Hi, in ''What is a Decision Feedback Equalizer (DFE)?'' at ''7:25'' you gave a definition of DEF which differs from this video at ''5:55''. The difference is one coefficient 1/h0 . I already asked another question below the other video. Thanks
All of the channel "taps" (h-values) need to be measured, and their values of course affect the detector performance. Another factor that affects the detector performance is the noise power spectral density. When we write down a mathematical model, we sometimes scale the h values so that h_0 = 1 and also scale the noise power accordingly (keeping the SNR constant).
hello professor lain, I want to ask that what are the effects a channel cause to a signal in terms of amplitude and phase and frequency to a channel? and how does an Equalizer able to correct these effects of channel means basically what functions are performed by Equalizer
Channel amp and phase are affected by things such as multipath, which can also cause ISI, which is what the equaliser is needed to combat. Perhaps watch "What is Intersymbol Interference ISI?" ua-cam.com/video/I087FUvW2ys/v-deo.html and "What is Rayleigh Fading?" ua-cam.com/video/-FOnYBZ7ZfQ/v-deo.html
sir in case 4 u have taken OFDM so the sole purpose of ofdm technology is to eliminate ISI or convert a frequency selective channel to frequency flat channel, than what is the purpose of doing equalization as equilization is used to remove the effect of ISI by opening the eye in eye diagram which has been already done by dividing the channel in to frequency flat sub bands by OFDM systems??
I'm not really sure what you're asking sorry. In the case of OFDM there is no "eye diagram", since all the sub channels are transmitted at the same time. But yes, each sub channel is flat fading (once the FFT in the receiver is applied). The sub channels still need to have their amplitudes adjusted/scaled so that the overall gain is unity, so that the detector can map the received signal onto the constellation points. This overall FFT-with-subcarrier-scaling can be viewed as the "equaliser".
Sure, as long as the channel gain response is a one-to-one function. You need a filter which has an impulse response that is the inverse of the channel's impulse response (or approximates it). This is often not easy to find though, as the problem can be non-convex.
Very clear and detailed explanation. I love your videos. Can you please add the equalization gain of the mmse vs the dfe from a pure mathematic form? Meaning how much db do we gain with mmse vs dfe.
Thanks, I'm glad you like the videos. There's not a simple answer to your question, since the performance of the DFT is hard to quantify mathematically, due to the difficulty in modelling the error sequences that result from an incorrect single error being fed back. It depends on the amount of ISI in the channel and the SNR.
HI, Prof. Thanks for such nice lecture. Could you please explain 1) why do we put tail symbols and 2) why do we take it same as the training symbols: (do we put it in order to protect the next training symbols) or it also has some other purposes.
ISI in the channel spreads out each input symbol over multiple symbol time slots. So, if the detector cutoff its measurements exactly at the time of the last data symbol in a packet, then it would lose the energy from the last few data symbols that arrives after that time. That would make it harder to make decisions on those symbols (compared with the other earlier symbols). So the received vector typically includes a few symbols extra (and correspondingly, dummy "tail symbols" are added to the end of the packet before transmission, so it doesn't matter that they spread out). If you use the same "dummy" symbols for the last few symbols of the header, then the sampled channel matrix will be circular, which has advantages for implementation, but it's not necessary.
@@iain_explains Thanks. Got ur point. I have one more question. How should we choose the length of the data symbols and training+header for good equalization? Does this process of choosing the lengths of header+data+tail depends on the choice of equalization.
The choice of the lengths of the header, data, and tail depend on many factors. They are defined in the relevant standards, so if you are transmitting according to a standard, then you need to use the lengths defined in that standard.
@@iain_explains Dear Professor, As you mentioned in above comment Why would a detector even cutoff its measurements exactly at the time of the last data symbol in a packet?
thank you so much. Much needed content. Please make some video on detection of time offset and carrier freq offset for non-OFDM general modulation cases, like Gardner's method. I always find them difficult to understand from the books
Hi Professor, Enjoy your style of explaining communication concepts :). Wonder if you plan to give a talk on partial response? Look forward to your lecture on the topic.
@@iain_explains i think they meant partial response signalling. which is something like duobinary signalling according to what i've learnt. I haven't learnt this fully, so i'm not exactly sure what it means. From what i've found online i think it means a scheme which allows for ISI between two adjacent terms for some symbol in the sequence. Thanks for the lecture, it's very helpful :)
Hi, in ''What is a Decision Feedback Equalizer (DFE)?'' at ''7:25'' you gave a definition of DEF which differs from this video at ''5:55''. The difference is one coefficient 1/h0 . I already asked another question below the other video. Thanks
All of the channel "taps" (h-values) need to be measured, and their values of course affect the detector performance. Another factor that affects the detector performance is the noise power spectral density. When we write down a mathematical model, we sometimes scale the h values so that h_0 = 1 and also scale the noise power accordingly (keeping the SNR constant).
1:35: h are channel values. training symbols help estimate channel values
Clear message, clear structure, easy to understand, thank you
You're very welcome
Can you make a video explaining the differences between a symbol spaced equalizer and a Fractionally Spaced Equaliser? Thank you.aced equalizer
Thanks for the suggestion. I've put it on my "to do" list.
