Me spending 2 1/2 days trying to bring up the PHY.
*sent email to mentor*
10min later I see this in the schematic.😭
I had people double check this :(
#electronics
DJ-Bauer
Also:
Ich studiere in Bonn Informatik und beschäftige mich auch sonstig hauptsächlich mit IT.
Arch/RHEL und Thinkpad enjoyer
After having had another succesful meeting with my topical mentor for my Bachelor Thesis project (I'm doing practical EE and Embedded programming in an way more than computer-science bachelor-thesis scope) I am positive I'm on a good track for a sucessful project.
I'm building a network-DAC, that outputs control-signals for laser-projectors and gets feed from Ethernet.
I don't have to worry about the networking and protocol side of things as I port the OpenIDN Laserdac framework onto my custom STM32H743 based PCB.
Most glue-logic for LwIP and the microcontroller-MAC was already done by somebody else and I could build upon their work to get networking going.
My big software-side is implementing an output-adapter that can stream samples over to 2 SPI-connected DACs with per-sample perfect timing.
I am now on a path where I have a HW-timer & DMA based solution, that only requires CPU-intervention to refill the sample/timing buffers or to apply double-buffering.
I got so deep down the rabbit hole I explored almost all possible memory optimizations accounting for DMA-FIFO behaviors and their implications on the 1st sample timings.
So far the things I learned about is, that realtime audio is a walk in the park compared to this...
If anybody is interested in the project, both Firmware and PCB design files are on my codeberg (see bio) (everything is still WIP and the PCB def. needs at least another change)
Open for any questions if interested as that would obviously help me find interesting/important points I'll point out in my thesis.
Who could have guessed that my "investments" into counter strike boxes as a 16yo (not opening them and targeted purchases and "market analysis) allow me to buy a Steam machine now guilt free.
I need to get that money out of my account balance somehow
Für 2025 das erste mal Steuererklärung als Kleinunternehmer gemacht.
"Ihr Konto ist ausgeglichen"
First try 💪
Gestern bin ich meinen ersten Iron Viking gelaufen.
Ein klasse Hindernissmarathon bei dem es um sehr viel mehr als nur Laufen geht.
Wir sind in einer Gruppe mit 4 gelaufen mit unterschiedlichen Niveaus, sodass alle mal an ihre Limits gekommen sind.
Leider war die Strecke etwas kurz bemessen, sodass ich mit 39km zwar einen neuen persönlichen Distanzrekord, aber immer noch keinen Marathon geschafft habe. Hoffen wir mal auf #2 in Gent am 12. September :)
I had a very pleasant experience getting my Pixel 8 battery replaced by the 2 year google warranty simply dropping it off at my local media markt and them handling the rest.
I didn't even purchase the phone directly through MediaMarkt, so getting that service for free was a nice suprise :)
Now it's finally time to setup GrapheneOS
I am now being borrowed 2x18TB HDD's so that i can move my 3x12TB mdraid with xfs to #zfs.
When I setup that storage zfs did not have the features, that are important for me.
Being able to expand the vdev was not possible until not so long ago.
I'm excited to finally be able to use the QoL of ZFS everywhere
#homelab
Can someone with #linux tape experience help me understand this perculiar behavior.
I'm in the process of manually extracting backups from tapes created using amanda.
So the first 32k bytes block of the file is header that i want to skip followed by the tar archive.
So I'd assume that `dd if=/dev/nst0 bs=32k skip=1 | tar -tvf -` would do the trick.
Instead tar always complaints about an incorrect archive.
Now here comes the kicker:
When I first copy a couple blocks of the tape into a file and then try the same it works.
`dd if=/dev/nst0 bs=32k count=10 of=test.bin && bd if=test.bin bs=32k skip=1 | tar -tvf -`
Obviously in all cases the tape was positionend at correct beginnings of files.
I could solve it by using `tail -c +32769` but that just seemed unelegant.
So it appears to me as if a skip of a block is not really of a 32k block, but instead of a different unit of block device specific
#archive #homelab #admin
First time using kibot to render a #kicad design in the CI.
https://dj-bauer.codeberg.page/OpenIDN-DAC
So you're telling me when I now run a git push Codeberg tells my NUC at home to please run this container in #podman that does a bunch of magic including downloading nonfree 3D models and spinning up a virtual X-Session using Xvfb to run pcbnew and ray trace pretty please and then upload all of that back to Codeberg pages into a nice dashboard?
The only way I can see to top that is running the blender module to use hardware accelerated physically based rendering
The first time bringing up an STM32H7. After having broken the first attempt during assembly apparently the second try worked flawlessly.
I am already glad I purchased a Blackmagic Probe as that thing is amazing for usability.
Especially the integrated Serial Adapter makes hooking it up via USB-C really easy having SWD and UART in one.
This project was the first one where i actively probed by hand with the oscilloscope under the microscope. So the tools are coming together.
I'm still getting used to the workflow of STM32CubeMX and STM32CubeIDE, but it seems decent except eclipse running a gdb session in the background pinning a core at 100% utilization
So, dass erste mal AISLER statt JLCPCB ausprobiert.
Ich habe mir immer wieder mal PCBs quoten lassen, aber jetzt ist das erste mal wo der Preis dem Projekt gerecht wird.
Aber wenn ich dann schon am Montag den Brief im Kasten haben würde, wäre das natürlich ein großes Argument gegenüber der China-Preise.
Heureca! I finally got around to compile my first true Baremetal program for #arm .
From a previous project, I had enough parts and PCB to quickly assemble a SAMD21G15 based test circuit.
I finally fully understood, how linkers work too and how my bootup sequence first has to load the .data segment from flash into sram and zero out the .bss section.
Remember: just because your logging doesn't have to be fast doesn't mean you shouldn't let it sit around BLOCKING INTERRUPTS BY CALLING HAL_UART_Transfer.
Well, my printf statement informing me a new udp stream has opened was enough to drop the next couple ethernet frames.
Which was subsequently logged to uart.
Which caused even more packages to be dropped...
Turns out: I never actually fully implemented a ring buffer.
Now my _write syscall to stdout or is blazingly fast.
So you're telling me Mailman3 advertises itself as "too fast moving for distributions" so that its advertised everywhere to prefer a local venv install over using the package managers....
And then the Python version from DEBIAN is too new for it???
@PersonOfInterest@nrw.social Ich habe sehr gute Erfahrung mit garmindb gemacht.
Das haut dir das ganze in eine lokale sqlite.
Ich habe mal angefangen eine ganz rudimentäre django app zu bauen, die auf garmindb aufbaut.
https://codeberg.org/dj-bauer/garmindash
Funny chasing bugs like the Hardware MAC not properly setting UDP package checksum, which lead to DHCP to work in 1/3 networks