Digital Signal Processing

good response from Hitesh Shukla . triangular wave = sum of infinity sinusoidal signals

I think the main issue is lack of consumer demand specifically for a "do anything" device. People prefer their gadgets to have the long battery life which which specialised hardware can deliver. However bedside clock radios added functions for DAB+ WiFi streaming media years ago. Hardware is now 'manufactured' as a (soft) IP design, e.g. Imagination: IP CPU design -> Frontier Silicon: radio software and integration as System-on-Chip -> Cisco's Pure Digital: end-user devices. From their comments in the press Frontier Silicon uses SDR to give it the flexibility to specialise products to meet an increasingly fragmented market.

Yes, ok but orders of magnitude of throughput increase will only come from massive extensions of mimo technology, or massive mimo. So that's one to watch.

your question can be subdivided into two possibilities: analog or digital image (in the visible range) down converted to the IR range. Analog method depends very much on the material properties, and down conversion of the ranges may not be uniform for the entire visible range. but digital images has no problem. Using FPGA (http://www.mpi-inf.mpg.de/~strzodka/papers/public/KlErHu_02fpga.pdf) , or specialized media processor eg, http://www.ti.com/lit/ds/symlink/tms320dm6437.pdf from TI, it is possible to digitally process the image and down convert the frequencies to any range you like. Processing in real time may be technically challenging (even with the current GPU state of art, but I may be wrong :-).

simply put together convolution gives average overlapping area between two signals,which is nothing but low pass filtering. when we convolve 2 signals, one is considered to be impulse response of LTI system and other as input signal. Hence is equivalent to passing a signal through lowpass filter. whereas, correlation is simply comparison of the signal with itself(autocorrelation) or with other siganl (cross correlation) with respect to time.

I suggest you the SimPowerSystems MatLab module. this module has a controlled voltage source. You can generate the desired signal using this source.

Hi, The widespread method is to use Chen and Millero's equation (UNESCO). It calculates the sound speed from CTD data. The accuracy is 0.25 m/s. Or you can design a device using the time of flight between two hydrophones. Regards, Nicolas

Thank you so much, this helped a lot Ana and Chris :)

In some sense, FFT can be seen as a linear transformation between two infinite dimensional functional spaces generated by orthogonal functions (basis vectors): the "time domain", generated by an infinite set of impulse functions at (infinite) different delays; and the frequency domain generated by infinite complex exponential functions at (infinite) different frequencies. The powerful idea is that the second basis set is much more efficient for signals originated by anything that rotates (and maybe later nonlinearly transformed). This includes electrical signals (generators rotate), vibrations (most vibrations in industry are generated by rotating elements), etc. Moreover, smooth nonlinear transformations of sines usually produce a few more sines of other frequencies. Also, complex exponentials are eigenfunctions of most usual differential operators (in simpler words, they are the solutions of linear difference equations); this adds more signals that are composed by only a few complex exponentials (decaying sinusoids). Such sine-biased composition of signals makes that "coordinates" referred to the exp(jw)-basis are only nonzero for a few basis functions. In other words, in the frequency domain we get sparse (hence simpler and more compact) representations of many real life problems. If we add to the point the fast computation that the FFT provides to move between domains we get an outstanding contribution of Tukey to science and technology.

Hi Mahmood, EVM just gives you the average difference of received points compared to the ideal point of a constellation diagram (E.g., provides information about first order spreading). If your distribution of points can be described closely by a Gaussian distribution, EVM could be calculated from mean and std quite accurately, and would have a close relation to the actual BER. If, however, you have a case where only a few points are far away from the ideal point, while all the others are close to it (as you described in your case), EVM does not relate well to BER. Similar discrepancies could occur for other statistical distributions such as truncated Gaussian, or bean-shaped due to NL-phase noise. So I think, BER is the only true performance measure - EVM could serve as estimate for certain cases, but would fail for others.

You cannot reconstruct the spectral values outside the (-Fs/2, Fs/2) bandwidth, because you cannot accurately quantify any values outside this bandwidth with the information that you have. I further assume that you used an anti-aliasing filter in your setup. If so, responses are further filtered.

Thanks Guys you all have been a great help I will approach this issue at the computational level, in parallel. Design the node to calculate in the field to verify optimal signal processing.

