|whirlpool|core|slashdot|google linux|freshmeat|traceroute|
|melbourne wireless|ldp|port search|

welcome to the anti-fsck wireless broadband homepage

Contact anti-fsck  

  ICQ #114584073
public gnupg key

   wlan -av

Wireless AP: PIII 500MHz, 256Mb RAM, 20Gb HDD, RealTek 100baseT, Compaq WL200, Enterasys Roamabout, Compaq WL200 Texas Instruments cradle.

Notebook: Compaq Armarda M700 750MHz with 196Mb RAM, 12Gb HDD, Compaq WL110 PCMCIA card.

Access point: Netgear ME 102.

Antennas: homebrew 8dBi omni, 5dBi omni, modded 19dBi Galaxy, infant formula tin, Pringles can, Sony NetMD promotional can (don't steal muzak).

Next project: tba.

LAST UPDATED 25 SEPTEMBER 2003

    what's the rub?

    Soz, lost access to this website for a while. Attempting a 70km wireless link in about a week's time, will provide an update on how it works out.

    13 MAY 2003

    Node BCA is no more

    The mast and antennas came down at the weekend thanks to the efforts of vak and rbritt. The mast was up about six months in total and would have served at least several gigabytes of data (wish I'd kept precise logs of exactly what traffic went via the link, but http traffic alone was about 35Mb-70Mb a day) to seven users up to about 4km away. There was a bit of a postmortem on the gear after it came down that showed a watertight seal isn't necessarily watertight. The homebrew omni fared very well despite showing signs of slight water impregnation.
    rbritt takes to the chimney, vak and rbritt start removing the mounts.
    The omni with the water seal tape partially pulled off, the omni with the n-connector exposed, and the bubble inside the omni's base which indicates water impregnation.
    cable 1, cable 2, cable 3, cable 4.
    There's also pics of vak's cardbaord box AP protector that sat for six months inside the roof crawl space. There's a few signs that water was getting inside the roof, a fair bit of dust, but it did surprisingly well. Couple more raw pix here and here. Soz, these pics here haven't been resized, so they're ~300k each.
    Update: I've resized these pics to here, here, here, here and here. There's also a pic of all the crap used to run the wireless node packed up in the back of the car here.

    11 MARCH 2003
    There's a pic of yesterday's innernorth meeting at node FPU here. Umm, it's a 300k file, haven't had time to resize it. In summary, the meeting decided to encourage the migration to melbwireless-allocated IP addresses by March 17, hit a couple of nodes to get them up and working, to not standardise any vpn protocols, to start and use dhcp, and that's about it.
    Heads in the pic are (from left) vak, (I think this is how you spell it) Jaymie, and Grant.

    1 MARCH 2003
    Node BCA is closing down
    Well, relocating really. We've bought a house that's a fair distance away from where we're now located. The move is on May 29, so there will be a gradual decommissioning of node BCA that will culminate in the dismantling of the mast.
    The move means it is not possible to keep the node part of the Melbourne-based network.
    On the plus side, our new house has very good LOS over Warragul, which means the prospects of setting up a wireless community network to serve the town are very good.

    25 FEBRUARY 2003
    Had some feedback from Craig regarding how he solved the issues with my wireless access point. He writes:
      this isn't actually correct.
      the IDE problem and the WL-200 issue were entirely unrelated.
      the WL-200 problem was a configuration issue. pcmcia-cs needed to be configured to use PCI interrupts rather than ISA bus interrupts. on a debian system, that is done by editing /etc/default/pcmcia so that it has the following line:
      PCIC_OPTS="irq_mode=1"
      the IDE problem is that you have a combination of drive and motherboard IDE controller which cause file system corruption when DMA mode is enabled. the corruption is directly related to disk activity - the heavier the disk activity the sooner and more severe the corruption. a new motherboard would almost certainly fix the problem, or a new drive might fix the problem. replacing both would definitely fix it.
      i didn't fix it, i just worked around it by disabling DMA mode.
      craig

    So I had it pretty wrong. Unfortunately, the budget means I'm stuck with both the harddisk and the motherboard indefinitely.

    9 FEBRUARY 2003
    Well, it's been a while since the last update, mostly because I forgot my password for this account :/ But we're back up and going.
    In summary, the WL200 (hostap_cs) is now up and working under Debian. This is largely thanks to Craig Sanders, who discovered that I had a rare Ultra IDE motherboard that had serious issues under Linux, and was able to work around it by disabling IDE during the boot process. The box the WL200 is in is completely headless, thanks to an Enterasys (orinoco_cs) in the TI cradle. The box is sitting in the front wondow and is connected to the homebrew 8dBi omni via 10m of LMR 400.
    The mast and antennas, the antennas from the side, and the public access point where it sits in the front window. Soz, the pics are pretty big at the moment.
    Aside from the public access AP I also now have a DLink DWL900+ AP (vak's AP, not mine) in the roof with a direct 19dBi connection to node BFF, a distance of 1.7km. He's using a DWL900+ at his end as well, although his has been modified as a mast-mount POE solution. The link works fine in AP-client mode, although there appears to be a software/hardware flaw that is causing excessive packet duplication. One to work on.

    19 OCTOBER 2002
    I've set up a page for today's antenna raising. I'll update it as often as I can thruout the day - with pics - but can't guarantee it will be *that* often.

    28 SEPTEMBER 2002
    Having problems with the WL200 under Debian. If anyone has clue++ about what's going on here, please e-mail me.
    Here is the output from uname, iwconfig and dmesg:
      packetstorm:~# uname -a
      Linux packetstorm 2.4.19 #1 SMP Wed Sep 25 15:12:23 EST 2002 i686 unknown
      

      packetstorm:~# iwconfig wlan0 Warning: Driver for device wlan0 has been compiled with version 13 of Wireless Extension, while this program is using version 12. Some things may be broken... wlan0 IEEE 802.11-b ESSID:"packetstorm" Nickname:"melbwireless_nodeBCA" Mode:Ad-Hoc Frequency:2.457GHz Cell: 02:23:7B:07:33:94 Bit Rate:5.5Mb/s Tx-Power:5 dBm Sensitivity=1/3 Retry min limit:8 RTS thr:off Fragment thr:off Encryption key:off Power Management:off Link Quality:0/92 Signal level:-90 dBm Noise level:-90 dBm Rx invalid nwid:0 Rx invalid crypt:0 Rx invalid frag:0 Tx excessive retries:0 Invalid misc:0 Missed beacon:0 packetstorm:~#

