We finally got our 900mhz 450i put up... so far, I'm impressed. We only have two clients on it at this point, but if the link tests are anything to go by, it's a bit better than the PMP100 it replaced...
VC Downlink Uplink Aggregate Packet Transmit Packet Receive Actual Actual 18 95.00 Mbps 14.60 Mbps 109.61 Mbps, 7916 pps 68661 (6866 pps) 10501(1050 pps) VC Downlink Uplink Aggregate Packet Transmit Packet Receive Actual Actual 19 31.93 Mbps 6.50 Mbps 38.43 Mbps, 2767 pps 23034 (2303 pps) 4645(464 pps) I have it running on a 20mhz channel right now, just because I can (there isn't any other 900mhz in the area), the second connection is a little on the weak side because of some terrain issues, but it actually wasn't too much worse running on a 10mhz channel (about 30mbps aggregate, if I remember right). It's not the best time of year to be testing NLOS connections, but signal levels are pretty close to what they were with the PMP100, so I have a pretty good idea what to expect... I'm hoping that dual slant will cut through the leaves a bit better than single H-pol did, but even if signal levels are only as good as the PMP100 was in the summer, this is going to be very usable. On Sat, Dec 19, 2015 at 8:24 PM, Eric Kuhnke <[email protected]> wrote: > That's pretty good, I was expecting an average of more like 18-20ms. The > occasional spike to 250 for 1 or 2 seconds at a time is not unexpected at > 900MHz with noise and retransmits. > > > On Sat, Dec 19, 2015 at 4:36 PM, Ken Hohhof <[email protected]> wrote: > >> SM is bridged and has an RFC1918 IP. Test would have to be run on-net. >> >> I just tried a few ping tests from the (Cisco) tower router. 100 pings >> typically comes back around 4/10/35 for min/avg/max. I ran 1000 pings >> though and got 4/11/48, ran another 1000 and got 4/12/248. Zero packet >> loss, but apparently noise can cause an occasional latency spike, probably >> due to upstream. >> >> >> *From:* Eric Kuhnke <[email protected]> >> *Sent:* Saturday, December 19, 2015 6:15 PM >> *To:* [email protected] >> *Subject:* Re: [AFMUG] 900 MHz 450i update with better results :-) >> >> Any chance of running 'mtr' against the CPE IP of house #2 and leaving it >> for a few hours? I'm curious what the min/max/average latency looks like. >> >> On Fri, Dec 18, 2015 at 9:23 AM, Ken Hohhof <[email protected]> wrote: >> >>> So yesterday I did some tweaking on the sync parameters and was able to >>> get better results. I'll put the details in a second post, but I wanted to >>> update the grim picture I painted the other day. >>> >>> As initially set up, I was only getting 1X MIMO-A, even on a 5 MHz >>> channel and after trying a bunch of frequencies, and the SMs would not >>> reliably stay registered. >>> >>> After tweaking the sync parameters and channel, I was able to get one >>> sub to 4X MIMO-B and the other to 2X MIMO-B, and increase the channel width >>> to 7 MHz. They have stayed registered now for 18 hours and while the speed >>> varies a bit they hold 4X and 2X. Here are linktest results I just ran: >>> >>> subscriber #1 >>> 2 miles with a few bare trees and apparently some multipath issues >>> 2X MIMO-B >>> 8.6M down, 4.4M up, 13.0M aggregate >>> >>> subscriber #2 >>> 8 miles with 2 lines of trees and a house in the path >>> 4X MIMO-B >>> 14.6M down, 4.6M up, 19.3M aggregate >>> >>> The AP and subscriber #1 have bad interference across the band, >>> interference is not as bad at subscriber #2. >>> >>> These numbers may not look great compared to 100M aggregate capacity, >>> but they are around a 6 times improvement over what we had with FSK. I'm >>> sure we would see better results at another site with less interference, in >>> fact I have another tower where we can run 2X FSK but 4M aggregate just >>> isn't enough capacity. >>> >>> >> > >
