You can view the Siemens press release (http://www.siemens.com/index.jsp?sdc_p=cfi1232554 lmno1232554ps5uz1&
)
I'm not quite sure why this is news for slashdot. Maybe it's a quite news time.
A couple of points:
It's not existing hardware. It needs multiple receivers and multiple transmitters.
They use 100 MHz bandwidth, so would expect to get a factor of 5 increase over the 54Mbps delivered by 802.11a simply through that.
They use 4x3 MIMO which has been around for a while, see particularly the BLAST work done by Chris Nicol & co at Lucent.
(Curse the submit button being right next to the preview button)
It's a lame answer, but... it depends. Going from 2 to 5 GHz does result in an increase in attenuation, particularly the exponent in the expression. Another obvious change is the multipath. At 5GHz it's somewhat narrower band than at 2.4GHz (simply due to the relationship between delay and carrier frequency), which has benefits in the OFDM rates because the FEC across subcarriers works better with narrow-band fading. At 5Ghz, the signal beam doesn't defract quite as much as 2GHz, so trees etc are slightly more opaque, but off the top of my head I can't say just how much.
As a general comment, the frequency dependence of range exists, but isn't as significant as the extra signal needed by the OFDM rates.
I... never experience any symptoms of interference
That's most likely because, as you said, your traffic requirements are low, and possibly the traffic on the newtworks you can see isn't particularly heavy. If you have access to the PHY layer, you will see that collisions are in fact very common. The standard provides a couple of ways for dealing with this. (I'm sorry if I'm teaching you how to suck eggs here - I don't know what you know, so I'm aiming low).
At the base level, each data packet is acknowledged by the recipient. If it isn't acknowledged (an ACK) then the packet is re-sent. Depending on the particular manufacturers implementation, the retries continue, and if still unsuccessful, at some point the rate is dropped (the lower rates are more robust). Eventually, if there's no acknowledgement, the packet is dropped.
At the higher level the MAC can take advantage of RTS/CTS. In this mode, before sending a data packet, a small "Request to Send" packet is transmitted - telling all the radios in range that it is about to send a packet and it will take such and such a length of time to do it - so please stay off the air. The recipient (the STA to whom the eventual data packet is addressed) responds with CTS - Clear-to-Send. Then the data transfer goes ahead. The RTS/CTS mechanism is designed to reduce collisions in heavily loaded networks. Unfortunately, it relies on all the radios which can make noise being able to hear and succesfully demod the RTS/CTS. Adjacent channels have the annoying property that they are load and interfere with you, but they are also incomprehensible. As a result RTS/CTS doesn't help with one of the more common mechanisms for collisions.
In summary, collisions do happen, and adjacent cells/frequencies do lower throughput. Many users never notice this because of their low data requirements. As many posters have noted, even 1Mbps is more than most people need at present. However, as requirements go up (streaming video etc), this will become a much more visible problem.
This is, unfortunately, a common misconception. Increasing the range of 11b means that the available bandwidth has to be shared more widely - meaning that each user gets less bandwidth, which is to say data throughput.
Imagine if you will, a world where you could hear everyone talking within a block of you. Sounds great - you can hear your stereo from a mile away (well, this already happens). Unfortunately you can also hear everyone elses stereo, and everyone else talking, and their refrigerators humming, and their dogs barking. The net result is that you can't hear anything clearly. It's very much the same as being in a very full pub towards the end of a horse race - everyone's making a tonne of noise, and you can't understand much of it at all.
The solution is to have smaller cells, in the limit pico-cells, where an individual user gets the full channel to herself, and 10m away another user gets their own full channel. An alternative is to have large cells, but with APs having different non-overlapping frequencies - so that the cells are isolated in the same manner as for the pico-cells. Sadly, the lack of spectrum for 11b at 2.4Ghz makes this second solution unrealistic.
No. 11a, 11b and 11g are physical layer (PHY) components of the 802.11 standard. The Medium Access Layer (MAC) which sits above the PHY layer is common (with very small differences) to all the PHY layers.
Security is specified by 802.11i (ratified in the last week or so) and applies equally to both 11a, 11b, and 11g. Until recently, they all shared the same poor excuse for security - WEP or proprietary extensions such as Cisco's LEAP. In the future, they will all share (when firms implement it) the Robust Secure Network properties of 802.11i.
