Re: [ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
Hi, Yes you can do this. The trick is changing the antenna config for each frame you send out. If you do it with an rs232 port then you have to only hand the hardware one packet at a time. If you hand it multiple packets to different nodes, you won't be able to do per-destination antenna configuration. Yes, I wish there were some NICs that just exposed the antenna configuration bits, RX_CLEAR line and GPIO pins. That would make this particular thing easier to do. adrian On 14 May 2013 08:38, Mofolo Mofolo mofol...@gmail.com wrote: Hi Adrian, Thanks for the reply. I am realize there might be hardware constraint for me to use sectorised antenna. I currently have a NIC with only one antenna exposed . I have reconsidered my scenario and now I am planning to use switched beam antennas (specifically ESPAR antenna). This antenna technology has only one feed line similar to the omnidirectional antenna. In this case, I don't have to worry about modifying HAL code. ESPAR antenna can be configured as either omnidirectional or directional. Directional beams can be achieved by electronically controlling the parasitic elements, for instance by sending a byte value via a serial port on a per packet basis. If I can rephrase the idea, this is what I would like to explore. I want to create a scenario where upon start up, each node uses omnidirectional antenna. Then, once a neigboring node requests association, i want to perform a scan using each of the 4 antenna beams and based on the beam which gives me max RSSI, I assume it is the best sector/beam to be used to communicate with that neighbor. So an extra parameter can be introduced in the neighbor's table to flag the antenna sector/sector which has to be used whenever a node communicates with that specific neighbor. I am assuming the all nodes receive using omnidirectional antenna configuration but transmit (per packet level) using directional beam configuration. It would be more interesting if such a routine is automatically performed within the driver for every station that is added into the neighbor's table. For simplicity I am assuming static network topology. The directional beam configuration is achieved by sending appropriate byte value via serial port (RS232) to control and direct the antenna radiation pattern towards desired direction. I have already achieved this by modifying MADAWIFI. I have to try it using mac80211 and ath5k driver. The only challenge is achieving neighbor discovery especially in ad-hoc mode. With thanks, Mofolo On 13 May 2013 19:29, Adrian Chadd adr...@freebsd.org wrote: Hi, I figure I'll give a 30 second introduction into sectored antenna support. So the AR5212 era chips have four antenna control bits. They're ANTA - ANTD. Now, there's some switch table registers. By default the HAL code sets these up so BB_ANTENNA_CONTROL, BB_SWITCH_TABLE1 and BB_SWITCH_TABLE2 map to idle, stuff with antenna 1 and stuff with antenna 2. This is for non-sectored operation. I can go into more detail here if you'd like. For sectored operation, you don't leave ANTA:ANTD up to the switch table. You control the TX antenna configuration and you use the single omni antenna for receiving frames (normal RX, RTS, listening for ACK too I guess.) There's four antenna config bits in the TX descriptor for each rate attempt, along with DEFANT for the omni antenna configuration. In this mode (sector AP mode) its up to the driver and wifi stack to track the per-client, per-sector behaviour and choose an optimal sector configuration for that. The 11n chips have something but not quite similar for this kind of antenna selection. So, if you have a modified NIC that exposes ANTA-ANTD and you can make up a sectored antenna configuration + omni antenna configuration, I can attempt to help you. I haven't done this myself; I'm just going on what I have here. Thanks, Adrian ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel
Re: [ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
.. in fact, what I'd really _really_ like is if someone would build an open source phased antenna array that's intended to be used by three-stream (ie, three transmit/receive chain) antenna devices. That way we could do both phased array station operation (when it directs its transmit power _to_ a specific location _AND_ leave the default receive antenna configured to the best receive antenna configuration) and phased array AP operation (when it directs its transmit power to specific antenna configurations, one per associated station; with a default omni receive configuration.) If someone's happy to design, implement and open source that, I could have my arm twisted to hack on the basic rate control and driver code to make that stuff work. For ath5k NICs you only have 4 bits of antenna control. For ath9k NICs you have 24 bits of antenna control shifted out two GPIO pins. Adrian ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel
Re: [ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
Hi Adrian, Yes, I am assuming a simplified unicast scenario (peer to peer communication). In that case, one packet can be handled by the hardware. Upon completion of directional transmit process, the antenna is always kept at omnidirectional configuration.On the other hand, cases like broadcast/multicast can be handled with setting the antenna to omnidirectional mode. I am not sure how practical or realistic is my assumption. For the neighbor discovery, I was only able to trace the function calls involved for the antenna to be added into the neighbor's table. The challenge I have is to force a specific scan within the driver once a new station is associated. The scan should be done using all available/possible directional beam configurations. The proposed antenna design (open source phased antenna array that's intended to be used by three-stream antenna devices) sounds interesting. I will discuss it with some guys working on antenna design With thanks, Mofolo On 14 May 2013 18:10, Adrian Chadd adr...@freebsd.org wrote: .. in fact, what I'd really _really_ like is if someone would build an open source phased antenna array that's intended to be used by three-stream (ie, three transmit/receive chain) antenna devices. That way we could do both phased array station operation (when it directs its transmit power _to_ a specific location _AND_ leave the default receive antenna configured to the best receive antenna configuration) and phased array AP operation (when it directs its transmit power to specific antenna configurations, one per associated station; with a default omni receive configuration.) If someone's happy to design, implement and open source that, I could have my arm twisted to hack on the basic rate control and driver code to make that stuff work. For ath5k NICs you only have 4 bits of antenna control. For ath9k NICs you have 24 bits of antenna control shifted out two GPIO pins. Adrian ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel
Re: [ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
Well, I can help you with the hardware tweaks for sectored antennas. But first you need to tell me how you plan on hooking up said sectored antenna to the ath5k NIC. Adrian On 7 May 2013 00:27, Holger Schurig holgerschu...@gmail.com wrote: You're probably 5 years too late. The MADWIFI driver is quite out of fashion nowadays and few people are still hacking on it. I wouldn't base any new development on it. What you describe is a bit like diversity: receive with all antennas, determine which antenna had the highest RSSI, use that antenna for transmits towards the client. For this you'd need to maintain your own mac-hashed neightborhood table in the driver. However, to my best knowledge the hardware doesn't support 4 antennas. You'd need some circuitry to use only some or all of those antenna (all antennas in the beacon case). And that for the tx and rx case. As the chips usually only suport 2 antennas, you'd need some koax switching circuitry, and sync that to the transmits. Things get ugly and complex pretty fast. In the end it might be cheaper to just use one AP with one sectional antenna ... and simply put 3 or 4 of those APs into one casing. :-) 2013/5/6 Mofolo Mofolo mofol...@gmail.com Hi, I am integrating sectorised antenna with a Wi-Fi node, basically a PC running Ubuntu. The network network is setup on ad-hoc mode. Currently installed driver is MADWIFI Please advise on how can a node perform neighbor discovery while using sectorised antennas. That means for every station in a neighbor's table, an appropriate antenna sector has to be recorded so as for every packet transmitted, the node can select and use suitable antenna sector. For example: If we say each node is equipped with an antenna with 4 sectors, then the node has to determine which sector to use when communicating with neighbors. in the case where nodes A, B, C and D are aligned in a straight line; node 'B' has to use a different antenna sector when communicating with node 'A', as opposed to when communicating to node 'C'. A B C D How can this mapping of neigbouring nodes with specific antenna sectors be achieved, if maybe RSSI is used as the determining factor during building up of neighbor's table? Thanks. ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel
Re: [ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
Hi, I figure I'll give a 30 second introduction into sectored antenna support. So the AR5212 era chips have four antenna control bits. They're ANTA - ANTD. Now, there's some switch table registers. By default the HAL code sets these up so BB_ANTENNA_CONTROL, BB_SWITCH_TABLE1 and BB_SWITCH_TABLE2 map to idle, stuff with antenna 1 and stuff with antenna 2. This is for non-sectored operation. I can go into more detail here if you'd like. For sectored operation, you don't leave ANTA:ANTD up to the switch table. You control the TX antenna configuration and you use the single omni antenna for receiving frames (normal RX, RTS, listening for ACK too I guess.) There's four antenna config bits in the TX descriptor for each rate attempt, along with DEFANT for the omni antenna configuration. In this mode (sector AP mode) its up to the driver and wifi stack to track the per-client, per-sector behaviour and choose an optimal sector configuration for that. The 11n chips have something but not quite similar for this kind of antenna selection. So, if you have a modified NIC that exposes ANTA-ANTD and you can make up a sectored antenna configuration + omni antenna configuration, I can attempt to help you. I haven't done this myself; I'm just going on what I have here. Thanks, Adrian ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel
Re: [ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
You're probably 5 years too late. The MADWIFI driver is quite out of fashion nowadays and few people are still hacking on it. I wouldn't base any new development on it. What you describe is a bit like diversity: receive with all antennas, determine which antenna had the highest RSSI, use that antenna for transmits towards the client. For this you'd need to maintain your own mac-hashed neightborhood table in the driver. However, to my best knowledge the hardware doesn't support 4 antennas. You'd need some circuitry to use only some or all of those antenna (all antennas in the beacon case). And that for the tx and rx case. As the chips usually only suport 2 antennas, you'd need some koax switching circuitry, and sync that to the transmits. Things get ugly and complex pretty fast. In the end it might be cheaper to just use one AP with one sectional antenna ... and simply put 3 or 4 of those APs into one casing. :-) 2013/5/6 Mofolo Mofolo mofol...@gmail.com Hi, I am integrating sectorised antenna with a Wi-Fi node, basically a PC running Ubuntu. The network network is setup on ad-hoc mode. Currently installed driver is MADWIFI Please advise on how can a node perform neighbor discovery while using sectorised antennas. That means for every station in a neighbor's table, an appropriate antenna sector has to be recorded so as for every packet transmitted, the node can select and use suitable antenna sector. For example: If we say each node is equipped with an antenna with 4 sectors, then the node has to determine which sector to use when communicating with neighbors. in the case where nodes A, B, C and D are aligned in a straight line; node 'B' has to use a different antenna sector when communicating with node 'A', as opposed to when communicating to node 'C'. A B C D How can this mapping of neigbouring nodes with specific antenna sectors be achieved, if maybe RSSI is used as the determining factor during building up of neighbor's table? Thanks. ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel
[ath5k-devel] Integrating sectorised antenna with a Wi-Fi node
Hi, I am integrating sectorised antenna with a Wi-Fi node, basically a PC running Ubuntu. The network network is setup on ad-hoc mode. Currently installed driver is MADWIFI Please advise on how can a node perform neighbor discovery while using sectorised antennas. That means for every station in a neighbor's table, an appropriate antenna sector has to be recorded so as for every packet transmitted, the node can select and use suitable antenna sector. For example: If we say each node is equipped with an antenna with 4 sectors, then the node has to determine which sector to use when communicating with neighbors. in the case where nodes A, B, C and D are aligned in a straight line; node 'B' has to use a different antenna sector when communicating with node 'A', as opposed to when communicating to node 'C'. A B C D How can this mapping of neigbouring nodes with specific antenna sectors be achieved, if maybe RSSI is used as the determining factor during building up of neighbor's table? Thanks. ___ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel