Hacker News new | past | comments | ask | show | jobs | submit | jdevy's comments login

Interesting. I've been watching various Wi-Fi HaLow modules for a couple years now. They are interesting alternatives to LoRa and Zigbee which seem to have low data rates (from a quick Google Search) of best casees 50kbit/s and 250kbit/s respectively vs. Halow's best case 15Mbit/s. Obviously, this all depends on a lot of factors (coding scheme, range, noise, etc.).

It's unfortunate that the modules cost so much (single unit costs):

Alfa Network ($20): https://asiarf.com/product/wi-fi-halow-sub-ghz-wireless-modu...

Morse Micro ($30): https://www.mouser.com/c/?marcom=191003943

Silex ($33): https://www.mouser.com/c/?marcom=191003943

As with all these types of unpopular modules, the SDKs and software documentation are not that freely available, except for these $70 SparkFun Pi HATs: https://www.sparkfun.com/products/19956

Those all have FCC certs, but if you get modules from China, without certifications, it can be a lot cheaper ($5): https://www.alibaba.com/product-detail/New-Generation-900MHz... But who knows how they actually perform or what documentation they actually have.

I would enjoy making a long-range Wi-Fi camera with these (assuming you can buy a base station Wi-Fi adapter for your computer, with drivers). If this random Google Image is anything to go by, https://www.macxdvd.com/mac-dvd-video-converter-how-to/artic..., depending on if you encode video into H264 or H265, you might be able to fit a 60fps 1080p into a stream at a decent range.


I'm using HaLow WiFi to connect two houses in a mountainous region. I'm only using it for a link about 800m distant, but it's done what PtP links couldn't do reliably. I only get 13-14Mbps but it's enough for the light use we make of it.

I started testing the modules but then discovered that there are companies on AliExpress selling the same thing as "Wireless CCTV IP bridge", but they can be used for any kind of data if you accept the losses.

I've more recently seen old Ubiquiti 900MHz kit turning up on AliExpress, I'm tempted to try it instead but I am also concerned about counterfeiting and/or if the firmware has been tampered with.


To add, there's a new CAD kernal that is in slow development called truck [1]. Is is written in Rust and you script/make models in Rust. It doesn't have any geometric constraints though.

There was a very short attempt at starting to make a GUI for truck [2] but development has seemed to have halted. There was discussion and hype about this here a couple months ago [3].

[1]: https://github.com/ricosjp/truck

[2]: https://github.com/CADmium-Co/CADmium

[3]: https://news.ycombinator.com/item?id=40428827


Ahh! That was the name, thanks. I was trying to remember it to find the blog post for [3] again. It is uncommonly well-written.

I have 100s of hours of FreeCAD experience from my day job designing injection molded parts for toys.

For some background, for about ~3-4 years (~5 years ago) I started using FreeCAD 1.18.1 in my job (and even more before that for hobby use). I am used to using open-source software with bad UI, so that's not my major complaint. As long as you stick mostly in the Part Design, Part, and TechDraw workbenches, you should get used to the UI. I used the main branch of FreeCAD up until 1.19.1 but then switched to RealThunder's LinkBranch [2]. I switched for the topological naming fixes (some introduced in this 1.0), assembly workbench (not the same as in this 1.0), and other many quality of life fixes (multiple solids per body and 3D offsets for most Part Design boolean operations). It was never great but it got the job done. As long as you never need complex organic 3D surfaces, FreeCAD can work - or at least the LinkBranch did for me, I'll have to test 1.0.

However, my biggest complaint is with the CAD engine FreeCAD uses: OpenCASCADE (OCCT) [1]. As with most CAD engines, this thing is OLD. It does not like to make NURBS surfaces with true tangency to other faces, and it really doesn't like when fillits cross edges into other faces. You will spend hours adjusting cosmetic geometry so that dress up features like fillits and chamfers will apply. Unless some group of PhDs with some hardcore C/C++ experience come along or the company that develops OCCT gets some major funding, I don't think FreeCAD will improve enough for day-to-day design of complex parts for a long time.

Nowadays, I use Fusion 360. I prefer SolidWorks but Fusion is all my job offers me currently. For a CAD package, and coming from years of FreeCAD, Fusion 360 just works. I have tools for making arbitrary complex surfaces (could still be better), I can create fillets that cross into other faces (most of the time), and I can go back in history and edit features and my model will rebuild itself (to a limit, but even the FreeCAD LinkBranch had more issues than Fusion even though it was better than vanilla 1.19.1 and 1.20 FreeCAD). Fusion also has a proper assembly system, which is essential! You can cheat and create parts in FreeCAD by linking sketches to geometry in other parts, but it can only get you so far before you need to go back in time and everything breaks upon a rebuild.

I hate to say it, but FreeCAD has a lot of work to do other than the UI. I want to use FreeCAD but it wastes too much of my time for professional work. I would still use it for simple hobby projects.

I could talk for hours on this stuff.

[1] https://github.com/Open-Cascade-SAS/OCCT

[2] https://github.com/realthunder/FreeCAD/releases


I've heard people compare FreeCAD to KiCad (a PCB design tool). KiCad has been usable for a long time but it's only recently gotten good enough where you might choose to use it over the other choices because it's so good. I've heard FreeCad still has a ways to go before you might choose it over Fusion or something like that.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: