CPE210 Frequent dropouts?
This is a point to point link between two buildings. It is approx 300m and line of sight. Problem is that the link drops perhaps once a minute. I have a ping runing for a computer at the other end. Most of the time the ping replies in 2-3ms, but occasionally it goes up to over a second and sometimes there is a complete break for 5..10s. When there is a break the Web UI for the closest CPE210 (slave) is unresponsive and does not reply to pings.
The link is configured as 802.11n, 20/40MHz, MCS 15 and MaxStream enabled.
Have tried almost every channel, currently on 11, noise -95dBm, S/N 44dB and quality between 80..100%
Servers -- CP201_Access_point ---air----CP210_Client---switch----Me
Any ideas?
- Copy Link
- Subscribe
- Bookmark
- Report Inappropriate Content
Over 50m (you wrote ~300m first) a CPE sees even APs located in the opposite direction due to reflections. 50m is no big distance for a CPE. You could try to decrease Tx power on such a short distance, but it will need lots of tests to get best settings for ideal throughput and IMO it's questionable whether this is worth the time to gain just a few Mbits/s.
Remember: WLAN is a shared medium. You can't expect an ideal environment. See what happens with the CPE's throughput if I send data to another WLAN link in the same room during a speed test between the CPEs (still at 40 MHz channel width with WiFi rate 300 Mbps):
Throughput on second (different) WLAN:
$ iperf -t 600 -i 10 -c 192.168.5.10
------------------------------------------------------------
Client connecting to 192.168.5.10, TCP port 5001
TCP window size: 129 KByte (default)
------------------------------------------------------------
[ 3] local 192.168.5.205 port 64342 connected with 192.168.5.10 port 5001
[ ID] Interval Transfer Bandwidth
[...]
[ 3] 70.0-80.0 sec 22.0 MBytes 18.5 Mbits/sec
[ 3] 80.0-90.0 sec 19.8 MBytes 16.6 Mbits/sec
[ 3] 90.0-100.0 sec 20.0 MBytes 16.8 Mbits/sec
[ 3] 100.0-110.0 sec 19.6 MBytes 16.5 Mbits/sec
[ 3] 110.0-120.0 sec 11.4 MBytes 9.54 Mbits/sec
[ 3] 120.0-130.0 sec 7.75 MBytes 6.50 Mbits/sec
[ 3] 130.0-140.0 sec 19.6 MBytes 16.5 Mbits/sec
[ 3] 140.0-150.0 sec 19.5 MBytes 16.4 Mbits/sec
[ 3] 150.0-160.0 sec 19.9 MBytes 16.7 Mbits/sec
[ 3] 160.0-170.0 sec 640 KBytes 524 Kbits/sec
[ 3] 170.0-180.0 sec 13.9 MBytes 11.6 Mbits/sec
[ 3] 180.0-190.0 sec 20.9 MBytes 17.5 Mbits/sec
[ 3] 190.0-200.0 sec 19.5 MBytes 16.4 Mbits/sec
[ 3] 200.0-210.0 sec 19.9 MBytes 16.7 Mbits/sec
[ 3] 210.0-220.0 sec 20.5 MBytes 17.2 Mbits/sec
You can never predict which actual throughput you will get, since you never know who is sending data around your place on the same (or an overlapping) channel. That's why 20 MHz channel width (4 channels required) can result in better throughput than 40 MHz (8 channels required) in the 2.4 GHz frequency band.
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
This is almost certainly caused by interferences. It's not unusual that lags happen over wireless links; this always happens on each wireless link no matter which device. Much more important then ping round trip delays (RTD) is the overall throughput of the wireless link (well, except probably for gaming). I would not call high ping RTD a »drop-out«.
You can try to minimize lags by setting a fixed 20 MHz channel width and selecting a channel which isn't overcrowded. If there are too many nearby APs, changing to the 5 GHz band might help to improve the link quality and especially the throughput. But since WiFi is a half-duplex shared medium, you cannot avoid lags at all, thus you always will have a higher latency compared to a full-duplex medium such as Ethernet.
MAXtream TDMA (it's not MAXStream) makes sense only in PtMP setups – not so much in a PtP setup –, but even with MAXtream enabled your CPEs share the wireless channel with all other nearby APs visible to the CPEs, which use the same channel.
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
AndersG wrote
OK: Thanks. At this point are the dropouts my main concern and unfortunately would going 5GHz mean that I need to change devices, right? As for dropouts, what I mean is that the unit closest to me becomes totally unresponsive. It does not answer to pings on its own address, nor is the web UI responsive.
Yes, 5 GHz would require CPE510.
With »unit closest« to you do you mean the CPE wired to your local network? There should be no drop-outs at all over a cable and if they happen, it does not have to do with the wireless link. If the CPE's web UI becomes unresponsive, I would check IP settings (of the whole network, not only the CPE) and routing. Maybe you can run tcpdump or Wireshark to find out what happens in your local network when the CPE becomes unresponsive.
If this is not what you meant, please post a network diagram so that we know what we are talking about. In your opening post you wrote about pings to a computer on the remote site (over the WiFi link), so I'm not sure I understand what you mean with »unit closest« to you.
- Copy Link
- Report Inappropriate Content
I drew a diagram in a previous posting:
Servers -- CP201_Access_point ---air----CP210_Client---switch----Me
So, "Me" is my computer. I have two pings running. One is to the "CP210_Client", other is to "CP201_Access_point". When the dropouts happen both stop responding. "Me" is connected to "CP210_Client" through a 1Gb switch. There are three more computers and a printer connected to the same switch.