      packetstorm:~# dmesg (snippage of irrelevant stuff) NET4: Linux TCP/IP 1.0 for NET4.0 IP Protocols: ICMP, UDP, TCP, IGMP IP: routing cache hash table of 512 buckets, 4Kbytes TCP: Hash tables configured (established 8192 bind 8192) IPv4 over IPv4 tunneling driver GRE over IPv4 tunneling driver Linux IP multicast router 0.06 plus PIM-SM NET4: Unix domain sockets 1.0/SMP for Linux NET4.0. NET4: Ethernet Bridge 008 for NET4.0 VFS: Mounted root (ext2 filesystem) readonly. Freeing unused kernel memory: 124k freed Adding Swap: 489940k swap-space (priority -1) eth0: Setting 100mbps full-duplex based on auto-negotiated partner ability 45e1. Linux PCMCIA Card Services 3.2.1 kernel build: 2.4.19 #1 SMP Wed Sep 25 15:12:23 EST 2002 options: [pci] [cardbus] [apm] Intel ISA/PCI/CardBus PCIC probe: PCI: Found IRQ 5 for device 00:0a.0 PCI: Sharing IRQ 5 with 00:0e.0 TI 1410 rev 01 PCI-to-CardBus at slot 00:0a, mem 0xdd024000 host opts [0]: [isa irq] [pci irq 5] [lat 64/176] [bus 172/0] ISA irqs (scanned) = none!<6> PCI card interrupts, PCI status changes PCI: Found IRQ 11 for device 00:08.0 Cirrus PD6729 rev 00 PCI-to-PCMCIA at slot 00:08, port 0xa400 host opts [0]: [ring] [1/5/0] [1/20/0] host opts [1]: [ring] [1/5/0] [1/20/0] ISA irqs (default) = 3,4,7,12 polling interval = 1000 ms cs: memory probe 0x0d0000-0x0dffff: clean. hostap_crypt: registered algorithm 'NULL' hostap_cs: hostap_cs.c 0.0.0 2002-09-12 (SSH Communications Security Corp, Jouni Malinen) hostap_cs: (c) Jouni Malinen hostap_cs: setting Vcc=33 (constant) hostap_cs: CS_EVENT_CARD_INSERTION prism2_config() hostap_cs: setting Vcc=33 (from config) Checking CFTABLE_ENTRY 0x01 (default 0x01) Config has no IRQ info, but trying to enable IRQ anyway.. IO window settings: cfg->io.nwin=1 dflt.io.nwin=1 io->flags = 0x0046, io.base=0x0000, len=64 cs: IO port probe 0x0100-0x04ff: excluding 0x290-0x297 0x378-0x37f 0x3c0-0x3df 0x4d0-0x4d7 cs: IO port probe 0x0298-0x0377: clean. cs: IO port probe 0x0380-0x03bf: clean. cs: IO port probe 0x03e0-0x04cf: clean. cs: IO port probe 0x04d8-0x04ff: clean. cs: IO port probe 0x0800-0x08ff: clean. cs: IO port probe 0x0a00-0x0aff: clean. cs: IO port probe 0x0c00-0x0cff: clean. hostap_cs: index 0x01: Vcc 3.3, irq 3, io 0x0100-0x013f hostap_cs: Registered netdevice wlan0 prism2_hw_init() prism2_hw_config: initialized in 9466 iterations wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=0) wlan0: hfa384x_cmd: command was not completed (res=0, entry=c5d9ade0, type=0, cmd=0x0021, param0=0xfd0b) wlan0: hfa384x_get_rid: CMDCODE_ACCESS failed (res=-110, rid=fd0b, len=8) Could not get RID for component NIC wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=0) wlan0: hfa384x_cmd: command was not completed (res=0, entry=c5d9ade0, type=0, cmd=0x0021, param0=0xfd02) wlan0: hfa384x_get_rid: CMDCODE_ACCESS failed (res=-110, rid=fd02, len=8) Could not get RID for component PRI wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8021 wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 wlan0: hfa384x_get_rid: CMDCODE_ACCESS failed (res=-110, rid=fd20, len=8) Could not get RID for component STA wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8021 wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 wlan0: hfa384x_get_rid: CMDCODE_ACCESS failed (res=-110, rid=fd20, len=8) wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8021 wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 wlan0: hfa384x_get_rid: CMDCODE_ACCESS failed (res=-110, rid=fc01, len=6) wlan0: could not get own MAC address wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8000 wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 wlan0: hfa384x_set_rid: CMDCODE_ACCESS_WRITE failed (res=-110, rid=fc00, len=2) wlan0: Port type setting to 6 failed wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8000 wlan0: hfa384x_cmd: entry still in list? (entry=c5d9ade0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 wlan0: MAC port 0 enabling failed wlan0: prism2_open prism2: wlan0: operating mode changed 3 -> 1 wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8000 wlan0: hfa384x_cmd: entry still in list? (entry=c4e661e0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 wlan0: hfa384x_set_rid: CMDCODE_ACCESS_WRITE failed (res=-110, rid=fc00, len=2) wlan0: prism2_close prism2_detach hostap_cs: detach postponed, 'wlan0' still locked prism2_release wlan0: hfa384x_cmd_issue: cmd reg was busy for 1000 usec wlan0: hfa384x_cmd_issue - timeout - reg=0x8000 wlan0: hfa384x_cmd: entry still in list? (entry=c3f9a5e0, type=0, res=-1) wlan0: hfa384x_cmd: interrupted; err=-110 hostap_cs: Shutdown failed release - done hostap_cs: Netdevice wlan0 unregistered hostap_cs: Driver unloaded hostap_crypt: unregistered algorithm 'NULL' (deinit) Trying to free nonexistent resource <0000a400-0000a401> unloading PCMCIA Card Services Linux PCMCIA Card Services 3.2.1 kernel build: 2.4.19 #1 SMP Wed Sep 25 15:12:23 EST 2002 options: [pci] [cardbus] [apm] Intel ISA/PCI/CardBus PCIC probe: PCI: Found IRQ 5 for device 00:0a.0 PCI: Sharing IRQ 5 with 00:0e.0 TI 1410 rev 01 PCI-to-CardBus at slot 00:0a, mem 0xdd024000 host opts [0]: [isa irq] [pci irq 5] [lat 64/176] [bus 172/0] ISA irqs (scanned) = none!<6> PCI card interrupts, PCI status changes PCI: Found IRQ 11 for device 00:08.0 Cirrus PD6729 rev 00 PCI-to-PCMCIA at slot 00:08, port 0xa400 host opts [0]: [ring] [pci irq 11] [1/5/0] [1/20/0] host opts [1]: [ring] [pci irq 11] [1/5/0] [1/20/0] PCI card interrupts, PCI status changes cs: memory probe 0x0d0000-0x0dffff: clean. hostap_crypt: registered algorithm 'NULL' hostap_cs: hostap_cs.c 0.0.0 2002-09-12 (SSH Communications Security Corp, Jouni Malinen) hostap_cs: (c) Jouni Malinen hostap_cs: setting Vcc=33 (constant) hostap_cs: CS_EVENT_CARD_INSERTION prism2_config() hostap_cs: setting Vcc=33 (from config) Checking CFTABLE_ENTRY 0x01 (default 0x01) Config has no IRQ info, but trying to enable IRQ anyway.. IO window settings: cfg->io.nwin=1 dflt.io.nwin=1 io->flags = 0x0046, io.base=0x0000, len=64 cs: IO port probe 0x0100-0x04ff: excluding 0x290-0x297 0x378-0x37f 0x3c0-0x3df 0x4d0-0x4d7 cs: IO port probe 0x0298-0x0377: clean. cs: IO port probe 0x0380-0x03bf: clean. cs: IO port probe 0x03e0-0x04cf: clean. cs: IO port probe 0x04d8-0x04ff: clean. cs: IO port probe 0x0800-0x08ff: clean. cs: IO port probe 0x0a00-0x0aff: clean. cs: IO port probe 0x0c00-0x0cff: clean. hostap_cs: index 0x01: Vcc 3.3, irq 11, io 0x0100-0x013f hostap_cs: Registered netdevice wlan0 prism2_hw_init() prism2_hw_config: initialized in 9468 iterations wlan0: NIC: id=0x8002 v1.0.1 wlan0: PRI: id=0x15 v0.3.0 wlan0: STA: id=0x1f v0.8.0 Using data::nullfunc ACK workaround - firmware upgrade recommended wlan0: LinkStatus=2 (Disconnected) wlan0: prism2_open prism2: wlan0: operating mode changed 3 -> 1 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=00:50:8b:d0:33:94 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=44:44:44:44:44:44 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=44:44:44:44:44:44 wlan0: prism2_close wlan0: prism2_open wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=02:23:d1:06:33:94 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=44:44:44:44:44:44 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=02:23:f7:5e:33:94 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=02:23:cd:4a:33:94 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=02:23:69:57:33:94 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=44:44:44:44:44:44 wlan0: LinkStatus=2 (Disconnected) wlan0: LinkStatus: BSSID=44:44:44:44:44:44 packetstorm:~#

    22 SEPTEMBER 2002
    Been a while since the last update.
    The gateway had a visit to the last melbwireless meeting to see if anyone could have a play with it and fix a few outstanding issues. Well, the end result is that it is still sporadically segfaulting, although dhcp now works properly. Next job is to scrub Red Hat and replace it with Debian to see if that improves the situation.
    Holding an antenna installation day on October 12. I really need to get this damn thing up, so it's free barbecue and beer until the job is done. In the meantime, the pressure is on to get the gateway issues resolved.

    26 AUGUST
    This is an ex-Galaxy eaves mount mast.

    18 AUGUST 2002
    The wireless gateway went live yesterday for a test. No one in my immediate area said they were able to connect, which is a bit of a disappointment. Vak, who is about 800m away from me, said he attempted a connection but was unable to see the node. However, vak and I were later able to establish a connection over about 700m from a local park that, as far as we could tell, had no direct LOS to the antenna. The connection's signal strength was good, and the pings <10ms. Not bad for a homebrew omni. We also played around with how much tolerance the connection had from a badly-aimed directional antenna. Over a distance of about 150m, the antenna could turn more than 30 degrees from the omni and still receive a shitty signal. We were also surprised to see that you could "bounce" the signal off nearby buildings to bend it through a 90-degree arc. Polarisation also is important - turning the antenna 90 degrees from vertical lost about half the signal.
    I think I learnt at least one valuable lesson from yesterday's test: height is king. The higher you can get the antenna, the more chance you have of being seen. With that in mind, let's go shopping for a 6m mast.
    Soz, no pix. The AP was up for about 10 hours in total. Might try again another day.

    16 AUGUST 2002
    A simple lesson in why cable modems and babies *don't* mix. If you have a sensitive stomach, don't click here. I'd moved the modem from the middle of it all by the time I took the photo. It still works fine, just smells a bit.
    Update: Saffie is still spewing (10.30pm). Also, forgot to upload a pic of the ultimate in wireless mobility. Hey, I needed something to carry all the stuff.

    14 AUGUST 2002
    It is done. The WL200 is in the PIII 500MHz running in master mode. This means I have an access point working under Linux. This site was a big help.
    Running two separate APs on channels three and five doesn't work. So they are now five channels apart. Hopefully I'll be setting up a samba share later this week and then testing it with the 8dBi omni. I can also wardrive with the 5dBi omni to see what reach I can get out of it.
    Anyone from the innernorth mailing list can look for SSID MelbWireless_NodeBCA operating on channel one. I'll set up DHCP and put a couple of isos on the samba share for download testing. Depending on time I'll also set up a nice-looking html page to serve rather than the default apache install page.
      [root@localhost root]# uname -a
      Linux localhost.localdomain 2.4.18-3 #2 Tue Aug 13 22:48:21 EST 2002 i686 unknown
      

      [root@localhost root]# iwconfig wlan0 wlan0 IEEE 802.11-DS ESSID:"MelbWireless_NodeBCA" Nickname:"localhost.localdomain" Mode:Master Frequency:2.412GHz Access Point: 00:50:8B:D0:33:94 Bit Rate:2Mb/s Tx-Power:7 dBm Sensitivity=1/3 Retry min limit:8 RTS thr:off Fragment thr:off Encryption key:off Power Management:off Link Quality:0 Signal level:0 Noise level:0 Rx invalid nwid:0 Rx invalid crypt:15 Rx invalid frag:0 Tx excessive retries:1301 Invalid misc:1221 Missed beacon:0 [root@localhost root]#

    27 JULY 2002
    Tried to upload a couple of pics of the omni mounted on the roof, but bur.st's ftp server appears to be having some sort of crisis. Will try again soonish.
    Update: Done it. Here's one pic and another from a different angle.
    Trying to build the wireless gateway under FreeBSD but running into problems. It's strange, as it will see the ME 102, but won't pass any packets. More on that later.

    17 JULY 2002
    Vak has tested the LMR-400 omni. It gets 8dBi with three degrees at 3dB. That means a spread of +/-70m at 1000m. w00t!
    Now to work how to mount the sucker.

    12 JULY 2002
    Here's the first pic of the LMR-400 ghetto omni [here's a big pic if anyone wants a closer look], professionally hand-modelled by vak, who has done most of the work on it. Because LMR has a different velocity to RG-213 the segment lengths had to be re-calculated. In short, using LMR instead of RG means the antenna will end up being a little bit longer - but hopefully better gain. This one be 11 segments long, so about 8dBi we hope.

    07 JULY 2002
    Warchalking has hit the streets of Melbourne's CBD.

    29 JUNE 2002
    Mkay, few quick things. I've added a text version of a Netstumbler log of a wardrive around Melbourne's CBD. It's basically sweeping the CBD between La Trobe Street and Flinders Lane using a 5dBi omni. Secondly, I'm going to attempt to build a ghetto 8dBi omni using a length of LMR 400 (not RG-213 even though it looks much easier to work with).

    11 JUNE 2002
    Just a brief note to say nothing has been happening for a while. Wombat of Wrath has been working on the wireless gateway's kludgy software. Still have three more galaxy antennas to get, saw an easy-to-get 24dBi antenna that I will have to suss, time to think about some wireless security. More about that later.

    10 MAY 2002

    modding a pacific monolithics antenna

    Preamble: Thanks to Dad for helping me with this. You might notice some of the pics are a bit fuzzy. Yeah, yeah, the camera was too close.

    OK, we've done how to open the Pacific Monolithics dipole (scroll down further you n00b). Now we have to attach a pigtail.

    Once you have the dipole out of the housing, take off the bit we need. I have found it best to cut off the dipole just past where the mount holds the dipole housing onto the arm of the antenna. This allows you to use the screw to refit the modded dipole.

