Viterbi equalizer

Started by roblkc3 2 years ago4 replieslatest reply 2 years ago183 views


I have been watching an old video from fred harris (this one) and there the Viterbi equalizer used in GSM is mentioned as one of the best ideas ever. However, that was recorded in 2011. Is there still any interest in the Viterbi equalizer (or, more generally, in Decision Feedback Equalization) ten years later?

We have plenty of computational power now, so this is not an issue for sure. I have read a couple of quick introductions to DFEs and I guess they can quickly diverge if they start making a set of bad decisions (but I do expect that someone figured out clever ways to prevent that). Also, since we have more and more multi-carrier protocols, one-tap equalizers might suffice to get decent equalization (and, therefore, no need to get the added complexity) -- but still, some "important" protocols, such as SC-FDMA, are single-carrier.

Are there any other reasons for not seeing these DFEs (and the Viterbi equalizer in particular) everywhere?

Thanks in advance ! :)


[ - ]
Reply by Tim WescottMay 6, 2021

I did one project where I was literally trying to pull valid data out of the mud*.  I used what was essentially a DFE for this -- but each decision came with an estimate of quality, and if the quality was low then the decision wasn't used, and the equalizer adaptation was frozen until good decisions started happening again.

* No, really -- it was mud-pulse telemetry for drilling rigs.

[ - ]
Reply by friedmanMay 6, 2021
SC-FDMA can be viewed as a flavor of OFDM (DFT-precoded), therefore can be equalized similarly - no need for expensive time-domain techniques.

I guess you could find such time-domain equalization techniques in other (non-OFDM) contemporary modems (e.g. Bluetooth).

[ - ]
Reply by roblkc3May 6, 2021

Thanks for your replies!
In the video harris indeed says that the equalization complexity pushes people towards OFDM-style equalization, and since those are more and more common nowadays, mine could simply be a distorted view of the field :)

@Tim Wescott: thanks, you really made me laugh hard with your mud-data-pulling :D

Have a nice day!!


[ - ]
Reply by Mannai_MuraliMay 8, 2021

Instead of single carrier when we use OFDM the symbol duration becomes large to combat delay spread.For the same mobile speed and hence the doppler the normalized doppler (Doppler in Hz*Symbol Period in sec) becomes large.So fading becomes fast fading and ICI hurts.In that case don't we need more complex equalization technique in frequency domain (Not simple one tap frequency domain equalizer).Does this not increase the complexity?What we gained in time domain has been lost in frequency domain.Any comments or discussions welcome.