802.11a... range is typically even more limited than 802.11b/g
In reality, 11a range, for the same transmit power, is not that bad compared to 11b. The limit to range is the sensitivity of the receiver - at what received signal strength does the packet error rate drop to an unacceptable level - usually taken as 10% but in reality somewhere between 1% and 10%.
For a well designed 11b receiver, the receive sensitivity for 1Mbps data rate is around -94dBm (i.e, the signal is about as strong as the noise at the receiver, and here we mean the noise floor due to the sky itself plus the receive amplifiers etc).
For an 11a receiver, the receive sensitivity for its lowest rate (6Mbps) is around -87dBm, which is to say the receiver needs 7dB or about 4 times as much power to receive 6Mbps as 1Mbps (not a bad tradeoff when you recall that you're receiving 6 times as much data).
Solutions to this: up the transmit power, or use high gain antennas for the 11a. (This is limited to an extent by FCC transmit spectral mask rules). At the same time, drop the transmit power on the 11b. This means that
The 11b range has been reduced - sharing the spectrum with fewer users, thus upping the throughput each user gets
The 11a range has been increased so that it is larger than the inter-cell(11b) spacing, allowing the cell APs to be meshed.
Beautiful.
(FWIW, I don't have any relationship with these people)
This solution seems to be quite a clever approach. A fundamental problem with 802.11b is the lack of spectrum. Although the channels are labelled 1 to 11 (in the USA), the numbers refer to a spacing of 5MHz chunks over 50MHz in the 2.4GHz ISM band. The problem is that the 802.11b signal uses almost 4 of those channels to actually transmit data. As a result, in order to have systems on different frequencies which are not interfering with each other, you end up with three effective channels - 1, 6, and 11. (If you have a WiFi AP accessible, check what channel it's on - most likely it will be one of those). Due to the low number of channels, it's impossible to do much in the way of channel planning. The result is that adjacent APs have to share the spectrum. The net outcome is that the data rate that users get between their client and the AP is reduced.
802.11a at 5GHz was supposed to solve this. The 5GHz band is notable because of the extra spectrum it has. Compared to the 3 effective channels at 2.4GHz, the 5GHz UNII band has (again, it depends on your country) at least 8 usable channels of 20MHz. Additionally, the link rate is between 6 and 54 Mbps (as compared to 1 to 11Mbps for 11b, although this is somewhat moot given the growing preponderance of 11g solutions at 2.4Ghz). However, the 802.11a market never really took off and killed the 11b market the way we (engineers) expected it to. Mostly due to good (if slippery) marketing of 11g.
As a result, there's a lot of unused 11a spectrum begging to be used. There are a lot of people with 2.4GHz equipment who want more range without losing data throughput. Using the 11a spectrum to extend the 11b/g range is what these guys have done. Neat - they get to use a superior technology with cheap chips available, to leverage a large market (albeit of dullards wed to an inferior solution).
From the herald article (first line), the child is one year old. So what are the chances given that age? (I recall from a friend doing a PhD torturing kittens that early visual development is critical, and one year sounds maybe a little late to start).
To the contrary, it's not at all uncommon in NSW (and the other states) for costs to be granted against the unsuccesful party in a civil matter. The amount of costs is however determined by the court.
For the interested./ers (and I can't imagine there are many who would cherish the thought of reading legislation) Part 11 of the
Legal Profession Act 1987 discusses some of the issues of costs as applicable in NSW.
One thing in the article is cute - the Power Line people invited the ARRL to be involved
The ARRL became involved in Spencer's case after United Power Line Council President William R. Moroney invited the League in mid-March to keep his organization in the loop on any cases of BPL interference that were not being satisfactorily addressed.
and the ARRL have repaid them by asking the FCC to close them down and fine them $10,000:
rhe ARRL's formal complaint to FCC Enforcement Bureau Chief David H. Solomon called on the Commission not only to close down Alliant's BPL field trial system but to fine the utility $10,000...
Nice. I'm sure comms companies all over the US will jump at the chance to get the ARRL's contribution and involvement in future.
Either way, it's great to see that the FCC is standing firm to protect sad lonely guys holed up in their bunkers listening to strangers over the airwaves from the interference of sad young(er) lonely guys holed up in their bunkers looking at strangers over the ether.
802.11h is a work in progress - referred to as Task Group h. After a task group completes and its subject is ratified, it is referred to (for instance) as 802.11h, until then, it's just a task group.
No names, no packdrill: A small number of steps down from Charlie.
