Slashdot Mirror
Full-Duplex Radio Integrated Circuit Could Double Radio Frequency Data Capacity
Zothecula writes Full-duplex radio communication usually involves transmitters and receivers operating at different frequencies. Simultaneous transmission and reception on the same frequency is the Holy Grail for researchers, but has proved difficult to achieve. Those that have been built have proven complex and bulky, but to be commercially useful in the ever-shrinking world of communications technology, miniaturization is key. To this end, engineers at Columbia University (CU) claim to have created a world-first, full-duplex radio transceiver, all on one miniature integrated circuit.
Why? Is the spectrum that crowded that we need this? Making this work in the real-world will not be a piece of cake...
Karma: Bad
You should have posted a reply in your own post.
Tic-Tac-Toe, Global Thermonuclear War, and relationships all have the same winning move.
foxconn robots do it, and the first fully digital transciever too.
"all on one miniature integrated circuit."
That would be as opposed to all of the really big integrated circuits.
The article is misleading. Transmission and reception on the same "frequency" is done today. However, there's some other "discriminator" in the signal. Either modulation method, phase, shift, orientation, or "something" is different so that the receive and transmit don't collide.
This article -- despite its misleading introduction -- talks about a limited application whereby RX and TX can occur using the same frequency *BAND* (they say "spread spectrum") and allow full-duplex communication. The advance is that this is all on one chip.
What would be truly revolutionary, like the example of two people talking to each other at the same time, is the ability to transmit and receive using the *same* exact method by both transceivers. THAT would be the holy grail.
Not there yet.
E
Ar = Bt - At and Br = At - Bt
To make room for the possibility that both At and Bt are transmitting a max-modulation symbol at the same time, there will need to be additional bandwidth in the frequency range to accommodate. The additional bandwidth would be just about the same bandwidth as would be needed for a second frequency.
If there is a free lunch in this invention, it hasn't been adequately explained in the article.
(Yes. I actually don't know what I'm talking about, but feel free to remind me.)
Reducing the transmitted signal the 60-80 dB needed to bring the received signal into detectable range is quite a challenge. A cellphone transmits a three watt signal. It's rather hard to fight something like that.
that'd be like talking and listening at the same time using mouth only. no ears.
the conditions of test in the abstract listed were not mentioned. this is routine in cross-polarization (transmit vertical, receive horizontal) or in time-division (squintillions of telco, digital TV applications) multiplexing. but if this was going to happen on one little dinky antenna in the side of a smartphone at the same time on the same chip, the echo cancellation algorithm on chip would be the size of a SUV. and echo cancellation is sort of TDM all in itself.
I'll believe it when the flying pig hands the document off to my unicorn for translation.
if this is supposed to be a new economy, how come they still want my old fashioned money?
The Plessey Groundsat had simultaneous same channel repeat (insert technical name here) back in 1980. I wonder if the lack of real world applications today are limited by crazy patents on this kind of technology.
The (copper) Gigabit Ethernet PHY transmits and receives simultaneously on four wire pairs. It accomplishes this with a hybrid that subtracts the transmitted signal from the one being received. Last year some newer WiFi access points debuted that could do the same thing with RF. (Gigabit Ethernet is technically RF too because each of the four wire pairs operate at around 125MHz. WiFi access points operate in the 2.4GHz and 5.4GHz bands.)
By definition its not full duplex if its using a shared channel to transmit and receive.
If you define full duplex in your own bizarre way, then, yes this is not full duplex. The more common definition, however, is that transmitting and receiving can be done simultaneously. And that is exactly what is going on here. Obviously using a shared channel, otherwise it would not be news. And even that is nothing new. Echo canceling and circulators have been existing for ages. The novelty here is the size of integration.
The issue is that a strong transmission in the same band as a receiver can desense the receiver. This can also be done with a cavity duplexer if you need input and output in the same band on adjacent frequencies, but you pay for it with geometric space (since cavity duplexer dimensions are a fraction of the wavelength in free space multiplied by the materials velocity factor). This can be problematic on HF and VHF bands, but UHF and microwave can get away with duplexers the size of a brick. Unfortunately, that's still too much for mobile phones since it's too big to fit in someone's pocket.
In the era of Big Data, the current frequency spectrum crisis is one of the biggest challenges researchers are grappling with...
Articles that throw buzzwords around are annoying, but irrelevant buzzwords are even worse!
The article is misleading. Transmission and reception on the same "frequency" is done today. However, there's some other "discriminator" in the signal. Either modulation method, phase, shift, orientation, or "something" is different so that the receive and transmit don't collide.
Actually, bidirectional, simultaneous transmissions using exactly the same polarisation, modulation etc have been possible for a long time, using circulators/hybrids and echo cancelers. I imagine they had limited succes because typically the power difference between transmitted and received signal is too high for the echo canceler to deal with, but in theory, this "holy grail" is certainly possible.
Apart from that, as you mention correctly, the novelty here is the size.
true, but "channel" is not necessarily equivalent to "frequency", and it sounds like they're just looking at sharing the frequency. TDM would indeed count as fast switching half duplex, but polarization might not.
The word geek is in your username yet you come here and ask WHY? What the hell is wrong with you? You aren't a geek, you're a poser. Hand over the card!
As described, It is full duplex - both sides transmit simultaneously and have to cancel their own signal and its echoes from the combined received signal. Twisted-pair Ethernet already works this way, but in the radio medium the echoes must be even more challenging to model and cancel.
Not, really. It's more like screaming at someone that is far away while he is screaming at you and trying to hear at the same time.
The problem is that you hear your own screaming a lot better than the one far away.
One would think that it is easy to just subtract the sent signal from the received one, but with all the amplifiers necessary the tiniest noise in the transmitter path will just drown out the receiver completely.
Hi all, I was perusing through all the comments, and as one of the authors of the work, I thought I would clarify some of the points that were raised to aid the discussion: 1. The chip targets same-channel full duplex, meaning the transmitter and the receiver work in the same frequency channel at the same time, and are not separated by polarization, modulation format etc. Therefore, since transmitted signals are around +20dBm and receiver sensitivity levels are around -90dBm, nearly 110dB of suppression through isolation (across a pair of antennas or a circulator) and echo (aka self-interference or SI) cancellation must be achieved (as one of the people above has correctly pointed out). Such a high degree of SI cancellation requires that SI cancellers be implemented in all domains (RF, analog and digital, each yielding a part of the total SI suppression). 2. As one of the people above has pointed out, even if the signals were separated in modulation format for instance, the transmitter SI would be so powerful that it would saturate the receiver front end before modulation-format-based separation can be achieved in the digital domain. So echo cancellation at the receiver front end is required. 3. As someone points out, circulators and echo cancellers have existed for quite a while and have been implemented in many ways. The innovation here is that we perform echo or SI cancellation at RF in a single chip, which has not been done before. 4. Moreover, the SI cancellation approach can tackle echos that experience significant delay (as high as 20ns) while still fitting with an IC form factor through the use of on-chip reconfigurable high-Q filters, enabling cancellation of wideband signals (>20MHz enabling use for WiFi). 5. Finally, indeed the varying environment is a challenge and the RF and digital SI cancellers need to be reconfigured periodically (milli-seconds). Hope this helps.
Bingo. The "obvious" solution is to automatically develop a transfer function between the transmit side and the receive side, so that any signal transmitted can be subtracted from the "received" signal. The problem is that the transmitted signal is many orders of magnitude stronger than the received signal, making the signal-to-noise ratio very poor. Any noise within the circuitry that performs the transform/subtract process needs to be absolutely miniscule in order for this to work. That is challenging enough with large components that can be made to perform with low noise, but as the components get smaller, thermal noise becomes significant, limiting performance. That this can now be done on a single IC sounds like an advance in low-noise amplification techniques, more so than advances in transmission technology.
C'mon folks, this is just doing with one IC chip what has been done before with more IC chips, and even years ago with discrete components. Nothing to see here, move along!