For SaO2, normally one LED is red, with wavelength of 660 nm, and the other is infrared, 940 nm. To measure bilirubin a blue LED, 460 nm or to measure carboxyhemoglobin, a 500 or 550 nm LED might be added. Looking at: http://www.andor.com/learning-academy/ccd-spectral-response-(qe)-defining-the-qe-of-a-ccd The infrared measurement might be a problem - although at 50% sensitivity this still might be OK. Since there are isosbestic points at 586 nm and near 808 nm one might do with a 586 nm LED. Accurate control over wavelength is critical though. For normal range heart rates, I'd suggest you could make a pretty good measurement, at 80-120 fps, up-sampling to accurately make peak measurements. You may even have an advantage, since rather than strobing, and interleaving LEDs at each frequency, you might run the LEDs simultaneously and identify frequency by placement of the LED in the CCD frame. Frame sampling jitter between frequencies would then be non-critical. Resolution could be enhanced by estimating intensity distribution across multiple pixels.

Check this: A Short Biography of Joseph Fourier and Historical Development of Fourier Series and Fourier Transforms http://dx.doi.org/10.1080/0020739X.2011.633712 For the history of FFT, check this other: Historical Notes on the Fast Fourier Transform http://www.signallake.com/innovation/FFTHistoryJun67.pdf

Thanks

Manoj, The sampling theorem says that you may reconstitute entirely a signal if you upper sample the signal and if the signal is sufficiently periodic (not available for distribution). Entropie is an informations amount evaluation the signal carries. For example, if you under sample the signal then you lost information and the signal entropy decrease. h(i)=pi.log(1/pi) decrease because : pi statistically decrease or increase away from the maximum of h(i) / pi curve. But you can continue to use your signal. It is like you have noise on the signal. Sampling theorem is just a limit as : ------------------------------------------------- Low Bit rate but low efficienty <= Sampling Theorem = High efficienty rate <= High bit rate but low efficienty This type of theorem just show us how to raise efficienty up to n = 1. If you under sample then n<<1. If you upper sample then n<<1. If you are close to the theorem then n=0.99 for example. Best regards. Jean-Paul Cipria http://www.nanotechinnov.com/

STFT is basically a DFT (or FFT) with sliding window. So the time-frequency resolution tradeoff is that of the DFT, that is df=1/T, with df the frequency resolution and T the transform time interval or window. The DFT is a bank of filters with frequency responses centered each on its own frequency bin (point of the frequency axis), overlapping in between and spaced by df. This said, when you move the DFT windows, you can overlap them by the amount you want, and this time interval between the center points of two successive DFTs is the time step of the STFT. at time instant tn we have DFTn centered on tn and time width T, moving to t(n+1) = tn + tstep, the new DFT(n+1) is centered on t(n+1) and has still a width of T seconds. tstep is established by the amount of overlapping you chose, depending on e.g. how do you track non-stationary phenomena: tstep = (1-p) T, where p is the overlap fraction, from 0 (no overlap) to 1 (100% overlap, so you do not move on the time axis). The tstep is an artificial time resolution because the values of the various DFTs are correlated. This is more complicated and it is also related to teh smoothing windows you use to control frequency leakage (e.g. hanning, Blackman-Harris, etc.). These things are explained in many books of Signal Processing, such as Lyons, Kay-Marple, Proakis-Manolakis, and also in Matlab (Signal Processing Toolbox)

Thanks 'Oliver Tremois' , really it is very inportant to understand, but i need a description with VHDL program. Thanks for comprehension

Buffer

It appears that in the equations 6.62 and 6.63, in the fraction outside the brackets variable m should be replaced by the variable q.

Great. Thanks

Because it is the causal nature of the real world physical meaning time signals....simple as that. That said you can model filters for non-causal phenomena but fundamentally that would be the strongest of reasons.

There would be rather than one algorithm various types of approach depending on the philosophy behind conception of your model. Fundamentally: Deterministic and stochastic-statistical approaches. First one is based in any case on "source footprint" knowledge and therefore you look for the closest match for your model in the signal and the second one relies in the forever beloved, simple yet mighty correlation -can you perhaps guess my taste from my comments? haha. In earth sciences for us is really impossible to have source prediction in the strictest of senses so we estimate source by correlation and extrapolation algorithms. Yet in NDT techniques quite often you model the source.

Good references are: E. Lutz, M. Werner, A. Jahn, "Satellite systems for personal and broadband communications", Springer-Verlag, Berlin, 2000. G. Maral and M. Bousquet, "Satellite Communications Systems: Systems, Techniques and Technology, 5th Edition", Wiley Both are books which are not freely available. In case you do not have access to these books you can check the documentation of the propagation program at JPL, NASA which is available online: http://descanso.jpl.nasa.gov/Propagation/propagation.cfm