I can perfectly well understand that I cannot ping "CP201_Access_point" if the link goes down, but "CP210_Client" should answer, no matter what.
I can draw a better diagram tomorrow and also run Wireshark to see if I can find anything odd on the wire.
The article you wrote elsewhere on the CPE510 was really interesting and impressive and it might make sense to go to 5GHz in the long run, but I will investigate this first. But from what I understand would 5GHz lead to a marked improvement in throghput? Especially if the ether is congested?
Thanks also for explaining the finer issues of WiFi and why a unit needs to be 300Mb/s to be able to run 100Mb/s :)
One possible culprit would be if something else on the net had the same IP address as "CP210_Client", intermittently, but in that case would changing MAXtream not make a difference right?
- Copy Link
- Report Inappropriate Content
AndersG wrote
When the dropouts happen both stop responding.
Ah, I see, you run two pings simultaneously. Yes, you're right, if the ping to the remote CPE over WiFi lags or has drop-outs because of interferences, the ping to the local CPE should continue independently. So, there is probably another cause if both pings don't get a reply at the same time.
I once saw such a behavior (packet loss) even over Ethernet on a VLAN-aware switch where broadcasts from one VLAN leaked into another VLAN due to membership of switch access ports in an additional Default VLAN.
But from what I understand would 5GHz lead to a marked improvement in throghput?
Depends on the region and environmental parameters. In my country, the 5 GHz band was not much populated when we deployed the wireless link I described in the story. What's more, in Europe it's guaranted to have 20 MHz channel width (e.g. ch100 = 5490-5510 MHz, ch104 = 5510-5530 MHz etc.).
This avoids overlapping channels like they happen in the 2.4 GHz band (e.g. ch1 overlaps with ch2, 3 and partially 4 for 802.11b and ch2, 3 with 802.11g/n). There were only few foreign 5 GHz APs (~5) nearby, albeit nowadays there are considerably more (~35).
We did tests with two CPE210 on the same link and througput was as low as only 3-5 Mbps (~130 foreign APs visible to the CPE) compared to 80-90 Mbps with two CPE510.
Also, CPE510's antenna beam width is a bit narrower (45° H, 30° E) compared to the CPE210's beam wdith (65° H / 35° E) and they can have up to 23dBm transmit power yielding 30dBm (1W) EIRP in the U-NII-2C frequency band (channels 100 to 144).
One possible culprit would be if something else on the net had the same IP address as "CP210_Client", intermittently, but in that case would changing MAXtream not make a difference right?
Yes, MAXtream is just kind of a Time-Division Multiple Access protocol, where the base station WBS or CPE divides the available AirTime into smaller time slots to allow multiple remote CPEs share the same channel without too much collisions. It won't make a difference if two CPEs have the same IP.
When using CPEs in AP and Client modes, they must have different IPs, since they share the same broadcast domain.
- Copy Link
- Report Inappropriate Content
Thanks again for your very detailed explanations. I just wish I had known this before we bought the 210s. OTOH are two 510's approx what we would pay per month for a p-p link from our telco so it might be worth investigating. Less crowded and a narrower beam should make a difference. Right now I see 30-50Mb/s on the link.
There certainly are several APs nearby, but nowhere near what you had. I would guess 10-20 of them visible in WiFi analyser on my phone. I can check exactly today. I can also check what is visible on the 5GHz band.
This is Finland ro regulations should be pretty much like what you have there in Germany.
Yes, both ends have different IPs. The client now has DHCP with fallback to a static address, but I think I better change that to static permanently.
- Copy Link
- Report Inappropriate Content
@AndersG, with 30-50 Mbps throughput I would not consider changing a CPE210. The 2.4 GHz band has advantages, too, such as no requirement for DFS or better signal distribution when there are obstacles in-between both CPEs such as rain or snow.
Try to align the CPE's antennas more exactly, use 20 MHz channel width, set the distance to 0.4km for 300m, change the WiFi channels, preferably to channel 1 or 13. And of course, always use static IPs for stationary devices. But I still can't imagine what causes excessive packet losses (except probably wrong IP assignments, cable problems or a subtle hardware defect).
- Copy Link
- Report Inappropriate Content
I realised that I had forgotten the grounding cable to the client unit so I fixed that. I also turned off the wireless router connected to the same network, but outside the firewall. We used that for clients that visit us. Ran a ping for several hours. Roundtrip time 1-3ms and no packet loss. Both units are outside a window. Sitting on a mount on the window-sill. Bith windows appear to be metallised so having them inside led to excessive damping.
I now realise that i had judged the ditance wrong. It is just 49m according to google maps :) So should I change the distance set in the unit? I guess it alters the characteristics of the signal somewhat?
And yes, 5GHz would probably be more affected by heavy rain or snow.
I wil di a proper LAN benchmark later. For now I use one targeted for consumers of ADSL and such. Tests to a server in Stockholm:
Tidigare mätningar
14 december 2019, 12:06
57.557 Mbit/s 67.8885 Mbit/s
14 december 2019, 12:05
55.0147 Mbit/s 64.1953 Mbit/s
13 december 2019, 19:47
20.1266 Mbit/s 1.25653 Mbit/s
Figures are down and up. As you can see is there a marked improvement! But as I said: Better measure against a local server over the link to get more real values where latency out to the rest of the world does not matter.
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
Information
Helpful: 0
Views: 6401
Replies: 27
Voters 0
No one has voted for it yet.