One common issue people contact me about is slow speed on their Three connection.  A massive drop-off in speed after 8pm is usually a good indication of contention, which can be either on the mast, interference from other masts or contention further upstream.  In some cases, the drop-off in speed is due to traffic shaping, where certain services such as CDN traffic (YouTube, CloudFlare, Amazon, etc.) is given priority over other traffic such as file downloads and streaming from less well-known sources.

Where the slow speed is due to traffic shaping, getting around it involves trying to find a VPN server on a host that gets priority.  CloudFlare's WARP+ service is an example, but is currently only available for Android and iOS at this time of posting. 

After testing various VPNs, I found that the Hide Me VPN gets priority when using its Dutch and German servers using the SSTP protocol.  I have yet to try other servers as I have been testing using their limited free account.


Testing for traffic shaping

Before signing up to a VPN service, I first recommend checking if traffic shaping is the culprit.  The most reliable test on the Three network involves timing how long it takes to download the 100MB file on the following two links (use a stopwatch):

Leaseweb: https://mirror.leaseweb.com/speedtest/
Vultr: https://ams-nl-ping.vultr.com/

If the Vultr file downloaded much faster, this issue is very likely traffic shaping, in which case I suggest try the Hide Me VPN with a free account and configure the Network Protocol to SSTP as mentioned below. If you currently have a VPN service, check if there is a setting to use the SSTP network protocol, which might also work.

With one Three site I tested around 10pm, the difference was massive.  I chose the 10MB Leaseweb file as the 100MB file was taking too long:
   

The Vultr 100MB file for comparison:
   

If there is little difference in the download speed (i.e. both are slow), the issue is not traffic shaping.  In this case, the issue may be contention on the cell tower itself or with its backhaul feed, particularly if the speed drops off in the evening.  Another issue could be interference from other masts.  If the SINR signal reading is regularly dipping below 0, interference is the culprit, i.e. consider getting a directional outdoor antenna.  


Configuring the Hide Me VPN

Click the Settings cog wheel icon:
   

Go into VPN Protocol on the left (1), then tick SSTP (2):
   

Connect to the Netherlands VPN server and try re-downloading the Leaseweb 100MB file to see if it downloads any quicker.  


Test-run with an affected Three site

When most people think of running speed tests, they head straight to Ookla's Speedtest.  While it is great for finding what ISPs can potentially deliver, personally I don't find it that useful for diagnostics or troubleshooting.  Anyway, here's the Ookla speed test results without the VPN, via the Amsterdam VPN server and finally via the Germany VPN server, all using the same Blacknight speed test server:


The following are via the Amsterdam and Germany VPN servers, letting Ookla automatically choose the server:


Going by Ookla's test results, it looks like the VPN doubles the speed.  Well... far from it!  

Let's look at some more real world tests...

Personally, my favourite diagnostic speed test is testmy.net.  It basically simulates a stopwatch timing by downloading a specific block size and calculating the speed based on how long it took.  

TestMy Germany server - Direct connection, VPN server in Netherlands, VPN server in Germany:


TestMy Texas server - Direct connection, VPN server in Netherlands, VPN server in Germany:


TestMy Australia server - Direct connection, VPN server in Netherlands, VPN server in Germany:


Now let's look at a screenshot of the Leaseweb 10MB file downloading without the VPN:
   

Downloading the larger 100MB file from Leaseweb with the VPN connected to Amsterdam: 
   

Stopwatch timings with Leaseweb test files:
Direct download (10MB file): 42.44s (=1.9Mbps, 7m04s for 100MB!)
Amsterdam VPN server (100MB file): 31.9s (=25.1Mbps, =13.3x faster)
Germany VPN server (100MB file): 35.9s (=22.3Mbps, 11.8x faster)

That's the difference between being able to stream at standard definition and 4K!
Note: I do not recommend streaming in 4K on a wireless network such as 4G, fixed wireless or public Wi-Fi as this can have a significant impact on other users using it in the area.

Stopwatch timings with Vultr NL test files:
Direct download (100MB file): 23.3s
Amsterdam VPN server (100MB file): 31.2s (75% the speed)
Germany VPN server (100MB file): 41.9s (56% the speed)

As the Vultr files were already prioritised on Three, there is nothing to be gained from using the VPN with these.  On the other hand, these still make good test files to see how the VPN server compares with a direct connection.  As I'm using a free account, it's likely these servers are more congested than those available with a paid subscription.

Stopwatch timings downloading Kodi 18.6 for Windows 64-bit (60.2MB):
Tested: 19th May 2020 ~9:55pm
Direct download: 4m28s
Amsterdam VPN server: 22s
Germany VPN server: 22s

As many Three users are well aware, there are many areas where peak time speeds get frustratingly slow.  With some areas, this is mainly due to high contention on the cell or sector, in which case there is little that the user can do other than try using another mast (e.g. outdoor directional antenna) or try another provider such as Eir or Vodafone. 

There are areas where there is plenty of spare capacity on the mast, but where traffic shaping further upstream affects the speed.  For example, if you notice that YouTube plays just fine in 1080p while at the same time some websites are slow to open, this is a good sign of traffic shaping taking hold.  

As I mentioned in another thread, Three's traffic shaping lets some CDNs run at full speed, including Cloudflare. 

Besides accelerating websites, Cloudflare can accelerate your data connection through its 1.1.1.1 App.  At present, it is only available for Android and iOS and costs €3.99/month with unlimited traffic for up to 5 devices.

Test run:

At home, I pick up two separate Three masts.  One mast shows little variation between day & night.  The other mast I pick up further away nosedives in performance after 8pm, but shows clear symptoms of traffic shaping.  

At just after 10pm on the 3rd May, I performed a 6MB test download on every server on TestMy.net.  I then connected to WARP+ and reran the tests.  As the speed was substantially quicker, I run 25MB test downloads on every mirror over WARP+.  Here are the results:
   

Warp+ itself seems to be capped around 20Mbps, but even still, that is a massive improvement with every server apart from the Colorado Springs server, which seems to get priority. 

The following chart shows just how many times quicker each download ran:
   

Here is a direct comparison of the Australian server download test (direct vs Warp+), which had the highest impact:


Finally for the Ookla Speedtest fanatics, here is a speed test with the same Irish Blacknight server, first with a direct connection, then over Ookla's own VPN service and finally over Cloudflare's WARP+:
   

I have not tested this yet on Android TV, but if the App works there, it should help get around streaming issues in cases where traffic shaping is the culprit. 

Cloudflare plans upcoming Windows and Mac Apps, which should bring similar performance to the PC, assuming Three does not manage to exclude it from its traffic shaping.

Note: Before purchasing Warp+, I strongly recommend first checking if traffic shaping is the culprit. This thread gives a handful of tests to try. If there is little to no speed difference in the tests, then the issue is not traffic shaping, i.e. Warp+ will unlikely offer any improvement.

I have often noticed that whenever my Three 4G connection went down, I could not open the web interface of my Huawei B525 on my PC.  This affects Firefox, Chrome and Edge, which just endlessly shows a blank page with the status "Connecting to 192.168.8.1".  Turning off/on the router generally solved the issue - The internet came back and I could access the web interface again. 

When I got a new Netgear Nighthawk MR1100 (M1) router, I ran into a similar issue.  It would display the header of the web interface, but not the login screen.  However, this did not affect my laptop or my phone, so I used my laptop to configure the Netgear.  Once it was online, I noticed I could access the web interface fine on my PC, ... until I changed a setting that knocked out the Internet connection.  

After trying a process of elimination on my PC turning different start-up applications off/on, I found Bitdefender Antivirus to be the culprit.  With its Protection Shield temporarily turned off, I could access the Nighthawk's web interface fine while the Internet was disconnected and the same also with the Huawei B525.  This issue has also been on-going with the B525 since before doing a clean Windows 10 installation on my PC as I have been using Bitdefender previously also.  My laptop has a different Antivirus, which explains why it was unaffected.  

Demo of Bitdefender blocking the router's web interface:

I rebooted the MR1100 with my phone's SIM inserted.  This SIM is PIN locked, so the router will remain disconnected until that PIN is keyed in.

Windows detects that the router requires action and brings up this page.  Notice how it shows the header, but no login form.  This also occurs if I try accessing 192.168.1.1 directly:
   

To temporarily turn off Bitdefender, open up Bitdefender, click the cogwheel icon and click 'Protection':
   

Turn the Protection Shield Toggle off:
   

The router's web interface should now appear:
   

Once finished working with the router's web interface and have Internet access, be sure to turn on the Antivirus again. 

I have my 3 broadband 4G modem in the attic, and live in a remote but high up location, so was getting a strong signal and 5 bars, but a SINR of only 2db and speeds of only 2-3 Mbps at peak hours. And that speed was after figuring out to hide the modem on one side of a water tank, otherwise the speed was much lower.

Anyway I was browsing this forum and the comments Sean made about band 20 being congested started to make me think. I found this Github project which enables you to switch LTE bands from within a browser like Chrome:
https://github.com/ECOFRI/HuaweiLTEBandSelector

I had to modify it to get it to work with Three's custom firmware which they installed on my B525 via an OTA update. 

