Forum

  • Have I got a fun one for you guys...

    I?¡é?€??ve got a customer whom I suspect is experiencing severe multipath interference and/or some type of EMI interference from large machinery. They are a metal machine shop with large drills, saws, and planers with overhead cranes moving back and forth to transport the metal from one location in the plant to the other. Oh and did I mention the walls were all sheet metal too? No plasma cutters or microwave type heating machines luckily...

    They are running 1242 IOS access points using 12.3.8 code, using primarily the AIR-ANT3213 pillar mount antennas. They do have one area with two 1728 antennas mounted correctly. The clients are desktop PC?¡é?€??s using a mix of Linksys WRT54G card/antennas (not diversity) and Cisco PI21ABG cards/antennas. A passive survey was done previously to determine coverage and placement and coverage seems to be fine. In the logs I see a lot of clients dis-associating due to Max Retries Limits reached. The radio stats indicate very high numbers of CRC errors and high retries (hundreds every 5 minutes). Voice is not in use here. Ping tests show a lost packet every 20 or so tries. They are running an ERP application where if it loses enough data over time it shuts down. Email, web and simple file sharing seem to be fine.

    I?¡é?€??ve run Air Magnet Laptop Analyzer and found higher signal to noise ratios than I would like (-50 to -60). The only thing it seems to notice is the high CRC errors and retries. The signal levels are constantly bouncing from -50?¡é?€??s to -60?¡é?€??s and are never steady. The signal distribution levels are bouncing all over as well, which from a quick look at Air Magnet?¡é?€??s knowledge base indicated possible RF or multipath interference. There is a high level of management traffic at the 1mbps data rate, but I suspect that is from either the script they are running to constantly ping the ERP server, or the frequent re-associations. I do not see any co-channel interference.

    So far to combat this I?¡é?€??ve had the customer move the pillar mount antennas off of the metal I-Beams which Cisco recommends. I?¡é?€??ve experimented with power settings and found that I could turn it WAY down (was at 25mW) to 1mW which actually still provides plenty of coverage shockingly. Turned on Long Preamble and increased the Max Retries thresholds. Once I did this, there was some decrease in the CRC errors and retries, but I?¡é?€??m still not comfortable with it.

    Does anyone else have any ideas to combat this? I?¡é?€??m considering trying a lower gain antenna, but reaching the antennas is a bit difficult without a lift. I?¡é?€??m even considering trying the 5ghz band as I believe it is a little less susceptible to interference. Not sure if the customer is willing to swallow that pill though.

    Thanks in advance?¡é?€?|

    Jake

  • Try using a spectrum analyzer like Cognio's. The retransmissions are more likely to be from EMI than Multipath given correct antenna use and placement. If EMI is the cause, try switching to another PHY, 5GhZ. If you can not do that try decreasing the fragmentation threshold and or moving the antennas.

  • Thanks for the info. I'm trying to get my hands on a spectrum analyzer, but not having much luck.

    Why decrease the fragmentation threshold, just curious? I've never heard that before.

    Thanks for the response!

    Jake

  • Fragments are smaller than the entire frame. Thus they are exposed to the EMI for a shorter period of time and are less likely to suffer from collisions with the EMI. Fragmentation does add more overhead, but can often be a lesser evil than the fixed source of non 802.11 noise. Changing your fragmentation threshold is free. Other solutions are not.

    If you want a good spec. device go with the Cognio ($3,500.00). If you just want to find the noise try WiSpy USB ($100.00). It is not as complete a scan but will find the noise.

    Good Luck and keep us posted.

  • Thanks for the explanation. I did change the threshold to 1024 and I'm watching the CRC's now. It's a bit lower, but the machines aren't running at this time. Once they get off break I'll monitor further.

    Thanks for the input!

    Jake

  • If you are going to experiment with fragmentation, then you will want to use a lower threshold than 1024. Knock it down to 300 or so and let us know. As Bryan said, it will have more overhead, but by testing it you won't lose service so there are no worries there.

    Gene

Page 1 of 1
  • 1