What a fantastic explanation. Dear Doctor. We need a video about channel estimation using deep learning
Thanks for the suggestion. I'll add it to my "to do" list.
Hi Professor, you explain SO WELL
Glad you think so!
hello professor lain,
I want to ask that what are the effects a channel cause to a signal in terms of amplitude and phase and frequency to a channel?
and how does an Equalizer able to correct these effects of channel means basically what functions are performed by Equalizer
Channel amp and phase are affected by things such as multipath, which can also cause ISI, which is what the equaliser is needed to combat. Perhaps watch "What is Intersymbol Interference ISI?" ua-cam.com/video/I087FUvW2ys/v-deo.html and "What is Rayleigh Fading?" ua-cam.com/video/-FOnYBZ7ZfQ/v-deo.html
sir in case 4 u have taken OFDM so the sole purpose of ofdm technology is to eliminate ISI or convert a frequency selective channel to frequency flat channel, than what is the purpose of doing equalization as equilization is used to remove the effect of ISI by opening the eye in eye diagram which has been already done by dividing the channel in to frequency flat sub bands by OFDM systems??
I'm not really sure what you're asking sorry. In the case of OFDM there is no "eye diagram", since all the sub channels are transmitted at the same time. But yes, each sub channel is flat fading (once the FFT in the receiver is applied). The sub channels still need to have their amplitudes adjusted/scaled so that the overall gain is unity, so that the detector can map the received signal onto the constellation points. This overall FFT-with-subcarrier-scaling can be viewed as the "equaliser".
hey Lain is it possible to equalize a non linear system (non linearities in phase and amplitude)
Sure, as long as the channel gain response is a one-to-one function. You need a filter which has an impulse response that is the inverse of the channel's impulse response (or approximates it). This is often not easy to find though, as the problem can be non-convex.
Thanks Mr Iain. Is this explanation in time domain or frequency domain. please help to confirm.
how did you do it can you share with me , thank you
Sorry, I'm not sure what you are asking.
Hello professor
Can you make a video on linearization of linear amplifiers
I guess you mean linearization of _non_ linear amplifiers, right? Thanks for the suggestion, I've added it to my "to do" list.
great comparison and easy to understand !
Glad you liked it!
Very clear and detailed explanation. I love your videos.
Can you please add the equalization gain of the mmse vs the dfe from a pure mathematic form? Meaning how much db do we gain with mmse vs dfe.
Thanks, I'm glad you like the videos. There's not a simple answer to your question, since the performance of the DFT is hard to quantify mathematically, due to the difficulty in modelling the error sequences that result from an incorrect single error being fed back. It depends on the amount of ISI in the channel and the SNR.
Some segments in the video are stamped not adjacent to each other
Sorry, I don't know what your comment means.
HI, Prof. Thanks for such nice lecture. Could you please explain 1) why do we put tail symbols and 2) why do we take it same as the training symbols: (do we put it in order to protect the next training symbols) or it also has some other purposes.
ISI in the channel spreads out each input symbol over multiple symbol time slots. So, if the detector cutoff its measurements exactly at the time of the last data symbol in a packet, then it would lose the energy from the last few data symbols that arrives after that time. That would make it harder to make decisions on those symbols (compared with the other earlier symbols). So the received vector typically includes a few symbols extra (and correspondingly, dummy "tail symbols" are added to the end of the packet before transmission, so it doesn't matter that they spread out). If you use the same "dummy" symbols for the last few symbols of the header, then the sampled channel matrix will be circular, which has advantages for implementation, but it's not necessary.
@@iain_explains Thanks. Got ur point. I have one more question. How should we choose the length of the data symbols and training+header for good equalization? Does this process of choosing the lengths of header+data+tail depends on the choice of equalization.
The choice of the lengths of the header, data, and tail depend on many factors. They are defined in the relevant standards, so if you are transmitting according to a standard, then you need to use the lengths defined in that standard.
@@iain_explains Dear Professor, As you mentioned in above comment Why would a detector even cutoff its measurements exactly at the time of the last data symbol in a packet?
thank you so much. Much needed content. Please make some video on detection of time offset and carrier freq offset for non-OFDM general modulation cases, like Gardner's method. I always find them difficult to understand from the books
Thanks for the suggestion, I've added it to my "to do" list.
Hi Professor,
Enjoy your style of explaining communication concepts :). Wonder if you plan to give a talk on partial response? Look forward to your lecture on the topic.
Thanks for the suggestion. Can you be a bit more specific though, about what you're unsure about? The term "partial response" can mean a few things.
@@iain_explains i think they meant partial response signalling. which is something like duobinary signalling according to what i've learnt. I haven't learnt this fully, so i'm not exactly sure what it means. From what i've found online i think it means a scheme which allows for ISI between two adjacent terms for some symbol in the sequence. Thanks for the lecture, it's very helpful :)
fantastic of fantastic
That's great to hear.
Character In the video It's great, I like it a lot $$
Glad you liked it.