Reading up on the differences between DDR5 and DDR4 PC Memory and they mention how DFE enables higher IO speeds and data rates! Thank you for explaining this.
Ah, that's an interesting application. Thanks for mentioning it. I always make a point of mentioning to students in my classes, that storage systems, such as hard drives, CDs, DVDs, RAM, ROM, etc., are all versions of communication systems (it's just that they have a very long delay in propagation - ie. the time between storage and whenever the data is retrieved).
Good question. You need to learn them by sending known (training) data, and estimating the channel. Here are some videos: "Channel Estimation for Mobile Communications" ruclips.net/video/ZsLh01nlRzY/видео.html and "Quick Introduction to MIMO Channel Estimation" ruclips.net/video/UPgD5Gnoa90/видео.html
L is the number of ISI terms. L times the symbol period equals the delay spread. For more details, see: "What is Intersymbol Interference ISI?" ruclips.net/video/I087FUvW2ys/видео.html
Thank you, Iain for your great video. As far as I know, the DFE can be used to estimate the phase noise in OFDM system, especially for the ICI. Can you explain how it works? Thank you.
DFEs only work well in conditions where the decisions are reliable (for example in high SNR). When data symbol decisions are reliable, they can be used effectively in exactly the same way as training data. This is called "decision directed training", and it can be used to estimate/track channel parameter changes, for parameters such as phase noise, carrier locking, and fading.
Sir do you familiar about Serdes? Can you please make videos on serdes equalizers, explained in both time domain and frequency domain so that we may have the visualizations
"SerDes equalizers" are just normal standard equalisers. Serialising a parallel data stream simply gives a single data stream at a rate that is N times higher than the rate of the individual parallel data streams. This means that ISI will be greater over the same channel (since the data symbols are shorter), which means equalisation is more important. But it doesn't mean the equalisation challenge/task is any different to "standard" equalisation.
Hi Iain, as usually very good explanation. In case of LTE OFDM, can we say that reference symbols are utilized as training symbols? If this is the case, that sounds as option 2 (from pilot symbols inserted throughout the data stream and/or across the frequency sub channels) from your reply how channel can be measured in three different ways. Thanks in advance and wish you all the best in 2023!
Yes, that's right. It's important to keep in mind that the standard defines certain things that need to be included in the data stream (by the transmitter), in order to enable certain particular functions to be performed at the receiver - but it doesn't stop the receiver from using those elements of the data stream for other purposes too.
Thank you, this very good explanation. Q: What is the difference between DFE and FFE? Why are called DFE non-linear vs FFE linear? If you have already made a video please can you point it out. I checked your page but could not find it. Thank you again for making all these amazing learning videos.
Great question. I've added it to my "to do" list. In summary, the Feed Forward Equaliser (FFE) is just a Finite Impulse Response (FIR) filter (which is a linear filter) where the tap values are chosen to try to "invert" the effect of the Inter-Symbol Interference (ISI) from the channel. In contrast, the DFE makes "hard" decisions and feeds them back, to cancel out ISI from the received measurements (as explained in the video). This "hard decision" process is non-linear.
Hi Iain,. Excellent explanation. I hope the next topic in the same chain can be decision feedforward Equalizer. I have one question. In implementation terms, given that we know y and x0 , how do estimate h0 and H1. Or are h0 and H1 estimated on training signals? If L is say 32 then do we require atleast 32 training symbols to estimate h0 to h31 ?
Thanks for the suggestion. I'm planning a video or two on channel estimation. In summary, the channel can be measured in three main ways: 1. from the training symbols at the start of the packet (or in the middle of the packet in some wireless standards), 2. from pilot symbols inserted throughout the data stream and/or across the frequency sub channels, 3. in a decision-directed way, using the estimated/detected data.
@@iain_explains Hi Iain, thanks for one more excellent video. Can we combine all three methods for channel estimation? for example initially we can send some known symbols and then periodically use decision directed estimation?
Hi Sir! In 6:03, how did u use the training symbols to estimate h0 and h1? Also, I noticed that the noise isn’t included in your estimate. Can u give an example where u show that a symbol is approximated on the constellation? Thx.
Channel estimation is still on my "to-do" list. Hopefully I can get to it soon. And thanks for the suggestion about detection, and showing the constellation diagram. I'll add that to my list too.
Reading up on the differences between DDR5 and DDR4 PC Memory and they mention how DFE enables higher IO speeds and data rates! Thank you for explaining this.
Ah, that's an interesting application. Thanks for mentioning it. I always make a point of mentioning to students in my classes, that storage systems, such as hard drives, CDs, DVDs, RAM, ROM, etc., are all versions of communication systems (it's just that they have a very long delay in propagation - ie. the time between storage and whenever the data is retrieved).
Concept I’ve never heard ab in my studies. Will definitely look into this more, thanks!
