Confused Bird Forum

Full Version: 4G contention demo with 2 routers and an antenna
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Two of the most frequently asked questions I get are:

1. My speed is terrible every evening, but is great early in the morning.  What's the best antenna to improve my speed in the evening?

2. If I get another 4G router and SIM card, can I connect it with my current router to give me better speed?   

To answer question number 1, an antenna may only improve the speed if it either picks up additional bands that the indoor router can't pick up inside, if the readings are very poor (e.g. negative SINR) or if there is another mast that the user can try aiming their antenna at.  As many rural masts are fed by microwave backhaul, aiming for another mast may not help either if the masts are all served by the same congested backhaul.  

To answer number 2, this may not work if both routers connect to the same mast as demonstrated below.  Even if the routers connect to two separate masts or different networks (e.g. Three and Eir), load balancing is quite complex to set up.  It can also cause problems with some websites such as online banking that don't tolerate connecting between two IP addresses.  One benefit however with using two routers would be to avoid latency spikes (bufferbloat) on the second router, such as for VoIP or video calling.  By having two separate 4G connections, both routers will not be sharing the same timeslots on the cell as the other router. 


Test set up

To give a demonstration of real world contention on a network cell, I set up two routers as follows:

MikroTik Chateau LTE12
  • Three bill pay broadband SIM
  • Iskra p-58 LOG antennae pair (outside on gable)
  • Desktop PC connected to primary Tenda node by Ethernet (i.e. no wireless link)
Netgear M1
  • 48 phone SIM
  • Placed in skylight window (older non coated glass)
  • Laptop PC connected by Ethernet

The following are the signal readings of the two routers:

[attachment=295]

As shown in the signal readings, the outdoor antenna gives a fair improvement in signal strength on all three bands.  This is also with the Netgear carefully positioned in the skylight where I got the strongest readings.  Even rotating the Netgear affects the readings with the band 1 and 3 readings falling below -120dB when turned about 90 degrees in the same spot.  If I move the Netgear out of the window, the band 20 reading falls towards -120dB and the band 1 and 3 readings fall below -130dB.  

Throughput test

To test the sustained throughput of each router, I started a download of a 1GB test file from the Dutch Vultr speed test host, one computer at a time.  This is one of the fastest hosts I've come across with most ISPs, which makes it great for running traditional stopwatch based speed tests as well as testing sustained throughput.   

As soon as the network graph filled up with the laptop, I stopped the download and started the download on the PC.  The following are two resulting throughput graphs:

[attachment=296]

The mast I'm connected to performed reasonably well around 8pm at the time I conducted these tests.  What is is interesting is how similar the throughput is with each graph. 

Apart from the occasional spike, why does the antenna not improve the speed?  

The answer is contention, as I'll now demonstrate. Wink


Contention demo

This time I restarted the download on the laptop.  When the throughput graph was about half filled, I started a download on the desktop PC that is connected to the Chateau router.

Notice the dip in the following throughput graph:

[attachment=297]

The red vertical mark is where I started the download on the desktop PC that is connected to the Chateau router.  As you can see above, the download speed on the laptop took a significant hit despite having a separate 4G connection.  

Let's see what happens when I repeat this vice versa.  In this case I download the 100MB test file from Vultr Germany on the laptop while the PC is downloading the 1GB test file from Vultr Netherlands:

[attachment=298]

The red start and end marks where the 100MB file started and finished downloading on the laptop. 

What this clearly shows is that even a high gain antennae cannot overcome contention.  Considering a HD stream consumes about 5Mbps, it would only take about 10 additional users streaming HD simultaneously on this cell to consume the available bandwidth at this time of testing. No antenna will overcome this, unless aimed at another mast.


Other observations: 

Both the 48 and Three network SIMs appear to get equal access to the network, at least with the mast I'm connected to.  This is also despite the 48 SIM being a phone SIM and the Three SIM being an official broadband SIM, i.e. broadband and phone SIM data also gets treated equally on the Three network.