The “normal” radiosondes are launched twice a day from capital city airports at the international standard of 45 minutes before 00Z and 12Z (midnight and noon UTC time).
However, on occasion we’ve received some interesting signals from sondes that appear to have been launched from the general area of a location in Broadmeadows - which a bit of judicious Google Street View stalking reveals to be BoM’s training facility.
These appeared to be a mixture of the old fashioned RS92s and the new, much more desirable RS41s. While RS41s are quite nice, further decoding of the RS92 signal revealed something even more interesting - it had an auxillary payload to measure ozone.
Noticing one launched early Monday afternoon (a surprise because we’d previously only seen launches from that facility on Wednesdays and Fridays), Michael Wheeler and I decided to attempt a retrieval mission. Due to work commitments, we couldn’t just drop everything and run, so we decided to head out after work and hope it was still sitting where it landed.
The signal was received down to about 1500m altitude - an encouragingly good result leaving us with a search radius of about 1km. The plan was to drive out there then drive around until we receive the signal, get GPS coordinates then find it.
We drove two hours out there, we received the signal but then there was one small problem.
I’m building a LTE to WiFi modem for the car. There are many existing devices to do this, including your mobile phone, so why build one?
I’d like something that can receive a better signal than a mobile phone - especially as our newfound hobby of radiosonde hunting takes us further afield even more than before. And most phones these days are no longer fitted with external antenna sockets. I’d also like it to be permanently fitted into the car, so it’s just “there”. Decent WiFi that extends out of the car would be a plus too.
Here are just a few of the commercially available options for doing this that I’ve collected over the years:
On the left we have the Huawei E585, one of the first such devices. On the right, the ZTE MF70. In the middle, from top to bottom, the Huawei E3872, the Netgear Aircard 785S and the Netgear AirCard AC790S. The E585, 785S and AC790S have internal batteries, and all except the E585 have external antenna sockets.
In this post, we’ll look at why I didn’t just use one of these devices (to be fair, the AC790S still gets a bit of use outside the car) and what hardware I’m planning to build my own with.
All was well, and performance seemed good. Until I decided to pull a large file from the server to my laptop. Despite being almost next to the 802.11ac access point, I was struggling to get even 5MB/sec. It appeared that the noise level on the WiFi channel was quite a bit higher than it should have been. Looking around with a SDR found what looked to be a wireless AV sender blanketing a wide chunk of spectrum in noise. Moving channel only increased performance to about 10Mbit/sec so I plugged into wired ethernet and tested with iperf.
Yeah, that’s not good. Further testing revealed it only affected traffic being switched from a 10Gbit to 1Gbit port.
Tried putting the “stable” and “latest beta” RouterOS on the CRS317, no change. Tried SwOS too in latest and older varieties - although I couldn’t get 2.3 to boot at all (known bug that crashes it on boot if you have any SFPs plugged in (!) but it seems worse than that for me).
I opened a thread on the Mikrotik forums, apparently it’s a “known problem”. I see that the patch notes mention a similar issue on the CRS326…
I’ve changed the network layout to route between two separate subnets, one for 10Gbit and one for 1Gbit. Performance is actually better - in the same configuration, about 950mbit throughput.
Hoping this is fixed soon, because until then I can’t really recommend the CRS317 on a “mixed” network.
So I’ve got an outdoor workbench that has a surplus-to-requirements office chair sitting in front of it. The only problem is that the workbench is at standing height, so the chair’s just not high enough.
I originally thought I’d have to get a whole new chair… but it actually turns out that most gas-lift struts for office chairs come in standard sizing - something which actually surprised me as a software developer.
A bit of measurement with calipers showed that the bottom of the strut measured 43mm, and that the top was 28mm. A quick search of eBay then revealed this one with very similar looking measurements. I ordered the HC200 model, and it arrived a few days later, somewhat amusingly with the top section sticking out the side of the package (something had pressed the button in shipping and the strut expanded).
Unlike the kind of hardware I usually deal with, the disassembly process can be summed up as “hit it with a hammer until it comes loose”. If this proves especially difficult, a larger hammer may help.
Reassembly, as they say, is the reverse of removal - reassemble, then sit on the chair to push both ends into place.
(you may wish to place the chair upright before use…)
(also, you may wish to purchase a foam mat in a better colour…)
With the taller strut, I probably wouldn’t lean back on the chair all the way - but it’s now the right height to sit at the workbench for extended soldering sessions, and I didn’t have to buy a whole new chair.
In the bedroom, there’s a media centre box attached to a projector. The projector in question is attached to the bed, pointed upwards at the ceiling, but that’s another story for another day…
Originally, I wanted the projector to be automatically switched on and off with the media centre. The projector has a RS232 serial port for control, so it was just a case of connecting this to the media centre with an appropriate cable and writing a few scripts to send commands on startup and shutdown.
However, I’ve just replaced the ancient AMD E350 based Mini-ITX box with a shiny new OSMC Vero 4k, an ARM-based box that you don’t really shut down.
Although I could still do this on the Vero, I decided to implement it as a standalone device. Take an ESP8266 NodeMCU board (a nice breakout board for an ESP-12 module) and combine it with a TTL to RS232 adaptor. Then, write firmware for it that connects to a MQTT server, and we’ll be able to control the projector from the web interface of HomeAssistant or any of the Amazon Echo devices.
So it seems that consumer affordable 10GBaseT is closer than I thought in my previous post. Asus has recently released a US$200 10GBaseT card which also supports 5 and 2.5Gbit. Following this, they’ve released another card, the cheaper XG-C100C with a MSRP of US$99 - although in the AU market, it seems to be selling for only about $30 less than the first one.
I’ve got a fileserver with two ZFS pools on it. Many terabytes of storage and much more reliable than individual disks. Single drives aren’t fast or reliable and I can’t fit all my Steam games on my RAID0 SSD pair. But the fileserver can read and write at close to 500MB/second. However, it’s on the other end of a gigabit ethernet connection, so it tops out at 120MB/sec - barely faster than a normal HDD. So what can we do? Isn’t 10GbE still crazy expensive?
I ended up with a drive letter on my desktop mapped to a SMB share on the fileserver and (most) Windows apps treat it no differently to a local drive. With a Steam library on it, I’m not going to be short on space for a while. But every time I go to load something, it’s bottlenecked by the gigabit connection.
(why ‘panzer’? The default name of a zfs pool is ‘tank’ and there already was one on the server)