    Once you have cut it, you need to break the circuit track to isolate the dipoles. This is because a bit of the circuitry still is attached to the dipoles, and as I'm not sure what it dies to the antenna's performance if left attached, I cut the contact with the electronics. I just did this using a box cutter, cutting two parallel lines across the board and digging away the silver circuit until I could see the green of the circuit board.

    Because the RG-58 that I'm using for the mod is a little bit thick to slide in beside the mounting screw, I needed to modify the dipole housing a bit to make room for the cable. I had to cut away about a third of the plastic housing that the mount fits into and remove a couple of the plastic stays that held the moisture-absorbing gel capsule. You can also see here that I made a hole at the base of the dipole housing to bring the RG-58 out the back of the dish.

    Okay, we're ready to solder in the RG-58. The RG-58 already has a DSE RG-58 to n-type silver connector attached to it. If you look at the dipole, there are two runs of silver circuit running east and west. In between is a small hole that leads to, when you turn the dipole over, a small square. We need to feed the core of the RG-58 through the hole on the opposite side to the square and solder the core onto the square. The next step is laying the outer braid on the left-hand track (if you looked carefully before you would notice that the square that you soldered the centre core to is connected to the right-hand track) and soldering it down Once you're finished, the assembled dipole with the RG-58 should look something like this.

    Mkay, now's the messy bit. Fill both halves of the dipole with Silastic. Place the dipole back in the dipole housing, and close the case back up as tight as you can using cable ties or some other means. If you want, jam the silastic gun in the hole where the f-connector used to be and pump some more of the stuff in until it oozes out the joins. Let it cure for 24 hours and then test the modded Galaxy for gain.

    The end.

    27 APRIL 2002
    Well, my dad came to visit today, and he has much better solder tekniq than I. So I put him to work on an old project: the milk tin passive repeater. The premise of the design is two milk tin cantennas connected with a sort length of LMR-400 low-loss cable.

    Does it work?

    B00ty. It works. Sorta. It's more of a concept than anything else, but it appears to work. For people in a really bad LOS jam, it is an option. The repeater's gain was measured against the orinoco alone. The basis was orinoco-->pigtail-->pringles can-->milk tin-->LMR-400-->milk tin-->access point. Testing distance was about 20m and the signal was bent through a right angle to make sure we were seeing a repeated signal. One milk tin had a 12dB gain over the orinoco, another had an 8dB gain over the orinoco (its core was deliberately oversize to see what it did to the can's performance. In summary, the longer core sucked badly). The pringles can had a 10dB gain over the orinoco.
    Anyway, what we saw was a 6dB gain over the orinoco once the signal was channelled through the passive repeater. Signal loss should have been about, say, one for the cable and two for the n-connectors either end of the passive repeater, so the end result was not that bad. I really need to do a distance test to back this up, preferably with a couple of modded Galaxy antennas. If anyone can help with this, send me an e-mail.

    2 APRIL 2002
    wardriving kit Well, it was a nice, sunny day so I thought I'd put the Galaxy antenna on the roof and see what I could pick up when I pointed it at the city. First I had to cut away a vine that had this irritating sticky white sappy shit that oozed all over me. Then I broke three roof tiles walking across the roof (you can see some of the vine in this photo). That aside, I finally set up the Galaxy antenna to point at the CBD and laid out the deckchair city h4x0r's access point with more of those damn vines and seed pods in the background.

    So what did I see? Netstumbler showed fuck all. Nothing. Nada. Oh well, at least I learnt how to walk on tiled rooftops without leaving a trail of destruction behind me.

    There's a couple of pix on the shafted.com.au website of the Galaxy antenna doing its thing at Shafted 20.

    31 MARCH 2002
    OK, pictures from Shafted 20. I always regret not taking enough photos. This instance is a case of that.
    I meant to take a picture of the line of sight to the AP but forgot.

    So how did the AP hold up? Chooken ran a Counter-Strike server (hopefully I have a screenshot on the PC somewhere) on a wireless notebook. Pings were a little worse than on the wired LAN - about 20ms compared to 5ms to 10ms on the wires. Unfortunately pings blew out to between 200ms and 300ms once anyone started to download files via a wireless link. On the bright side, anyone downloading could get a sustained 200kbps peaking at 300kbps.

    29 MARCH 2002
    Here is the 3dBi omni antenna and Cisco Aironet access point (lower left) at Shafted 20. Should have a pic of the cantenna in the thick of combat soon.
    27 March 2002
    Added the Sony NetMD cantenna to the fold today. It's not quite as good as the Pringles can, but the removable base meant I didn't have to find someone with small hands to put the screws in for me. The tin is part of a promo for the .au version of NetMD.
    Also completed modding the 19dBi Galaxy antenna. Took it for a test ping and saw a 22dBm gain compared to the Orinoco card alone. The mod itself is pretty easy, but for the frigging base of the downconverter, which snapped when I was trying to ream it out to fit the width of the LMR-400 (it's made from very soft magnesium). Replaced it with a mod by Julian Featherstone, which has worked well. Oh yer, I must take a pic at some stage.

    15 MARCH 2002
    The great cantenna shoot-off

    Okay, after much drilling, filing and soldering, making up of infant formula and consumption of dehydrated potato-flavored snackfoods, it was time to test the result.

    Running up against each other were:

    I'll take pictures and put them up over the weekend.

    The testing was over a distance of about 20m to a Netgear ME 102 access point. The Orinoco's gain was used as the benchmark for the test. Signals were measured using Netstumbler. The Netstumbler log is here if anyone is interested.

    Devicesignal (dBm)noise (dBm)
    Orinoco silver-78-92
    Pringles can-68-95
    Milk tin-66-97
    Galaxy antenna-57-95

    So what's it all mean? I may be wrong in my interpretation (so correct me if I draw the wrong assumptions), but it looks as though the Pringles can had about a 10dBm gain over the Orinoco's antenna alone. It had more noise than the Orinoco, but held up quite well.

    The milk tin was about on par with the Pringles can, but had more noise. I was a bit disappointed with the result, but bear in mind that I bjorked the calculations and was about 4mm out in the location of the n-type connector's distance from the base of the tin.

    The Galaxy ruxed.

    On a closing note, had a bit of a wardrive just up Lonsdale Street while passing through the city yesterday. I had the lappy in the car with no external antenna plugged in. I picked up 15 access points, of which only three had WEP enabled.

    Feedback
    evilbunny writes: ummm your assumptions are wrong... The noise is in fact less not more...
    He adds: -92 is greater then -97 :) basically the bigger the spread between them the better it should be... usually the noise floor sits about -99, -100 for me, lots of ovens or other WLAN's where you are? -92 for the card itself is kinda bad... i've had it as low as -104... as for signal you want it to go the other way, -1 would be brilliant :)
    Perhaps I should have mentioned that I had no LOS to the access point for the test -- it had to go through several walls of the house including a window in an external brick wall.

    19 FEBRUARY 2002
    Hacked the infant formula tin for 2.4GHz. Tin of choice is S-26 Progress Gold infant formula (from 6 months) which, funnily enough, is made in Ireland. Never noticed that before. I also have a couple of S-26 infant formula (from birth) tins to mod if this experiment proves worthwhile.

    Warning: The images are raw from the camera, so I think they're pretty large in size. Yep, about 300k each.

    3 FEBRUARY 2002
    Modding an ex-Galaxy pacific monolithics antenna is sure to bring many years of joy and excitement at a very low cost. But the hardest step is getting the goddamn thing out so you can play with it. Here's how.

    Equipment:
    Snub-nose pliers
    Retractible box cutter with a locking blade
    Largish knife
    Hammer
    Block of wood
    Band-Aids
    Use the snub-nose pliers to unscrew the coax plug from the housing. Don't worry about burring the thread, as you just chuck it in the bin straight away. I found the easiest method was to trace around the welded join of the housing with an al Queda special (otherwise known as a retractible box cutter, but one that allows you to lock the blade) several times, and then use a hammer to tap a knife in the groove starting at the narrowest end. You're best to do this with the dipole resting on a block of wood, as you need to apply a bit of pressure to make an impression. Work around both sides of the dipole, leaving the leading edge of the dipole housing - the end that points to where the signal will be going - intact. This gives the rest of the housing enough flexibility to allow you to pull the housing at the opposite end of the dipole apart and pull the down converter out. That way you don't need to split the whole case open. Pictures to come.

    Next: Modding the dipole without a cluestick.

    31 JANUARY 2002
    Okay, have some pictures here of the Galaxy parabolic dish that I'm attempting to hack for 2.4GHz. So far, so good although I need to cut the plastic housing away so that I can cut the down convertor off the dipole and attach some cable. It's pretty sharp, pointy and dangerous stuff.
    Anyway, here's the dish and here's the plastic dipole housing if anyone is interested in looking at them. I'll take more pictures as things (slowly) progress. Both images are 640 pixels wide.

    13 JANUARY 2002
    Okay, it's crunch time. There's going to be a bit of a physical reshuffle in the household, thanks to the twins. The computers are all going to have to move into the smaller room next to where they all sit now. The difficulty is, how am I going to keep the cable modem, stuck on the end of about 1.5m of coax, connected to the network? I could do it wirelessly, but the twins will be in this room so SWMBO is not warming to that idea. It may come down to running some cat5 under the house and a couple of wall plates.

    On a different slant, I'm going to have to make a conscious decision as to what I want to use to serve the AP. I might give the P133 one more shot this weekend, otherwise I'm going to have to try and build it into the Netfinity and run a longer length of cable to the antenna. I've given up on the PowerBase as I don't have the original install disks adn the hardware is not very Apple OS friendly. Send it up to the in-laws' place and they can use it for surfing the net and stuff.

    The hunt for an ex-Galaxy antenna is going poorly. Everyone is either renting, can't speak English, or not home. I also had a bit of confusion with Optus' antennas that they now use to supply pay TV services in areas where the cable does not run (I think they're using MMDS over about 1.9GHz?). It's about half the size again of the Galaxy dish.

    I might do a photojournal of my efforts to get the P133 working and slab it in here.

    Pic of the wardriving kit has been blatantly ripped with no attribution whatsoever from the fine folk at Computerworld. Looks sucky with the border, but tough, it ain't your page.

    19 DECEMBER 2001
    Well, I installed Mandrake 8.1 and was quite impressed on two fronts: the stock-standard install recognised the wireless pcmcia card and the damn thing beeped as it was recognised on boot. Promising stuff. However, Mandrake uses wvlan_cs which - as I found out - causes a few problems with certain hardware configurations, namely a constant stream of Tx errors as the system fails to pick up the pcmcia card properly. I also have to boot each time from the boot floppy otherwise I get a stream of "40 40 40 40" spewing across the screen. Again, this appears to be related to the system failing to see the wireless card properly than an install problem.
    The next step is to replace wvlan_cs with orinoco_cs. A quick Google shows this is what appears to have solved the problem on a number of installs, so I'm optimistic it will do the same for me.

    13 DECEMBER 2001
    Hrm, I'm beginning to think my card is not compatible with the board that I have in the P133. Tried the zooma test (inserting the pcmcia card while the box3n boots) and I didn't get the beep that should indicate the card was recognised. (I should add that I don't know if it's a problem with the PCI bridge.) I might just throw the card in the XP box3n this weekend and get it up and running for the sake of just getting it up and running.
    So what does all this mean? Well, first of all, it looks as though I can only get this card working under Windows 2000. Which means I can't use the P133 as an access point. Which means I will have to get a box3n that will run Win2k. Which means I'm going to have to look at a PIII 400MHz with at least 128Mb of RAM. *Sigh*. Starting to think it may have been easier just to invest in a prepackaged AP than try and build one. What I lose is stability and the low-cost option by using Linux as the platform.

    Okay, just some introductory basics for what I'm trying to do here. I have two boxes that I'm trying to set up as a wireless access point to provide wireless coverage within about a 3km or so range of the house. Whichever box - the AST or the PowerComputing - gets set up first will host an omnidirectional AP. The second box will be used as a long-haul bridge to some other part of the Melbourne Wireless network that is slowly gaining a foothold.

    So far I'm toying around with a single card to see what I can get going. I have a Compaq WL210, which is a rebadged Orinoco chipset (Compaq WL110 pcmcia card) housed in a Texas Instruments PCI-1410 PCI bridge. It's only 30mW compared to the WL200's 100mW, but is an ideal starting point and, most importantly, has a female jack for an external antenna.

    At this stage, I'm hoping to set up the PPC clone as the AP. Yellowdog Linux has very good support for the Orinoco chipset, which is used extensively in Apple's Airport wireless APs (now with AOL support, schwiiiiiing!), so I'm hoping it's going to be a pretty seamless install once I work out how to see the CD-ROM drive.