On the basis of trusting that the AC truly is one of the authors (of the scholarly paper), I want to thank you for these clarifications and suggest to all to mod that post up. It definitely is better than score: 1, which is its current value at the time of my writing.
110 dB of SI cancellation is beyond impressive - it is approaching magical!
On the face of it, this capability will double capacity of any RF channel for which it will work. AC claims this can be made to work on channel bandwidths exceeding 20 MHz, therefore making it useful for WiFi.
But I think there are other advantages. If a traditional system uses FDD (frequency division duplex) to achieve duplex (simultaneous transmit and receive) operation, then this new technology reduces by half any discrete RF/IF filter hardware needed to reject out-of-channel energy. That will help make the electronics simpler and less expensive. For FDD, the cost of the filters goes up as the two channels (transmit and receive) get closer together (the closer TX is to RX, the steeper the filters have to be to achieve adequate rejection). With this all-silicon approach, the most you need is bandpass filtering for the ONE channel you are using. Big win!
But then maybe I am exposing my dinosaur-like thinking in even bringing up discrete RF filter components. A recent announcement at Mobile World Congress touted a silicon-only radio technology that didn't appear to need any discrete filtering at all.
Also my (dinosaur-vintage) thinking about cellular base stations is that they generally operate in the +40 to +50 dBm range (out of the PA, prior to duplexers, etc. and not considering antenna gain), which implies another 20-30 dB isolation is required (vis-a-vis the AC's claim of 110 dB) to achieve the same isolation one would need in a cellular system. But then I'm not considering antenna gain which seems (without thinking about it too hard) to potentially improve the isolation if separate TX and RX antennae are used at the base station. Then again, I'm thinking macrocells here. But for a single channel duplex RF technology to be deployable in cellular, I think one would need to cover the macrocell case - in any case.
Krishnaswamy adds. “It will be very exciting if we are indeed able to deliver the promised performance gains.”
flatulus: Thanks for the comments. They are spot on.
- It is true that there are benefits beyond full duplex, namely in reducing duplexer filter requirements for FDD. We have received commercial interest for this application as well. LTE provides support for 24 FDD bands, a lot more than 3G. Having 24 fixed-frequency duplexers in a handset is near impossible. So, there is interest in tunable duplexers that can cover multiple bands but inevitably have reduced isolation and greater insertion loss than conventional fixed-frequency duplexers. Self-interference cancellation can be used to enhance the isolation back to the 55dB levels seen in conventional fixed-frequency duplexers. In our ISSCC paper, which has not yet been uploaded to IEEExplore, we show measurements for the FDD use case as well.
- It is true that cellular base stations will require 20-30dB higher isolation. I think the higher-gain antennas do offer an isolation advantage but not as much as the increase in antenna gain because the antenna-to-antenna coupling is a near-field phenomenon. In base stations, where form factor is less of a concern, discrete-component based approaches can be used as a first line of defense, followed by IC-based fully-integrated cancellation. Also, WiFi base stations and cellular small cells have lower transmit power levels and so are more direct applications for this IC technology.
- I did not follow the MWC, but as far as I know, in the literature, SAW-less receivers for TDD have been reported and have made it into phones, but duplexer-less receivers for FDD have not yet been reported. The SAW-less receiver problem is easier because one has to deal with jammers picked up from the environment, which tend to be a lot weaker than transmitter self-interference.
I hate to burst CU'S bubble but this has been done numerous times over a decade ago by researchers at UC Berkeley and other institutions. Search on Google scholar for SiGe.
This can't work in multi-path channels. The RF circuitry and balun would take care of removing most of the first path, but in a multi-path channel the 2nd echo of the TX signal will still hit the RX saturating the ADC.... you just need a 32-bit ADC, an NNA (instead of LNA) and cancel the interference in digital domain? Come on, be serious.
And what about the TX signal of a device talking beside yours? How do you cancel that interference if you don't even know that signal?
This will never make it into a real product...
Can somebody provide the link to CosmicLab's work presented at ISSCC?