My receiver comprises the following parts: 1) a 10-bit ADC running at fs (M samples/symbol), 2) a time-domain "interpolation filter" based on a comb-integrator stage with leaky integrator (sampling rate increase by N) does "interpolation" and acts as a dc-blocking filter as well, 3) a Manchester decoder based on a oversampling & counting algorithm decoding the interpolated signal and running at Nfs. After "interpolation" the Manchester decoder is presented with NM samples/symbol which in principle may be seen as increasing the time resolution for the decoder, but, does time-domain interpolation really help Manchester decoding? Any comment is welcome. Jos�
Does time-domain interpolation help Manchester decoding?
Started by ●July 21, 2009
Reply by ●July 21, 20092009-07-21
Jos� <jjr.navarro@gmail.com> wrote:>My receiver comprises the following parts: > >1) a 10-bit ADC running at fs (M samples/symbol), >2) a time-domain "interpolation filter" based on a comb-integrator >stage with leaky integrator (sampling rate increase by N) does >"interpolation" and acts as a dc-blocking filter as well, >3) a Manchester decoder based on a oversampling & counting algorithm >decoding the interpolated signal and running at Nfs. > >After "interpolation" the Manchester decoder is presented with NM >samples/symbol which in principle may be seen as increasing the time >resolution for the decoder, but, does time-domain interpolation really >help Manchester decoding?It's not "helping" in an information-theoretic sense -- your interpolated signal does not contain more information than the original signal -- but it may help in an algorithmic sense, i.e. it's one way of getting at the zero-crossing / timing information you are looking for in your detector. Steve
Reply by ●July 21, 20092009-07-21
Jos� wrote:> My receiver comprises the following parts: > > 1) a 10-bit ADC running at fs (M samples/symbol), > 2) a time-domain "interpolation filter" based on a comb-integrator > stage with leaky integrator (sampling rate increase by N) does > "interpolation" and acts as a dc-blocking filter as well, > 3) a Manchester decoder based on a oversampling & counting algorithm > decoding the interpolated signal and running at Nfs. > > After "interpolation" the Manchester decoder is presented with NM > samples/symbol which in principle may be seen as increasing the time > resolution for the decoder, but, does time-domain interpolation really > help Manchester decoding?Some people think that the curved toothbrush is really useful, since it helps cleaning the teeth from behind a corner. It is certainly useful, but why do you have to invent such a muddy algorithm when a simple phase locked loop will do the job? VLV
Reply by ●July 22, 20092009-07-22
On Jul 21, 6:51�pm, Vladimir Vassilevsky <nos...@nowhere.com> wrote:> Some people think that the curved toothbrush is really useful, since it > helps cleaning the teeth from behind a corner. It is certainly useful, > but why do you have to invent such a muddy algorithm when a simple phase > locked loop will do the job? > > VLVHi, A simple phase locked loop consumes more energy than a simple interpolation filter. Furthermore, if your logic is implemented in a FPGA then you don't usually find built-in PLLs; if your logic is implemented in a ASIC then a PLL does take more effort and risk than an all-digital solution. I forgot to mention that the output of my interpolation filter is sliced to one bit before being processed by my Manchester decoder (MD). Also, my MD counts ones (Hi points) / zeros (Lo points) and decides whether it was a long one/zero or a short one. Based on this information decoding follows straightforward. For more info, see my Blog http://asicdigitalarithmetic.wordpress.com/2009/04/23/an-all-digital-manchester-decoder/ I really need to block DC after AD conversion since the DC component really worsens decoding. The leaky integrator is simple and combined with a comb stage does allow easily for interpolation. The robustness of the decoding algorithm depends strongly on the time resolution and therefore interpolation may help decoding. But, as Steve mentions it helps only in a "algorithmic" sense and not in a "information-theoretic" sense. Anyway, I would like to hear more impressions. Thank you
Reply by ●July 22, 20092009-07-22
Jos� wrote:> On Jul 21, 6:51 pm, Vladimir Vassilevsky <nos...@nowhere.com> wrote: > >>Some people think that the curved toothbrush is really useful, since it >>helps cleaning the teeth from behind a corner. It is certainly useful, >>but why do you have to invent such a muddy algorithm when a simple phase >>locked loop will do the job? >> >>VLV > > > Hi, > > A simple phase locked loop consumes more energy than a simple > interpolation filter.Eh? Eh? All you have to do is lock the sampling rate to the signal.> Furthermore, if your logic is implemented in a FPGA then you don't > usually find built-in PLLs; if your logic is implemented in a ASIC > then a PLL does take more effort and risk than an all-digital solution.WOW! When I was young, I used to decode manchester using a couple of pieces of 74xx logic, and it didn't really take much more then a small breadboard. Worked just fine.> I forgot to mention that the output of my interpolation filter is > sliced to one bit before being processed by my Manchester decoder > (MD). > Also, my MD counts ones (Hi points) / zeros (Lo points) and decides > whether it was a long one/zero or a short one.Incredible. What a monstrous solution for a simple problem. Based on this> information decoding follows straightforward. For more info, see my > Blog > > http://asicdigitalarithmetic.wordpress.com/2009/04/23/an-all-digital-manchester-decoder/ > > I really need to block DC after AD conversion since the DC component > really worsens decoding. The leaky integrator is simple and combined > with a comb stage > does allow easily for interpolation. The robustness of the decoding > algorithm depends strongly on the time resolution and therefore > interpolation may help decoding.:))))))) Keep it simple and elegant. Avoid the smart scientific words that you don't understand.> But, as Steve mentions it helps only in a "algorithmic" sense and not > in a "information-theoretic" sense. Anyway, I would like to hear more > impressions.BTW, in the SED newsgroup, they collect the examples of bad designs. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