Results are impressive to me. Both band 3 and band 1 are far faster than band 20, with band 1 giving the best performance, despite also having the worst signal - 4 bars and 0db SINR. Speeds are between 15 and 30Mbps at peak times, which are luxurious speeds for me. I have found that using the external antennas that came with the B525 also help improve speeds slightly on band 1 only. 

Hopefully this will mean no more pulling my hair out while trying to work from home in Citrix. Thanks Sean. 

Code is below. Select it and drag it onto the bookmark bar, then log in to your router and run it.

Code for three firmware:

A few people asked me for help troubleshooting their workplace or privacy VPN over mobile broadband.  The main symtom was that the connection would establish, but had no connectivity other than being able to ping hosts.  While this mainly affect connections over Eir and GoMo, one had this issue with Vodafone 4G also. 

After being able to reproduce the issue with privacy VPN service over Eir 4G, it turns out that Eir uses a smaller MTU size than most fixed fixed line connections.  This also affects GoMo, which uses the Eir cellular network.  Generally a VPN connects with an MTU size of 1400, so the additional packet overhead works fine with the common 1492 and 1500 MTU sizes of most fixed line connections.    

To fix the issue, the VPN MTU size needs to be reduced.  I generally find 1300 works fine with Eir.

Note: If using a company laptop, these steps require Administrator access.

1. Connect to the VPN
   
2. Click Start and type: CMD
   
3. Right-click "Command Prompt", click "Run as Administrator" and click 'OK'.
   
4. Type the following command: netsh interface ipv4 show subinterfaces
   
5. Look for the line with an MTU size of 1400 and a "Bytes Out" column greater than 0.  Use this interface name for step 6.  The following is an example:
      
6. Type the command using the interface name from step 5: netsh interface ipv4 set subinterface "interface name" mtu=1300 store=persistent
   
7. If it says 'OK', you can check that it saved by typing: netsh interface ipv4 show subinterfaces
      

One more thing I noticed researching all this 4G stuff online (pretty bored ) . There was people using a thing called speedify to bond 2 or 3 different 4G carriers and getting faster download/upload throughput. Then there was some other hardware versions looked a bit like a router , either a "load balancing router" or some other thing im not sure what it was called.
Example of it on linus tech tip channel using wired connection though : https://youtu.be/tqbnjgbtDl0?t=23

After upgrading to a new BenQ 4K monitor (model PD2700U), I started noticing intermittent stuttering while watching videos on Firefox regardless of the video resolution.  For example, 1080p videos that played just fine on my earlier BenQ 1920x1200 monitor now kept stuttering roughly every minute of playback.  Basically, the video appeared to play fine, then the picture would either stall or flutter for a second and play fine for another short while.

Thinking it was a Firefox issue, I tried Chrome and while the video didn't flutter like it did in Firefox, it still paused for about a second intermittently.  VLC gave the same issue playing a local video file, ruling out my Internet connection.

A few other things I tried included updating to the latest Intel graphics driver (I'm using on-board graphics), installing the BenQ monitor driver, tried refresh rates down to 30Hz and even closing everything I could think of except VLC and it still kept randomly stuttering.   

Anyway, as my Windows 10 installation is about 2 years old and taking over a minute to boot despite having a fast Samsung 970 EVO M.2 SSD and a Core i7 7700K CPU, I decided to do a clean Windows installation.  Having just installed all the Windows updates, Firefox and Chrome, I tried playing some video and yet again, it kept intermittently stuttering and Windows still took unusually long to boot.

I tried the next process of elimination - I disconnected all the USB cables apart from the keyboard and mouse.  Windows booted a heck of a lot quicker, but the video playback stutter remained.  I then tried plugging in USB cables one-by-one and once I plugged my Seagate 8TB Backup Plus in, Windows took over a minute to boot again.  So it looks like this hard disk is about to pack up.

From searching online for other possible clues, I saw some mentions that the Display Port cable could be the culprit.  As I didn't have another cable handy, I decided to try using video capture software to record the screen, wait for the video to stutter and played back the affected spot of the recording.  The recording had the stutter captured, which I could verify by replaying the same spot, whereas the original video would stutter at different spots during each playback.  This ruled out the monitor and cable as the stuttering obviously occurred before it left the graphics card.  

After taking a break from the issue, it just dawned on me - I wonder if the Windows desktop background is having an effect as it periodically cycles between wallpapers of record players, a theme from the Microsoft Store.  I went into the desktop's personalise options and saw that it is set to change the picture every minute.  For curiosity, I changed this to 1 hour.  I didn't have high hopes for this, but to my surprise, no more intermittent video stuttering or fluttering in Firefox, VLC or Chrome.

So for anyone else that has their Windows Desktop set to show pictures in a slideshow, try changing the "Change picture every" drop-down to an hour or just set a static picture background.  I still can't get over how this was causing video playback to stutter.  I'm not a gamer, so am not sure whether it affects video game playback.

Both Amazon and eBay URLs can be incredibly long, leading to people using potentially dubious URL shortening services.  The main issue I have with these is that it's hard to tell whether a shortened URL is legit or potentially booby trapped.  For example, if I receive something with a bit.ly URL, I usually end up testing it on VirusTotal first to see where it points to.

The reason URLs are so long is that these sites include all sorts of tracking in the URLs.  This lets them track users even when cookies are disabled or blocked, where the URL originated from if shared, which search terms were used to reach this product, if an affiliate is involved and so on.  

As surprising as it sounds, you can reach just about any product on Amazon and eBay with just a tiny portion of the URL.  No need for a URL shortener! 

Shortening an Amazon URL:

All you need is the domain (before the first slash), '/dp/' and the code in-between this and the next slash as highlighted in the following example:

   

This becomes:  amazon.co.uk/dp/B07NMXR6DG

Shortening an eBay URL:

Very similar for ebay:  Just keep the domain followed by '/itm/' and the code before the question mark as highlighted in the following example:

   

This comes: ebay.ie/itm/283222289932

While the shortened URL does not give a clue about the item, at least the recipient knows that it's a proper eBay or Amazon link.  This also eliminates any tracking trail from the link including any affiliate ID picked up along the way.

My friend living near Kilbrittain, Co. Cork wants to get mobile broadband into his outside office, which is at the edge of his farmyard.  From his upstairs bedroom window he can get 3 bars of 4G signal on his mobile phone when on the Three network.

I've had a quick look for him.  From his location the ComReg SiteViewer identifies 3 masts within 6km transmitting for Three Ireland:
   
• (~4.1km) Site ID: 3_CK0394, Operator: Three, Services: GSM, UMTS, LTE
          Site ID: THREE_CK0394, Operator: Three, Services: GSM
  (from house, mast is 15 degrees north of due east)

• (~4.7km) Site ID: 3_CK0386, Operator: Three, Services: GSM, UMTS, LTE
          Site ID: THREE_CK0386, Operator: Three, Services: GSM
  (from house, mast is 20 degrees west of due south)

• (~5.9km) Site ID: 3_CK0248, Operator: Three, Services: GSM, UMTS, LTE
          Site ID: THREE_CK0248, Operator: Three, Services: GSM, UMTS
  (from house, mast is 12 degrees south of due east)
 
This resource documents all three of those sites as having the following 'Band Plan' values in the last column:
 ○ "800MHz - LTE"
 ○ "900MHz - LTE"

This is confusing, as I thought that only 800MHz was being used by Three for 4G/LTE, with 900MHz providing 2G/3G ?

Putting that confusion aside, I don't see any 1800MHz transmission, and wasn't expecting it in that rural area, so 4G+ is not on the menu.

A county-Cork-based provider suggested the following kit to use the 800MHz transmission from Three:

• Poynting Xpol2 3G/4G Antenna - €150 (€121.95 + VAT)
• Teltonika Rut 250 Router - €145 (€117.89 + VAT)

I'd appreciate opinions on that Antenna and Router combo, and would love to hear from anyone with mobile broadband around Kilbrittain.

Thanks - opm

Hi so first of all , I ordered the sma female - ts9 male adapters to try out vodafone later in the week with my setup incase the unlock code I requested from three for my B525 doesnt work. Its suppose to come in 3-5 days so I will update whenever I get either of them. I assume vodafone is also on the 1800 and 800mhz frequency?. I made a short video of the continual ping test during downloading from vultr here https://streamable.com/9wt2k It appeared to be capped at around 10 mbps? it never went over that on the download. when I tried speed tests on other sites it was much quicker. If you get a chance to see it let me know what you think.  Since you said on your blog reply that you doubt three do carrier aggregation on masts so far apart, I found them both on site viewer here and gave a rough idea how far i am below. I'm not sure now whether I pick up the site in dungloe or arranmore, or both lol.

 What I noticed is, Three have 2 LTE masts listed, at both sites,( im guessing it lists 1800 + 800 seperately) whereas vodafone have one at each site. according to comreg. If you  can check them on there maybe you know what it means. Since vodafone have only one do you think that means they only operate one frequency like 800 mhz at each?  Look forward to your input, I enjoy learning about this stuff   .
Ill post the results of my testing whenever I get a hold of the vodafone sim/router and try get it hooked up to my antennas I currently have in the three router. Thanks