PTP550 vs PB5AC400 for 6.7km link

ubun3t · April 30, 2019, 10:55am

Yes, this comparison seems to be ridiculous or a little crazy, but....I am very surprised with the performance result for both hardware.
Here the result :

PowerBeam5AC-400
5645Mhz
40Mhz
Noise: -103

Note: this PB5AC400 with 30Mhz : 165UDP/150TCP.

PTP550
5510Mhz
40Mhz
Noise -104
Firmware: 4.3.0.1

PTP550_OFIC AP SAMAY • Master(5).png

udp-550

tcp-550

I never thought that the result of this comparison would be that. Maybe something is not well configured in the PTP550 or I will be a huge disappointment.

What do you think? What are your experiences?

Cheers,

Vyacheslav · April 30, 2019, 4:45pm

PTP 550 is configured as UL/DL=25/75. Also 245Mbps UDP overloads 40 MHz DL channel capacity ( 75% of UL+DL) that causes latency increasing,

PB 5AC channel is not UL/DL devided and it's throughput can not be compared with PTP550 with UL/DL=25/75.

TCP test shoud have minimum 5 streams.

Also PTP550 and PB 5AC work at different frequency with different interference level and their results can not be comparable.

ubun3t · April 30, 2019, 5:48pm

Hello, I will do the tests again on the same channel and with 50/50 (UL/DL) and "traffic-generator".
I share the results as soon as possible.

Vyacheslav · April 30, 2019, 8:59pm

In this case you should do duplex test UDP and TCP .

I would recommend you to use in PTP550 UL/DL 25/75 and test UDP and TCP by assymmetric duplex in the same proportion 25% Upload , 75 % Download with traffic limitation in MT bandwidth tester. Limit traffic ( especially in UL) to avoid overload wireless interface - avg RTT latency should not exceed 8-9 ms ( in case of PTP550).

In PB 5AC you have not UL/DL split. But you should test devices by the same manner - test UDP and TCP by assymmetric duplex in proportion 25% Upload , 75 % Download with traffic limitation in MT bandwidth tester.

In TCP tests use 10-20 streams. Pay attention, that MT BT is not able to generate more then 330-350 Mbps TCP simplex ( or UL+DL duplex).

In PTP 550 use 4.3.2 firmware.

ubun3t · May 7, 2019, 4:27pm

Hi, I'm going to share some additional data on this team's performance.

The only way to get this hardware decent throughput is in a single channel of 20Mhz or a channel of 40Mhz. No more, at least with the following data that as you can see I am using channels about 6Ghz very clean for testing (to avoid suspicions of "dirty channel", "little offset between channels")
And here we go:

And finally, if you wish, I can send you a video showing real time performance and how bonding works with this scenario and values, private.

Thanks for your time.

Regards,

Vyacheslav · May 8, 2019, 7:24am

Dowlink in both channels is working mostly on 64QAM5/6 and 256QAM 3/4.

Uplink -64QAM 5/6. It means that at RSSI -55/57 dBm ( it is enough for 256QAM5/6) you have high interference , especially in UL.

So channels are not clean.

Are you sure that test traffic passed via two channels? It seems in test only one 40 MHz channel is used.

Please, begin to test with only one channel 40 MHz. Choose channel, that gives you 97%-100% UL and DL 256QAM 5/6 modulation. Use CA to find clear channel.

Please show UDP/TCP test traffic in BW MT windows, not only traffic at ethernet MT. Jitter is growing when wireless interface is overloaded. Your traffic is not smooth and latency jumps when traffic jumps and exceeds throughput capacity.

Please look at https://community.cambiumnetworks.com/t5/PTP-Stories/PTP550-Field-Test-12-km-2x80MHz/m-p/102604#M341 how PTP550 test should be done properly.