Shoddy performance from UniFi nanoHD APs

[CntrlAltDel]

Sup peeps,

I’m troubleshooting inconsistent throughput on two UniFi nanoHD access points.

Setup

• UniFi UCG-Max (UniFi Network Appliance - 10.0.162)
• 2x UniFi nanoHD ceiling mounted 3.5m high (Over GbE CAT5E cable runs no more than 15m directly to UCG-Max with UniFi Gigabit PoE AF injectors)
• 5 GHz only, 80 MHz channel width (160 MHz is worse, I've even tried 20 and 40MHz, at 20MHz I get less than 50Mbps also inconsistently)
• Non-DFS channels
• Each AP on a separate channel, and the channels appear unused in my area

Symptoms

• Throughput varies heavily: ~50 Mbps to ~200 Mbps (Do want to make it clear that 200Mbps is fine, my issue is more that it behaves so inconsistently jumping around, on 2.4GHz my throughput goes to even below 1Mbps with jumps to 5Mbps and then all of sudden 50Mbps, something is off here)
• Reproducible on both APs, same behaviour
• Testing done standing directly under the AP
• Around 6 m away in the next room, RSSI is about -82 dBm and fluctuates, matching the throughput swings
• Clients roam/bounce between APs frequently despite good physical placement

What I tried

• Adjusting TX power (lower/higher across presets): no change
• Checked cabling/termination to APs: link appears healthy, nothing obvious
• Change channel width to 20/40/80/160MHz, lower channel widths just result in even lower throughput but equally as inconsistent.

Spoiler: Stats from right underneath the AP

Spoiler: Stats from less than 6m away through 1 wall

I also wondered if my Ethernet cable runs/terminations to the APs are the issue, but everything looks normal and links show as healthy.

 

[degrub]

i'm looking at this from a hardware perspective. Don't know the Ubiquity software.
North America based or EU or ?
How long have these been installed ?

Are these the only two APs you have ?

BTW, the two pictures labeled with spoiler alert are not showing the image.

in the OP, first you stated 5 GHz only (presume 2.4 GHz radio off) and then mention 2.4GHz results in the next section. Are both radios on ?

Are there other APs in the area that you see that have >-75 dB power on either band ?
What the power that each of your APs sees from the other ?

Did you use the Ubiquity coverage predictor heat map utility for initial layout ?

What client are you using for testing ? phone, laptop, tablet ?
How are you testing ?
ideally use something like iperf to test local throughput - client to a hardwired LAN PC and reverse. Don't depend on "speed tests" across the internet.

Are you using included power injectors or purchased separately ?

Just to confirm the cables and power delivery, take one of the APs if possible, and connect with known good/different cat5e cables (both from the switch to injector and from injector to AP) and a TP-link or other Af or stronger rated injector (48V, >0.3 A). Set the AP on a horizontal surface. Same results with local bandwidth tests ?

 

[CntrlAltDel]

i'm looking at this from a hardware perspective. Don't know the Ubiquity software.
North America based or EU or ?
How long have these been installed ?

Region is ZA, and I've had these two for less than a year.

Are these the only two APs you have ?

Yes, these are the only two APs I have.

BTW, the two pictures labeled with spoiler alert are not showing the image.

They're hosted directly through SNBForums, maybe Ctrl+Shift+R hard refresh the page?

in the OP, first you stated 5 GHz only (presume 2.4 GHz radio off) and then mention 2.4GHz results in the next section. Are both radios on ?

I've tested every scenario. I mentioned 2.4GHz because even it is behaving weirdly. But I've tested with 2.4GHz on/off behaviour is identical. Even 2.4GHz performance is terrible only fairly uncongested channels.

Are there other APs in the area that you see that have >-75 dB power on either band ?
What the power that each of your APs sees from the other ?

They're not running in wireless mesh so I dont think I can see what the signal strength is of my other AP. But I'm not sure, I'm checking Radio/Environment log and I dont see the other AP.

Did you use the Ubiquity coverage predictor heat map utility for initial layout ?

No, it was eyeballing from my end, the rough distance that each AP has to cover is just 5m, technically I shouldve been able to install one dead centre of my home and it should reach both ends of the house just fine.

What client are you using for testing ? phone, laptop, tablet ?
How are you testing ?
ideally use something like iperf to test local throughput - client to a hardwired LAN PC and reverse. Don't depend on "speed tests" across the internet.

I'm using WiFi Man, just because it allows for throughput testing separated from WAN internet speed tests.

Are you using included power injectors or purchased separately ?

The injectors I use are UniFi PoE-48-24W-G-WH which come with the APs.

Just to confirm the cables and power delivery, take one of the APs if possible, and connect with known good/different cat5e cables (both from the switch to injector and from injector to AP) and a TP-link or other Af or stronger rated injector (48V, >0.3 A). Set the AP on a horizontal surface. Same results with local bandwidth tests ?

Brilliant idea I didn't think of. I just did this test now, I used my known good UGreen CAT6 cables that I've done 1Gbps on before. And I even removed the AP from the ceiling and kept it right beside me, the issue is identical, jumpy throughput, it went up by 50Mbps more and the signal was about 15dB better. Ultimately the issue still persisted.

And here's a quick snapshot of what happened when I walked 5m away into another room.

 

[degrub]

Has this issue existed from day 1 ?
i think you mentioned you have had these for less than a year ?

That looks like very weak signal level if the client was still attached to AP2-Backdoor. What did the signal strength plot look like for the same walk ?
Low bit rate is not surprising at those low RF power levels.

Does the same issue occur if you turn off one of the APs ?

Just curious, with what are your walls constructed ?
Concrete based or plaster lathe based walls will absorb a lot of signal power. 2.4 GHz may not penetrate well and 5 GHz is unlikely to.

Is this a single family home separated from others ?

Use of the design tool with a plan view of the building and information on wall construction can be very useful for determining likely layout issues. Radiation power plots are based on open air and RF isolated environment.

http://dl.ui.com/datasheets/unifi/UniFi_nanoHD_AP_DS.pdf