Slashdot Mirror
Turbo Codes Promise Better Wireless Transmission
captain igor writes "IEEE is running a story about two French professors that have created a new class of encoding, called 'Turbo Codes,' that will allow engineers to pass almost twice as much data through a given communications channel, or equivalently, the same amount of data at half the power. The new codes allow the Shannon Limit (the theoretical maximum capacity of a channel) to be approached to, currently, within .5 dB. Scientists hope that this breakthrough will revolutionize wireless communications, especially with the coming reclamation of large swaths of the EM spectrum." As the article points out, such codes are in use now, but seem poised for much wider implementation.
Like the article says, these codes were introduced in 1993. This would have made a good story - back then.
The problem is that turbo codes are so computationally intensive that using them in consumer electronics is only new becoming feasible.
Did you know about it? Did most people know about it? If something is making it more feasible to use, then its news. Dr. Evil wanted to put laser beams on sharks *years* ago, but no one's really done it. If something made it possible to put frikking laser beams on sharks today, it'd be news.
It's not that they are just now feasible, it's that few people outside of satcom were hip to them. Processor performance is not much of a factor because you don't do turbo coding in software. It's accelerated in hardware and is performed in line as the datastream passes through the mod/demod sections. There are chips that are not too expensive that do this or a turbo coding core can be dropped into the ASIC that holds the mod/demod sections. That makes it pretty cheap for a consumer device.
People have been working on them for ages and yes there are significant advantages.
However, the latest word on source and channel coding though, is Space-Time Coding. Especially convolutionsal S-T codes are very very promising and quite naturally perform even better than block S-T codes...
What I don't understand is why this now? It's like running a feature on GSM, instead of writing about TDD or FDD in 3G, or even the discussion going on about how 4G will shape out to be...
/. Where the truth
I knew about it, as did many other people. But you have to realize that coding theory is a pretty funny and insular field. Related techniques had been used in other fields for many years prior to that discovery. Most people who work in this general area of statistics simply don't think about coding and aren't interested in it. One of the obstacles is that people who build communications systems generally are engineers thinking about fast, low-level processing; their first reaction to anything non-trivial and new is that it's too slow to be implemented in practice.
Turbo coding is ultimately not much of a theoretical breakthrough, but a compromise and algorithmic hack that happens to work fairly well for real-world problems and is expressed in a language that people who work on communications systems understand. But that's nothing to be sneezed at, since it will ultimately mean that we will get higher data rates and other benefits in real-world products.
1 dB of gain is equal to a million dollars of annual revenue for a wireless provider
.5 dB. The .5 dB that's left would bring in only $500,000 of revenue?
So, turbo codes have brought the gain differential (sorry, don't know the proper term) to
That doesn't sound like terribly much, considering how much money those companies are pulling in already.
This space intentionally left blank.
I always enjoy the moment in history when theory becomes practice.
1904 Einstein predicts the energy released from nuclear fission (E=MC^2). ~1938 first atom split, the equation was correct.
FM radio was therorized for many years, but until Amstrong came up with his Phase Lock Loop none could make an FM radio capable of broadcasting more than about a hundred feet.
CDMA has been around for a while too, Qualcomm doesn't own the pattent on it, just some techniques for practically implementing it which wasn't possible until microprocessor and DSP technology got sufficientlt small and powerful to fit in a cell phone.
This is the difference between science and engineering and I think it's very fitting that an engineering journal (that's the last E in IEEE) is pointing out that a bit of science has finally become interesting to a rather large group of engineers.
You must remember that LDPC codes rely upon block (Codeword) lengths of many bits, e.g. over 10,000 bits long in order to achieve performance better than turbo codes. So your parity check matrix is enormous.
I'm sure there are some efficient implementations, but for certain applications having packets that long can be prohibitive.
Just because a company's revenue continues to grow doesn't mean that company's profit margin is a constant percentage... A multi-billion dollar corporation could very easily be making only five-hundred thousand in profit. Despite the size of the comany, adding another five hundred grand to their profit is still twice what they were making.
tasks(723) drafts(105) languages(484) examples(29106)
I haven't heard the 1M figure before, but I bet that it is per tower, or per tower in densely populated areas.
Right, Gallager worked out LDPC codes in 1963. Then they were forgotton for 20 years until people realized that Digital Fountain's "Tornado" codes were LDPC codes.
LDPC codes will be behind DVB-S2, the new transmission system for digital satellite video distribution. Since they approach the Shannon limit so closely, there will be no DVB-S3.
I should say that the IEEE article is a little over-hyped, in that these codes really only buy about 2-4 dB additional gain, concatenated RS and convolutional coding were pretty close to the Shannon limit in AWGN, but those last couple of dBs were nice, but the remaining 0.5-1 dB beyond LDPC & Turbo Codes isn't worth much.
Much more important now are ways to handle "fast fading" channels found in mobile environments, this is what is driving OFDM.
Also, both Turbo Codes and LDPC codes are really computationally intensive to decode. They are currently only decoded at speeds below 20 Mbps, generally implemented as (expensive) FPGAs. We won't see real cheap ASICs for another year or two.
That's a fair answer :-) I always assumed that the latency on 1xEV-DO was too high to run voice over IP. (The reviews I've read on Verizon's EV-DO seemed to imply that the latency was too high to handle VoIP...however, it still is better than GPRS/EDGE)
:-)
Also, (please correct me if I'm wrong), wasn't EV-DV designed to treat all (non-legacy) calls as data? (What I am asking is that all EV-DV calls are VoIP calls, and voice & data calls share the same service option number?) (I'm also of the assumption that EV-DV and EV-DO are two complementary (and incompatible) technologies, but I assume both use Turbo codes to maximize spectral efficiency?)
Again, outside of IS-95 and 1xRTT, my knowledge on CDMA is quite low! When your company makes money off of currently deployed technologies, there isn't much incentive to retrain until there is a larger market demand...plus I haven't seen many good EV-DO/EV-DV documents online
Doh!