Are you sure about those prices? Cisco didn't sell 11a client cards for very long, so it's possible that once the decision was made to get out of that market and concentrate on APs that some of the remaining client cards were dumped, but I'm almost certain that the 11a cards were sold for more than COGs. Maybe they didn't sell for the gross margin Cisco usually wants? Did they sell for more than what it cost to make them in terms of software & MAC firmware resources on clients - probably not.
Interesting. What did you think the component count was, and what counts as insane? I've got one of the first cards here, and although it's not a bare board it doesn't look insane. What sort of workarounds had to be made? I wasn't aware of any significant ones.
On the market capture - the statement about time to market with respect to.11a came from one of the directors in that BU, so no, I'm not very wrong.
You're understating some elements and overstating others here. Cisco bought Radiata for its PHY layer 802.11a chips - RF and baseband. Cisco didn't buy them because they'd made a lot of marketing smoke, they bought them because they had done a lot of technical due diligence, had seen exactly where the chip development cycle was, and felt it fitted with their business. As to the chips being buggy - Que? I'd heard there were problems with the MAC chips in clients (particularly the cardbus interface and associated driver software across a range of laptops), I hadn't heard about any PHY bugs. Cisco didn't use Radiata's MAC approach for its 802.11 products, using instead an internally developed MAC from the Aironet acquisition.
At the time of the Radiata acquisition, there were two realistic contenders for providing 11a chips - Atheros, and Radiata. A year after the acquisition there were still only two companies with genunine working silicon, Radiata (within Cisco) and Atheros. Sensitivity of the two chipsets was similar, performance in multipath was similar (being better for one or the other depending on who you asked; I've measured them both and found that from spin to spin they tended to leapfrog each other).
On the time to market - Cisco believes that it creates markets - the 11a market wasn't that big two years ago. Cisco felt it could take its time with its own 11a solutions, and when it did start to ship in volume, that would drive the 11a market all on its own. So I'm not sure where the Cisco engineers claims on slower time to market came from. I'm not sure that they would have been any quicker buying chips on the open market, because there wasn't really that much of an open market.
Disclaimer:
I don't work for Cisco, so this is just my take on what happened, not a report from inside Cisco
Slashdot Mirror
- It's not existing hardware. It needs multiple receivers and multiple transmitters.
- They use 100 MHz bandwidth, so would expect to get a factor of 5 increase over the 54Mbps delivered by 802.11a simply through that.
- They use 4x3 MIMO which has been around for a while, see particularly the BLAST work done by Chris Nicol & co at Lucent.
(Curse the submit button being right next to the preview button)As a general comment, the frequency dependence of range exists, but isn't as significant as the extra signal needed by the OFDM rates.
At the base level, each data packet is acknowledged by the recipient. If it isn't acknowledged (an ACK) then the packet is re-sent. Depending on the particular manufacturers implementation, the retries continue, and if still unsuccessful, at some point the rate is dropped (the lower rates are more robust). Eventually, if there's no acknowledgement, the packet is dropped.
At the higher level the MAC can take advantage of RTS/CTS. In this mode, before sending a data packet, a small "Request to Send" packet is transmitted - telling all the radios in range that it is about to send a packet and it will take such and such a length of time to do it - so please stay off the air. The recipient (the STA to whom the eventual data packet is addressed) responds with CTS - Clear-to-Send. Then the data transfer goes ahead. The RTS/CTS mechanism is designed to reduce collisions in heavily loaded networks. Unfortunately, it relies on all the radios which can make noise being able to hear and succesfully demod the RTS/CTS. Adjacent channels have the annoying property that they are load and interfere with you, but they are also incomprehensible. As a result RTS/CTS doesn't help with one of the more common mechanisms for collisions.
In summary, collisions do happen, and adjacent cells/frequencies do lower throughput. Many users never notice this because of their low data requirements. As many posters have noted, even 1Mbps is more than most people need at present. However, as requirements go up (streaming video etc), this will become a much more visible problem.
Imagine if you will, a world where you could hear everyone talking within a block of you. Sounds great - you can hear your stereo from a mile away (well, this already happens). Unfortunately you can also hear everyone elses stereo, and everyone else talking, and their refrigerators humming, and their dogs barking. The net result is that you can't hear anything clearly. It's very much the same as being in a very full pub towards the end of a horse race - everyone's making a tonne of noise, and you can't understand much of it at all.
