2KHz up to 80MHz (2 x 4K FIFO). I need to process the two signals as fast as possible. My question : what would be the best method for processing these two signals (reference would always be good and quiet constant - signal from DUT changes continually) I need good dynamic range, repeatability in the measurement technique. (1) FFT :- would need to continually alter the sampling frequency and ....... ????? (2) FIR :- need to capture lots of samples, and change coefficents ...........???? (3) Both OR ..............?????? Thanks in advance.
General DSP processing question
Started by ●January 24, 2011
Reply by ●January 24, 20112011-01-24
> My question : what would be the best method for processing these two > signalsDFT ? http://www.embeddedFORTH.de/temp/d.pdf MfG JRD
Reply by ●January 24, 20112011-01-24
On Mon, 24 Jan 2011 07:20:42 -0600, "charlyb" <carl@n_o_s_p_a_m.eca.co.za> wrote:>Hello > >I am sweeping a signal from 10Hz to 1MHz (in 2.5% increments) and >monitoring the output signal (reference signal) as well as the signal >returned from a DUT (device under test) that may down by up to 90dB from >the original output signal. > >I am using my own dual 14bit with some logic to let me alter the sampling >frequency form 2KHz up to 80MHz (2 x 4K FIFO). I need to process the two >signals as fast as possible. > >My question : what would be the best method for processing these two >signals (reference would always be good and quiet constant - signal from >DUT changes continually) >I need good dynamic range, repeatability in the measurement technique. > >(1) FFT :- would need to continually alter the sampling frequency and >....... ????? > >(2) FIR :- need to capture lots of samples, and change coefficents >...........???? > >(3) Both OR ..............?????? > > >Thanks in advance.What sort of information are you trying to extract with the processing? Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com
Reply by ●January 24, 20112011-01-24
On 01/24/2011 08:09 AM, Eric Jacobsen wrote:> On Mon, 24 Jan 2011 07:20:42 -0600, "charlyb" > <carl@n_o_s_p_a_m.eca.co.za> wrote: > >> Hello >> >> I am sweeping a signal from 10Hz to 1MHz (in 2.5% increments) and >> monitoring the output signal (reference signal) as well as the signal >> returned from a DUT (device under test) that may down by up to 90dB from >> the original output signal. >> >> I am using my own dual 14bit with some logic to let me alter the sampling >> frequency form 2KHz up to 80MHz (2 x 4K FIFO). I need to process the two >> signals as fast as possible. >> >> My question : what would be the best method for processing these two >> signals (reference would always be good and quiet constant - signal from >> DUT changes continually) >> I need good dynamic range, repeatability in the measurement technique. >> >> (1) FFT :- would need to continually alter the sampling frequency and >> ....... ????? >> >> (2) FIR :- need to capture lots of samples, and change coefficents >> ...........???? >> >> (3) Both OR ..............?????? >> >> >> Thanks in advance. > > What sort of information are you trying to extract with the > processing?If you're looking for just the fundamental response (i.e. just the signal that comes through at the same frequency) then Rafael's suggestion is spot-on. But is that all you need? -- 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 ●January 24, 20112011-01-24
Hi Eric Thanks for responding. I am looking for magnitude difference between the two signals and phase Charlyb>What sort of information are you trying to extract with the >processing? > > >Eric Jacobsen >Minister of Algorithms >Abineau Communications >http://www.abineau.com >
Reply by ●January 24, 20112011-01-24
On 01/24/2011 11:40 AM, charlyb wrote:> Hi Eric > > Thanks for responding. > > I am looking for magnitude difference between the two signals and phaseThen the single-point DFT that Rafael suggested is your best bet. It's easy to understand, doesn't require much storage, and it's accurate. What more could you want? -- 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 ●January 24, 20112011-01-24
charlyb wrote:> Hello > > I am sweeping a signal from 10Hz to 1MHz (in 2.5% increments) and > monitoring the output signal (reference signal) as well as the signal > returned from a DUT (device under test) that may down by up to 90dB from > the original output signal. > > I am using my own dual 14bit with some logic to let me alter the sampling > frequency form 2KHz up to 80MHz (2 x 4K FIFO). I need to process the two > signals as fast as possible. > > My question : what would be the best method for processing these two > signals (reference would always be good and quiet constant - signal from > DUT changes continually) > I need good dynamic range, repeatability in the measurement technique. > > (1) FFT :- would need to continually alter the sampling frequency and > ....... ????? > > (2) FIR :- need to capture lots of samples, and change coefficents > ...........???? > > (3) Both OR ..............??????Using frequency sweeps is no good idea. Use discrete multitone test signals instead. The tones should be orthogonal over the measurement interval; various tricks could be applied. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●January 24, 20112011-01-24
Tim Wescott wrote:> On 01/24/2011 11:40 AM, charlyb wrote: > >> Hi Eric >> >> Thanks for responding. >> >> I am looking for magnitude difference between the two signals and phase > > > Then the single-point DFT that Rafael suggested is your best bet. It's > easy to understand, doesn't require much storage, and it's accurate.Nope. It requires the entire swept frequency FFT to deal with the time/phase effects, and a hell of the post processing (as the sweep is nonlinear) to approach the goal of 90dB accuracy. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●January 24, 20112011-01-24
On 01/24/2011 01:11 PM, Vladimir Vassilevsky wrote:> > > Tim Wescott wrote: > >> On 01/24/2011 11:40 AM, charlyb wrote: >> >>> Hi Eric >>> >>> Thanks for responding. >>> >>> I am looking for magnitude difference between the two signals and phase >> >> >> Then the single-point DFT that Rafael suggested is your best bet. It's >> easy to understand, doesn't require much storage, and it's accurate. > > Nope. > > It requires the entire swept frequency FFT to deal with the time/phase > effects, and a hell of the post processing (as the sweep is nonlinear) > to approach the goal of 90dB accuracy.That depends on how you take the word "sweep", and how long the measurement can take. If by "sweep" he means incrementing the frequency, then he just needs to camp on each frequency long enough to get the signal to his desired accuracy, then clear his integrators and move on to the next. Even if the sweep is continuous, if he sweeps at a low enough speed things will be find (with all due respect paid to filter ring time vs. sweep speed). -- 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 ●January 24, 20112011-01-24
Tim Wescott wrote:> On 01/24/2011 01:11 PM, Vladimir Vassilevsky wrote: > >> >> >> Tim Wescott wrote: >> >>> On 01/24/2011 11:40 AM, charlyb wrote: >>> >>>> Hi Eric >>>> >>>> Thanks for responding. >>>> >>>> I am looking for magnitude difference between the two signals and phase >>> >>> >>> >>> Then the single-point DFT that Rafael suggested is your best bet. It's >>> easy to understand, doesn't require much storage, and it's accurate. >> >> >> Nope. >> >> It requires the entire swept frequency FFT to deal with the time/phase >> effects, and a hell of the post processing (as the sweep is nonlinear) >> to approach the goal of 90dB accuracy. > > > That depends on how you take the word "sweep", and how long the > measurement can take. > > If by "sweep" he means incrementing the frequency, then he just needs to > camp on each frequency long enough to get the signal to his desired > accuracy, then clear his integrators and move on to the next. Even if > the sweep is continuous, if he sweeps at a low enough speed things will > be find (with all due respect paid to filter ring time vs. sweep speed).That is 17 octaves starting from 10 Hz, to 90dB accuracy. Make the numbers yourself. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