Glad it was helpful!
Great explanation. I still have one question. How h0 and h1 are known beforehand?
Good question. You need to learn them by sending known (training) data, and estimating the channel. Here are some videos: "Channel Estimation for Mobile Communications" ruclips.net/video/ZsLh01nlRzY/видео.html and "Quick Introduction to MIMO Channel Estimation" ruclips.net/video/UPgD5Gnoa90/видео.html
thanks for your practical videos.At 7:25 how did we know h0? h0 is not part of our training data!
All of the h's need to be measured using training data.
Hello, this is a excelent video. Please can you give what is L in this case i dont understand what is. I am very confused.
L is the number of ISI terms. L times the symbol period equals the delay spread. For more details, see: "What is Intersymbol Interference ISI?" ruclips.net/video/I087FUvW2ys/видео.html
Thank you, Iain for your great video. As far as I know, the DFE can be used to estimate the phase noise in OFDM system, especially for the ICI. Can you explain how it works? Thank you.
DFEs only work well in conditions where the decisions are reliable (for example in high SNR). When data symbol decisions are reliable, they can be used effectively in exactly the same way as training data. This is called "decision directed training", and it can be used to estimate/track channel parameter changes, for parameters such as phase noise, carrier locking, and fading.
very nice, could you further explain the strcture of the DFE?
Thanks for the suggestion. I'll put it on my "to do" list.
Sir do you familiar about Serdes?
Can you please make videos on serdes equalizers, explained in both time domain and frequency domain so that we may have the visualizations
"SerDes equalizers" are just normal standard equalisers. Serialising a parallel data stream simply gives a single data stream at a rate that is N times higher than the rate of the individual parallel data streams. This means that ISI will be greater over the same channel (since the data symbols are shorter), which means equalisation is more important. But it doesn't mean the equalisation challenge/task is any different to "standard" equalisation.
Thank you very much sir for the insight
Hi Iain, as usually very good explanation. In case of LTE OFDM, can we say that reference symbols are utilized as training symbols? If this is the case, that sounds as option 2 (from pilot symbols inserted throughout the data stream and/or across the frequency sub channels) from your reply how channel can be measured in three different ways. Thanks in advance and wish you all the best in 2023!
Yes, that's right. It's important to keep in mind that the standard defines certain things that need to be included in the data stream (by the transmitter), in order to enable certain particular functions to be performed at the receiver - but it doesn't stop the receiver from using those elements of the data stream for other purposes too.
very neat explanation as always. Thanks Ian!
Glad you liked it!
Thank you, this very good explanation. Q: What is the difference between DFE and FFE? Why are called DFE non-linear vs FFE linear? If you have already made a video please can you point it out. I checked your page but could not find it. Thank you again for making all these amazing learning videos.
Great question. I've added it to my "to do" list. In summary, the Feed Forward Equaliser (FFE) is just a Finite Impulse Response (FIR) filter (which is a linear filter) where the tap values are chosen to try to "invert" the effect of the Inter-Symbol Interference (ISI) from the channel. In contrast, the DFE makes "hard" decisions and feeds them back, to cancel out ISI from the received measurements (as explained in the video). This "hard decision" process is non-linear.
Very Good Explanation!
Glad it was helpful!
So helpful!❤ Thank you.
Glad you liked the video.
Hi Iain,. Excellent explanation. I hope the next topic in the same chain can be decision feedforward Equalizer. I have one question. In implementation terms, given that we know y and x0 , how do estimate h0 and H1. Or are h0 and H1 estimated on training signals? If L is say 32 then do we require atleast 32 training symbols to estimate h0 to h31 ?
Thanks for the suggestion. I'm planning a video or two on channel estimation. In summary, the channel can be measured in three main ways: 1. from the training symbols at the start of the packet (or in the middle of the packet in some wireless standards), 2. from pilot symbols inserted throughout the data stream and/or across the frequency sub channels, 3. in a decision-directed way, using the estimated/detected data.
@@iain_explains Hi Iain, thanks for one more excellent video. Can we combine all three methods for channel estimation? for example initially we can send some known symbols and then periodically use decision directed estimation?
very helpful,thx a lot😃
That's great to hear.
Hi Sir! In 6:03, how did u use the training symbols to estimate h0 and h1? Also, I noticed that the noise isn’t included in your estimate. Can u give an example where u show that a symbol is approximated on the constellation? Thx.
Channel estimation is still on my "to-do" list. Hopefully I can get to it soon. And thanks for the suggestion about detection, and showing the constellation diagram. I'll add that to my list too.
@@iain_explains Lian please include matlab programs in your vdeos to give insight into working and performance improvement of equalizers.
Schmidt Route
Reanna Lodge
57712 Thaddeus Parkway
Marcos Keys