[technique] Le gain de la Nanobeam M5 19 a l'air de dependre fortement de la frequence / omni

[Laurent GUERBY]

Bonsoir,

Cette nuit pour tenter de comprendre un soucis de performance
sur un lien AP-STA entre deux nanobeam M5 19 j'ai essayé toutes les
frequences (canal 20 MHz), avec des resultats surprenants
sur deux liens differents (et séparés par des montagnes) a Banyuls
(3.8km) et Montescot (1.8km) :

https://community.ubnt.com/t5/airOS-Software-Configuration/Nanobeam-19-M5-gain-depending-a-lot-on-frequency-higher-is/m-p/1349022

<<
While debugging a poorly performing link between two Nanobeam 19 M5
running XW 5.5.10 I ended up trying all frequencies and found out up to
18 dB RX difference between the worst (5240) and the best (5680)
frequency!
 
To be sure I wasn't dreaming I tested two Nanobeam 19 M5 links in
completely different locations and with different distances (3.8km and
1.8km) on our network and got a very similar frequency response profile
(if not identical), TX power constant at 20 dBm for these tests, only
one STA on the AP, WDS activated, WPA2-AES.
 
Is this kind of result expected?
 
3.8km link results per frequency STA - AP :
 
2 AP - 3 STA width 20 MHz 3.8 km TX 20 dBm 
path theory 20+19-120+19=-62
5180 3 RX -76 noise -98  2 RX -80 noise -98
5200 3 RX -82 noise -93  2 RX -82 noise -93
5220 3 RX -77 noise -97  2 RX -79 noise -97
5240 3 RX -83 noise -93  2 RX -84 noise -95
5260 3 RX -79 noise -97  2 RX -80 noise -97
5280 3 RX -78 noise -95  2 RX -80 noise -96
5300 3 RX -78 noise -96  2 RX -79 noise -95
5320 3 RX -79 noise -96  2 RX -80 noise -96
5500 3 RX -73 noise -95  2 RX -75 noise -95
5520 3 RX -73 noise -95  2 RX -74 noise -95
5540 3 RX -74 noise -95  2 RX -72 noise -96
5560 3 RX -73 noise -96  2 RX -72 noise -95
5580 3 RX -71 noise -97  2 RX -70 noise -97
5660 3 RX -66 noise -99  2 RX -69 noise -99 
5680 3 RX -65 noise -100 2 RX -69 noise -100
5700 3 RX -66 noise -100 2 RX -70 noise -101

Second link 1.8km results per frequency STA - AP :
 
2 AP - 3 STA width 20 MHz 1.8 km TX 20 dBm 
path theory 20+19-114+19=-56 
Note: AP nearby 5540/20
5180 3 RX -67 noise -97  2 RX -70 noise -98
5200 3 RX -74 noise -92  2 RX -78 noise -93
5220 3 RX -70 noise -97  2 RX -73 noise -98
5240 3 RX -75 noise -94  2 RX -75 noise -95
5260 3 RX -71 noise -97  2 RX -71 noise -95
5280 3 RX -72 noise -95  2 RX -73 noise -95
5300 3 RX -71 noise -97  2 RX -72 noise -95
5320 3 RX -71 noise -96  2 RX -72 noise -96
5500 3 RX -63 noise -96  2 RX -66 noise -96
5520 3 RX -66 noise -96  2 RX -67 noise -97
5540 3 RX -63 noise -97  2 RX -64 noise -97
5560 3 RX -64 noise -96  2 RX -66 noise -98
5580 3 RX -62 noise -97  2 RX -64 noise -98
5660 3 RX -58 noise -100 2 RX -60 noise -101
5680 3 RX -61 noise -100 2 RX -61 noise -101
5700 3 RX -59 noise -101 2 RX -61 noise -101

Note: Assuming 19 dBi antenna gain I never did better than 2-3 dB from
theory given the distance, so this gives an effective antenna gain more
like 17-18 dBi in the best case.
>>

Autant de petites variations de gain sur antenne en fonction
de la frequence sont attendues, autant la ca fait 9 dBi
entre min et max ce qui est conséquent (x2 sur un lien
d'ou le "18 dBi" dans le message).

Nous avions deja remarqué a Saint-Gaudens ce type de comportement
pour une Rocket + Omni 13 dBi AMO-5G13, avec autour de 10 dB d'ecart
dans les pires cas entre 5220 et 5580.

Je vais faire des tests similaires sur d'autres equipements
(Nanostation M5 XW et Powerbeam M5) cette nuit.

Sincèrement,

Laurent