I’ve been running Optus 4G as my primary internet connections for over a year now and most of the time it’s been fine.

This week the 2300Mhz band appeared to go down on my local tower, causing my modems to connect to another tower further away (I think). The performance was horrible and unusable.

Letting the modem decide what band to use resulted in it using 700Mhz (which honestly I didn’t know Optus even supported until I started writing this down). While it worked, it wasn’t great. I was getting about 5Mbps and latency spikes, but good enough to get stuff done.

I had a brilliant idea to get a higher gain antenna and see if I could improve my connection to this new tower so I purchased two DMM-7-38 Panorama Antennas.

Check out the amazing results on my B525 modems:

So that doesn’t look good. No difference.

I thought I was being all clever looking at the raw data, and concluded that I’d wasted my time and money. But then just ran a simple speed test using Fast

Okay, that’s more promising, how about 2300Mhz? Nope, no go, couldn’t even get DNS to work this time when connected on the 2300Mhz band.

So what does that mean?

• The Panorama antenna is good at 700Mhz but useless at 2300Mhz?
• The Modem lies about some of it’s stats?
• There’s more to high quality connection than those stats?
• 4G fluctuates too much that testing itself is unreliable?
• I need to make (or buy) some 2300Mhz Yagi Uda’s and investigate more?

I already don’t understand HF antennas, I don’t want to start not understanding UHF antennas too.

Got my nerves together and tested the YouLoop last night with the same settings as my magnetic loop.

Thankfully it was just okay (but still amazing for the price/performance/size).

But it turns out modern computer hardware is still awful at keeping time. I noticed that my PC had drifted almost 3s in the last day. That’s too much for FT8 to cope with and I could see WSJT-X missing strong signals.

Windows does have a NTP client to keep the time updated, but it clearly only expects a low amount of drift and by default only updates once a week.

You can check the difference between your local time and real time on Windows by running:

w32tm /stripchart /computer:time.windows.com /dataonly /samples:5

I’m running NetTime with a 30min resync to see how things improve tonight. Hope no one minds me hitting the NTP server 336 times more often that normal…

Clearly the real fix is to work out how to replace my PC’s awful crystal with a GPSDO…. Or just sync using a USB GPS dongle I guess.

I finally put together my small transmitting loop antenna. It’s held together with zip ties, sits on a PVC stand in my kitchen and overall is a real mess, but it’s a start and appears to (mostly) work.

Adjusting the capacitor feels almost impossible. I either need a great big reduction gearbox, or to build my own with a more reasonable capacitance range.

I tried to build a fancy coupling loop out of some nice expensive cable, but I don’t have a soldering iron powerful enough… So it’s a few strands of copper wire braided together.

It resulted in an awful SWR of 1.9 on 40m. But there are lots of things to play with.

I cannot actually (legally) transmit so I just ran some Rx tests on FT8 over 24h using a Airspy HF+ Discovery.

Was planning on running the same 24hr test using my YouLoop but I’m a little worried about the results…

Factorio v1.0 has been released this week, in time to finish off my most recent map (with the Rampant mod for enemies and weapons).

It’s been over 4 years since I first played, and I’ve only recently learned the most important lesson for keeping it fun: Embrace the spaghetti. Don’t get annoyed and think you have to refactor all the time, don’t worry about things lining up or being the most efficient. Just keep building and expanding, it’s fun to look back at the mess you’ve made.

For fun, here are some screenshots:

And a 30MB screenshot of my world (12,000 x 10,000):

(It was quite hard to push out, so no trains, just belts everywhere)

The final patch for Terraria, 1.4, came out last month. Thought I’d record my progress after a couple of weeks.

I’ve just playing on classic this time round. Think Empress and the Frost moon event is all that’s left.

Turns out I still cannot build anything that looks good.

But at least I’ve explored a bit on this large world

Made some changes in the last couple of days. I setup a new antenna outside and switched to different software running on a Pi that automatically feeds ADS-B Exchange). Was really painless using their Pi image, just flashing a SD card and editing some text files.

The only problem I had was related to my router/internet connections constantly making requests from different IPs. But that was a simple fix on my router (once I understood the problem anyway).

I cannot directly compare stats with my old setup, but I have received around 2,000,000 messages in the last 24 hours and furthest contact was 360Km which is quite the improvement.

I’m still not 100% happy with the results, I would love to see more activity near Tullamarine and Avalon. I don’t think there is anything in the way, so it might just be a case of needing a bigger mast.

24hr Map

Some stats

If any of this is interesting, it’s really easy to get into. There are a variety of cheap dongles, windows/linux software, and pre-setup Pi distributions. There’s also plenty of information about building your antennas like this and this

Been playing around with SDR using a RTLSDR dongle, and right now I’m receiving ADS-B from nearby aircraft.

Was shockingly easy to setup using ModeSDeco2 which even provides a local web interface (you would normally feed a site like ADS-B Exchange that combines the results from thousands of receivers around the world).

I’m using a simple dipole antenna stuck to a window with poor LOS to the sky (amazing planning), but I’d like to setup something a little better in the future and maybe feed too. So I thought I’d save some images of my current results to compare with later.

Map

Breakdown of contacts by distance

Coverage pattern

Summary over 24 hours

(Blue are messages, red are positions. It’s very quiet now as you’d expect)

In an attempt to keep my rubbish bin from being (accidentally) taken by my neighbors, I tried to make a 3D printed stencil to make my unit number a little more obvious.

The first version was just a simple stencil. This probably worked the best, but required you to be careful of over-spray.

Then I tried to be clever and make a two tone number and add a shield to make it easier to apply. The idea was you did the background using the new stencil, and then drop in the original stencil and do the foreground. This actually fixed the problem of over-spray nicely. But the result didn’t look very good.

The main problem is that you aren’t going to get crisp lines as the 3D print is never going to be flush with the surface (enough that spray paint wont get under). I suspect the best way going forward would be to cut it out by hand on vinyl (Or make a vinyl cutter holder for my printer).

Overall, at least it’s pretty obvious which bin in mine now :-)

(Note: No STL as it barely works and I’m not happy with it).

I can never remember tar command line arguments to do a simple backup of the filesystem, so this time I’ll write it down.

mkdir -p /backups
cd /backups
tar -zcvpf /backups/img.tar.gz --directory=/ --exclude=proc --exclude=sys --exclude=dev/pts --exclude=backups .
python -m SimpleHTTPServer

# wget http://$SERVER_IP:8000/img.tar.gz

Stolen from: https://old.reddit.com/r/webhosting/comments/475a9b/is_there_a_way_to_download_digital_ocean_server/

Trying to get my head around all the different NUCs and their performance/price.

Price was taken from Amazon (in USD) and performance from PassMark. Neither are very accurate, but should be adequate for an overview.