The solution is to have smaller cells, in the limit pico-cells, where an individual user gets the full channel to herself, and 10m away another user gets their own full channel. An alternative is to have large cells, but with APs having different non-overlapping frequencies - so that the cells are isolated in the same manner as for the pico-cells. Sadly, the lack of spectrum for 11b at 2.4Ghz makes this second solution unrealistic.
Security is specified by 802.11i (ratified in the last week or so) and applies equally to both 11a, 11b, and 11g. Until recently, they all shared the same poor excuse for security - WEP or proprietary extensions such as Cisco's LEAP. In the future, they will all share (when firms implement it) the Robust Secure Network properties of 802.11i.
For a well designed 11b receiver, the receive sensitivity for 1Mbps data rate is around -94dBm (i.e, the signal is about as strong as the noise at the receiver, and here we mean the noise floor due to the sky itself plus the receive amplifiers etc).
For an 11a receiver, the receive sensitivity for its lowest rate (6Mbps) is around -87dBm, which is to say the receiver needs 7dB or about 4 times as much power to receive 6Mbps as 1Mbps (not a bad tradeoff when you recall that you're receiving 6 times as much data).
Solutions to this: up the transmit power, or use high gain antennas for the 11a. (This is limited to an extent by FCC transmit spectral mask rules). At the same time, drop the transmit power on the 11b. This means that
- The 11b range has been reduced - sharing the spectrum with fewer users, thus upping the throughput each user gets
- The 11a range has been increased so that it is larger than the inter-cell(11b) spacing, allowing the cell APs to be meshed.
Beautiful.(FWIW, I don't have any relationship with these people)
802.11a at 5GHz was supposed to solve this. The 5GHz band is notable because of the extra spectrum it has. Compared to the 3 effective channels at 2.4GHz, the 5GHz UNII band has (again, it depends on your country) at least 8 usable channels of 20MHz. Additionally, the link rate is between 6 and 54 Mbps (as compared to 1 to 11Mbps for 11b, although this is somewhat moot given the growing preponderance of 11g solutions at 2.4Ghz). However, the 802.11a market never really took off and killed the 11b market the way we (engineers) expected it to. Mostly due to good (if slippery) marketing of 11g. As a result, there's a lot of unused 11a spectrum begging to be used. There are a lot of people with 2.4GHz equipment who want more range without losing data throughput. Using the 11a spectrum to extend the 11b/g range is what these guys have done. Neat - they get to use a superior technology with cheap chips available, to leverage a large market (albeit of dullards wed to an inferior solution).
Rather than reading a journalists munged interpretation of what Symantec said, you can look at Symatec's original statement
For the interested ./ers (and I can't imagine there are many who would cherish the thought of reading legislation) Part 11 of the
Legal Profession Act 1987 discusses some of the issues of costs as applicable in NSW.
Why would you put extra hard drives inside your computer when you've got firewire?
Nice. I'm sure comms companies all over the US will jump at the chance to get the ARRL's contribution and involvement in future.
Either way, it's great to see that the FCC is standing firm to protect sad lonely guys holed up in their bunkers listening to strangers over the airwaves from the interference of sad young(er) lonely guys holed up in their bunkers looking at strangers over the ether.
Are you sure about those prices? Cisco didn't sell 11a client cards for very long, so it's possible that once the decision was made to get out of that market and concentrate on APs that some of the remaining client cards were dumped, but I'm almost certain that the 11a cards were sold for more than COGs. Maybe they didn't sell for the gross margin Cisco usually wants? Did they sell for more than what it cost to make them in terms of software & MAC firmware resources on clients - probably not.
On the market capture - the statement about time to market with respect to .11a came from one of the directors in that BU, so no, I'm not very wrong.
At the time of the Radiata acquisition, there were two realistic contenders for providing 11a chips - Atheros, and Radiata. A year after the acquisition there were still only two companies with genunine working silicon, Radiata (within Cisco) and Atheros. Sensitivity of the two chipsets was similar, performance in multipath was similar (being better for one or the other depending on who you asked; I've measured them both and found that from spin to spin they tended to leapfrog each other).
On the time to market - Cisco believes that it creates markets - the 11a market wasn't that big two years ago. Cisco felt it could take its time with its own 11a solutions, and when it did start to ship in volume, that would drive the 11a market all on its own. So I'm not sure where the Cisco engineers claims on slower time to market came from. I'm not sure that they would have been any quicker buying chips on the open market, because there wasn't really that much of an open market.
Disclaimer: I don't work for Cisco, so this is just my take on what happened, not a report from inside Cisco