PTP550 Link Testing vs. iPerf testing

Link tests yield north of 90mbps on a 100mbps port. iPerf testing yields ~30mbps.

Link reviewed and looking perfect. PPPOE MTU is 1492 and Radio MTU is set to 1550.

Does anyone have any ideas why I’m still getting degrading speeds? Any suggestions are welcome!

I’ve ask already to Cambium…
550E have a serious BW problem… firmware depend…
We waiting a new firmware!


FWIW, this is our PTP550


That’s only 1.35KM link, and it’s very clean where it is, and signals are -52 with a 42 SNR, so that it get’s DS9/DS9 most of the time.

My links are around 4km. I’m not getting anything close to those stats. Any chance we can compare configurations via zoom call?

Have you used the link planner for this link, and if so, how does the prediction compare with the actual performance?

I did use link planner and the aggregate data throughput was calculated at 582mbps. This link yielded no more than 35mbps. Installed link testing showed capabilities of 90mbps+

Does the PTP550 have the same problem?

Was this connected to the network via DHCP or PPPOE?

How do the RF measurements (in particular the total path loss and received signal level) compare to the link planner predictions?

Have you seen the Link is unreliable or does not achieve data rates required section of the user guide?

RF Measurements all check out and match up more or less +/-3db. The link is sound.

Is the link operating in the expected modulation mode, but not achieving the expected throughput? Or is it not managing to reach that modulation mode, and running at a lower mode instead?

Can you please share the iperf command you are using for the tests. It is possible that this is a FW problem, yet I also have seen some unreliable results with iperf. Generally, you need to tweak your traffic a bit to get a clear result when testing wireless links with iperf.

UPDATE - The link was setup on a PPPOE account for authentication to the core. We found out during troubleshooting that using the account credentials as provisioned by our system increased speeds to alot closer to link testing speeds. Although a generic PPPOE system account could still authenticate and pass traffic, it was at greatly reduced speeds. Strange.

What are you trying to use this link for. Complete disclosure from a manufacturer regarding if they use bonding what options are available if any to tweak is necessary when considering using links like this. For example, with probably the most commonly used 802.3ad often mentioned as LACP for many reasons including typical low CPU usage often necessary on low power (wattage) less CPU powerful devices, if you are in a scenario such as a single client or multiple clients shoved through one pipe accessing one server 802.3ad will not typically work in a bonding scenario. There are write-ups about why that is true such as one I was recently re-reading on a Mikrotik forum under common mistakes or something like that. If available other bonding options may perform better. In most “dumb” wireless pipes as long as multiple users are going through it accessing multiple servers than 802.3ad will usually work very well. A company like Cambium could if they chose to give customers the things they need up front to know if a radio would work well in a scenario and maybe they do already such as what channel bonding they use. They also could in the GUI or otherwise let network operators know how much traffic is passing through each bonded channel. That’s usually necessary information in troubleshooting if a bonding solution is working well in a particular setup. I haven’t tried these radios myself but what has been shared thus far might be an indication although you may have iperf setup to not work well with bonding it may or may not be reflective in throughput depending on how well the bonding solution works with the traffic you are intending to pass. That’s often why if bonding is used such as in licensed links many operators such as myself would if possible prefer the radio not do bonding itself unless complete disclose and potential different bonding solutions are possible and instead would prefer one interface such as an Ethernet or fiber port capable of fulling supporting the wireless traffic on a particular band such as a 40mhz channel. that way if needed the operator can choose what bonding techniques would work best and appropriately invest in hardware that can meet those needs. If for example one server / router is the end point for traffic going through a link 802.3ad will not work well and investing in bonding horsepower not typically in lower power radios would be necessary. Food for thought anyways.