Question regarding DSP Theory / Computer Science / Forums

Forums

4hv.org :: Forums :: Computer Science

« Previous topic | Next topic »

Question regarding DSP Theory

previous 1 2

Move Thread

LAN_403

Andy

Wed Sept 18 2013, 07:11PM

Registered Member #4266 Joined: Fri Dec 16 2011, 03:15AM
Location:
Posts: 874

Hi Linas

You could try this code

for(i=0;i<0xff;i++) {
for(j=0;j<0x16;j++) {
temp = v1|v2
temp1 = v1&temp
temp5 = v2&temp
temp6 = temp5|temp1;
temp2 = i^temp6
temp4=temp2+temp4;
}
temp3 = 0xfff-temp4
count = temp3+count
printf(%4X,temp3)

}

printf("% 4X,count) , this value you use

It should find the differential of the noise and give a lineal response works on sha1 the avalanche factor was 1.64, the code might be abit wrong, can't remember how I did it last time,but say you have 0x41 = 01000001 that gets changed into 10111110, and then xored .If you reads 16 points then do that with each character, which then cycles throughout 0xff range, if you print out the character in C you should see a slope or hill.

Steve Conner

Wed Sept 18 2013, 08:52PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I tried running this code on my Raspberry Pi and it turned into a strawberry tart. :(

Andy

Wed Sept 18 2013, 09:37PM

Registered Member #4266 Joined: Fri Dec 16 2011, 03:15AM
Location:
Posts: 874

I tried running this code on my Raspberry Pi and it turned into a strawberry tart. :(

What???

WaveRider

Thu Sept 19 2013, 09:08AM

Registered Member #29 Joined: Fri Feb 03 2006, 09:00AM
Location: Hasselt, Belgium
Posts: 500

Linas,
Maybe posting a block diagram of your system would help us understand what exactly you are trying to do. I see your problem as an acquisition/tracking problem that should be straightforward to solve, but unfortunately your description is somewhat confusing (to me at least).

Steve Conner

Thu Sept 19 2013, 10:09AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Andy wrote ...

I tried running this code on my Raspberry Pi and it turned into a strawberry tart. :(

What???

Even if this code does calculate the avalanche factor of sha1, I can't see how that is the slightest bit relevant to Linas' problem. :(

Andy

Thu Sept 19 2013, 07:16PM

Registered Member #4266 Joined: Fri Dec 16 2011, 03:15AM
Location:
Posts: 874

Even if this code does calculate the avalanche factor of sha1, I can't see how that is the slightest bit relevant to Linas' problem. :(

Question, can i do something to make oscillator act as linear system, by doing deconvolution or other DSP magic, so even if oscillator has nonlinear response, i can correct my feedback voltage by knowing all phase correction values in past so it will act as linear ?

Oscillator is Laser. It has random noise

Running the code over a sine wave, would give a linear step, I just saw nonlinear to linear, about what it did, sha is meant to be random, and by design nonlinear between each proceeding character, if you imputed AAAA*4,/BBBB*4/CCCC*4 the output feed into the code, would give a linear number from a to b to c

Linas

Thu Oct 17 2013, 08:55AM

Registered Member #1143 Joined: Sun Nov 25 2007, 04:55PM
Location: Vilnius, Lithuania
Posts: 721

Is what i figure out.
I need to generate feedback, with well defined spectral components ( like one period of cosinus, simple counting up, counting up and down, window (aka delta with width), and so on.) .At the same time measure response. When i get response, i will make FFT of input signal (In(x), and output signal (H(x)), .
and because H(x)=In(x)*Foscillator(x) is in frequency domain, i simply get Foscillator =H(X)/In(x)

But because i have noise, and at the same time i have active oscillator stabilization circuit, what i would get is combination of two functions.

But if i run this algorithm only for my oscillator with no stabilization, it is possible that while i am changing phase with feedback, phase can go off my measuring window, since phase can goo anywhere from +-128pi, and FFT method can only calculate phase in +-pi limit. Of course oscillator phase will run out of my measuring windows quite slow ( i can get 2-20 measurements while phase is still in my measuring window, but it is a bit unpredictable

BigBad

Thu Oct 17 2013, 03:17PM

Registered Member #2529 Joined: Thu Dec 10 2009, 02:43AM
Location:
Posts: 600

Maybe you should ditch the FFT stuff and do optimal filtering to do peak edge finding and then use that to calculate the phase and use that to feedback. That way you won't necessarily have to wait for a whole cycle and you won't have the windowing issues that FFT give you (or at least not the same).

The trick with optimal filtering is that the parts of the signal where it changes most quickly need greater weighting.

Also as Steve says, characterising or understanding the nonlinearity in the system seems to be a top priority, unless you know how much pull you get for how much voltage input, you're never going to get it stable.

Linas

Thu Oct 17 2013, 04:40PM

Registered Member #1143 Joined: Sun Nov 25 2007, 04:55PM
Location: Vilnius, Lithuania
Posts: 721

well, not FFT, but to be more precise DFT fo single poiny. My new dsp processor can do 1.8MHz loop, while i only need 250KHz (without I/O, it will be slower)

Also i am writing code for DFT inside acquisition fpga, i bet i can make it work by using Signed integer table of coefficients, and only thing that itself is do MAC with 16b of data from detector (detector eads spectrum of laser, and phase is inbeded in that sygnal)

previous 1 2

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob