-
Finally getting back into some Homelab stuff, which now includes a lovely #Immich instance.
And, I have to say, that is a *seriously* useful tool. Was quick to set up, but far more importantly it is a real pleasure to use. The UI is clean, you can find what'cha need, the albums and maps ... Well, not 100% perfect but work great.
Oh but the context search?? What a REALLY nice tool, you can search for stuff like "Volcano" or "miniature airplane" and it *actually* works!
-
Wish us luck!
We want to go ahead and ask our soon-to-be PhD supervisor if we can slide in a side project of ours - a I/Q Down/Upmixer for the #Digilent #AnalogDiscovery3 !Our main PhD topic will be a high MHz to low GHz frequency range fluid analyser, more precisely to create a more tailored analog/digital frontend for it.
And I think we can make the reasonable case that a first prototype can be a generic I/Q downmixer, to try it out :>
-
One of the things we are interested in learning about are System-on-Modules, things like the Raspberry Pi CM3/CM4S, or maybe even a SoC
Their form factor would make them fantastic for a lot of projects, but the ugly truth is that almost none of them boot mainline linux, and most only boot a very specific, patched, outdated, vendor-specific version...
Except we finally found it.
The x86 SODIMM-format SoM <3 -
Oh, and there's been a lot of fun with new self hosted toys again!!
To aid with our Master Thesis writing, we set up our own #jupyter #jupyterlab server for data vis.
There's also #codeserver and our own #forgejo instance to always have access to the writing and code~And did you know Code Server works great with the #latexworkshop plugin, including PDF preview?
You can completely ditch Overleaf and their horrible 150€/year plan (which you need for git and bibtex!!)Good stuff <3
-
Exciting - didn't know how easy it was to generate #GitLab #JUnit #ci test files, to precisely see which tests failed!
Sure, the source of the files is a ... Somewhat creatively assembled bash script.
But it works!
We're currently helping set up automated CI simulation of #VHDL using a #GHDL docker image, to verify parts of the #CERN quench detection #FPGA code - and it's looking pretty good so far!
-
Goodness we love the #grafana pipeline.
You start off introducing it in a small project at work.
Next thing you know a colleague asks if you can help set it up for their item, then a few weeks later the Section Leader of your group asks if you can get it running for some of their systems.As of now, we are having a trial run of monitoring all 1000+ Dipole Magnets of the #CERN #LHC, using #grafana, #timescale and some #Python running the #dataacquisition
#Observability for the win!
-
And after adding in the complete set of all #Cern #lhc Power Lead temperature sensors, the graph now looks far more stellar!
With 220 sensors providing data we are helping keep the accelerator's #superconductor magnets safely powered - and the data it generates looks absolutely fantastic as well
This graph here no longer shows a plot per individual sensor, but rather a heatmap of all sensors - beautifully visualising the operational heartbeat with #timescale and #grafana for plots <3
-
The heartbeat of a machine can be seen in many locations - in this case, in the monitoring of temperatures inside the #cern #lhc magnet power connectors using #grafana and #timescale
These connectors interface between the outside world and the cosmically cold cryogenics of the superconductor inside. If not monitored and heated, they can freeze and form ice and dew!
As the machine ramps power up and down, so do their temperatures, forming an imprint of the daily operations in the graphs. -
We may once again be tasked with scraping/polling data off of a slow scientific "big data" database and into #Timescale for a smoother live plotting experience in #Grafana
You'd think CERN would use a dedicated time series store for monitoring data but no, its HDFS+Spark??
The kind of system that, when you ask people about, will say "Oh no I've managed to avoid worrying about it until now" and "It works if you're patient".
We'll give them a taste of sub-second query times~
-
Slighlty disappointing to see #Influxdb 3's release saying it's "45x faster than Open Source" (referring to the Influx OSS version), and moving to a fully cloud-hosted system... And yet another rewrite, of course.
#timescale's TSL license feels more sensible, keeping the code open source and usable. We think they found a sustainable middle ground for their software, especially compared to what Influx just did.
-
Hm... *Very* silly thought, but, how hard would it be to build a little Sample-And-Hold decimation system to add oversampling to the #digilent Analog Discovery 3?
>200MHz bandwidth components are cheap, same for ns-level time delays for phase shift ICs...
A separate circuit could time-stretch a very fast signal to something the AD3 could easily measure at ~1MHz, by taking samples at precise delays after a trigger and "Holding" them to be measured slowly.
Only works for repeating signals tho. -
Ok so it's still a bit buggy on windows, and ModelSim doesn't play nice.
But combined with #GHDL for open source #VHDL simulation and #gtkwave for wave viewing it's a surprisingly comfortable to set up tool chain, especially on Linux it's all just via package manager and pip~
Now to learn proper file structuring and documenting habits to raise the code quality bar a bit, probably add proper test benches~
Again #TerosHDL helps with built in docs generator <3