3 ms·

Without knowing the details (which are still secret for good reason), before transmitting a packet to a specific device, the system computes an ideal transmissi

by lutusp 13y ago

Without knowing the details (which are still secret for good reason), before transmitting a packet to a specific device, the system computes an ideal transmission wavefront from multiple antennas to maximize the signal quality at that specific device. This is not a new idea -- phased array radars do essentially the same thing -- but it's a new application of the mathematical methods, and the complexity is much higher than for a modern radar because the system has to deal with many different devices, freely moving around, and calculate the optimal combination of radio signals to produce a high signal level at that devices's current location. It is the high computation workload that makes Perlman emphasize the role of a data center, which would need to be added to an existing cell network, to make this method possible. Cell systems already know how to determine the location of a given device using signal arrival times at multiple antennas. This method builds on that knowledge base to improve the outgoing signals, and possibly also to phase-adjust the reception antennas, to maximize signal quality in both directions. In essence, the system computes a protocol for each outgoing packet, beyond the packet's data content. Each packet contains additional information for each antenna that will be used to transmit it, information about phase delay. Before the packet is transmitted, the required phase delay for each of the transmitting antennas is adjusted to correspond to the calculated best solution for that device at its present location. This process is performed for each packet, for each device, which explains the high computation workload. If physically separate antennas are used for transmitting and receiving (usually true in cell service), a separate computation would be required to most efficiently receive packets from each device. I don't know if the system has this added level of complexity, but it's possible in principle.

foobarqux 13y ago
> This is not a new idea -- phased array radars do essentially the same thing -- but it's a new application of the mathematical methods[...] Yes, it's called pre-coding, every modern wireless system does it. Doing it across wireless stations is distributed beamforming which Perleman did not invent.
- lutusp 13y ago
  > Yes, it's called pre-coding, every modern wireless system does it. Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new. > Doing it across wireless stations is distributed beamforming which Perleman did not invent. That's certainly true -- phased array radars have been doing this for decades. But Perlman should be able to get a patent anyway, based on the combination of ideas this scheme represents. Multiple, physically separate antennas, the use of a data center to compute solutions on the fly to accommodate multiple antennas and different devices with different locations, that's new.
  - foobarqux 13y ago
    > Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new. You are just describing MIMO pre-coding which has existed for more than 20 years. I don't see what is patentable: Distributed beamforming necessarily requires physically dispersed antennas. Adding a data center doesn't seem patentable. But I don't know anything about patents nor do I care that much.
    - lutusp 13y ago
      > I don't see what is patentable ... Being a critic of software patents myself, and given how much this idea relies on software and mathematics (mathematics is definitely unpatentable), I agree that it's an issue. I suspect that the entire method is patentable even though each element taken separately isn't.
      - darkmighty 13y ago
        Yea the novelty is essentially that he's (allegedly) managed to make MIMO work with "non-local" antennas, which is really tricky in practice due to the sub-nanosecond phase synchronization required.
toomuchtodo 13y ago
> This process is performed for each packet, for each device, which explains the high computation workload. Could this not be sped up in ASIC hardware? I'm not a hardware guy, but I would think a GPU could fit the bill for these sorts of geospatial on-the-fly calculations. EDIT: "Accelerating geospatial analysis on GPUs using CUDA" http://www.zju.edu.cn/jzus/opentxt.php?doi=10.1631/jzus.C1100051 http://www.zju.edu.cn/jzus/opentxt.php?doi=10.1631/jzus.C110...
cookingrobot 13y ago
If the wavelength at 700Mhz is 43cm, wouldn't this have to track the position of the phone to within ~5-10cm to get a benefit? Or at least relative distance between antennas.
- lutusp 13y ago
  The present cell system can locate a cell device within the same limits using time delay and phase detection of the received signal (this requires three antennas for a 2D fix). Many cell phones that claim to have GPS actually rely on the cell network to locate them, and the location method is reasonably accurate. That means the system already knows where each device is located using the same wavelength constraints required to make this idea work. Link: http://en.wikipedia.org/wiki/Mobile_phone_tracking http://en.wikipedia.org/wiki/Mobile_phone_tracking Quote: "Mobile phone tracking refers to the attaining of the current position of a mobile phone, stationary or moving. Localization may occur either via multilateration of radio signals between (several) radio towers of the network and the phone, or simply via GPS."

by lutusp 13y ago

foobarqux 13y ago

> This is not a new idea -- phased array radars do essentially the same thing -- but it's a new application of the mathematical methods[...] Yes, it's called pre-coding, every modern wireless system does it. Doing it across wireless stations is distributed beamforming which Perleman did not invent.

lutusp 13y ago
> Yes, it's called pre-coding, every modern wireless system does it. Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new. > Doing it across wireless stations is distributed beamforming which Perleman did not invent. That's certainly true -- phased array radars have been doing this for decades. But Perlman should be able to get a patent anyway, based on the combination of ideas this scheme represents. Multiple, physically separate antennas, the use of a data center to compute solutions on the fly to accommodate multiple antennas and different devices with different locations, that's new.
- foobarqux 13y ago
  > Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new. You are just describing MIMO pre-coding which has existed for more than 20 years. I don't see what is patentable: Distributed beamforming necessarily requires physically dispersed antennas. Adding a data center doesn't seem patentable. But I don't know anything about patents nor do I care that much.
  - lutusp 13y ago
    > I don't see what is patentable ... Being a critic of software patents myself, and given how much this idea relies on software and mathematics (mathematics is definitely unpatentable), I agree that it's an issue. I suspect that the entire method is patentable even though each element taken separately isn't.
    - darkmighty 13y ago
      Yea the novelty is essentially that he's (allegedly) managed to make MIMO work with "non-local" antennas, which is really tricky in practice due to the sub-nanosecond phase synchronization required.

toomuchtodo 13y ago

> This process is performed for each packet, for each device, which explains the high computation workload. Could this not be sped up in ASIC hardware? I'm not a hardware guy, but I would think a GPU could fit the bill for these sorts of geospatial on-the-fly calculations. EDIT: "Accelerating geospatial analysis on GPUs using CUDA" http://www.zju.edu.cn/jzus/opentxt.php?doi=10.1631/jzus.C1100051 http://www.zju.edu.cn/jzus/opentxt.php?doi=10.1631/jzus.C110...

cookingrobot 13y ago

If the wavelength at 700Mhz is 43cm, wouldn't this have to track the position of the phone to within ~5-10cm to get a benefit? Or at least relative distance between antennas.

lutusp 13y ago
The present cell system can locate a cell device within the same limits using time delay and phase detection of the received signal (this requires three antennas for a 2D fix). Many cell phones that claim to have GPS actually rely on the cell network to locate them, and the location method is reasonably accurate. That means the system already knows where each device is located using the same wavelength constraints required to make this idea work. Link: http://en.wikipedia.org/wiki/Mobile_phone_tracking http://en.wikipedia.org/wiki/Mobile_phone_tracking Quote: "Mobile phone tracking refers to the attaining of the current position of a mobile phone, stationary or moving. Localization may occur either via multilateration of radio signals between (several) radio towers of the network and the phone, or simply via GPS."