ffhein @ fhein @lemmy.world

Posts

12

Comments

311

Joined

2 yr. ago

gcode only contains high level instructions, e.g. "move 10mm on the X axis", which the firmware has to convert this to electrical pulses that can be sent to the stepper motors. If this is done naively the hardware wouldn't be able to follow the instructions due to inertia, filament viscosity, etc. Both firmwares have several methods to deal with this, but afaik Marlin does not have anything that does the same as Klipper's input shaper. Marlin has "linear advance" while Klipper has "pressure advance" which both try to solve the same problem but have different algorithms and you calibrate them differently. Using Klipper with calibrated input shaper will probably allow you to print faster than Marlin at an equivalent quality level.

Another difference is that all Klipper configuration goes into a text file, and you just restart Klipper to reload it. To change some settings i Marlin you have to re-compile the entire firmware and flash it.

Marlin's G29 command has different syntax depending on which ABL algorithm the firmware was compiled with, which IMO makes it more difficult to help other people who didn't compile their own firmware (i.e. who don't know which ABL algorithm their Marlin has).

I asked someone about this a few days ago, and they claimed to have over 30000 photos in Nextcloud without issues

I suppose "a few" is quite open to interpretation, but I have 50k photos now so if it can handle 100k without getting sluggish it'll probably be fine for the foreseeable future.

Does Nextcloud handle large numbers of photos nowadays? IIRC when I was comparing programs some years ago I read that both it and Owncloud struggled when you got to a few 10000s of photos.

I skimmed through the llama 2 research paper, there were some sections about them working to prevent users from circumventing the language model's programming. IIRC one of the examples of model hijacking was to disguise the request as a creative/fictional prompt. perhaps it's some part of that training gone wrong.

Thanks! I'm going to do some experiments and see if I get different results. I've been using TheBloke's format and it worked mostly well, but perhaps switching to meta-llama's format will eliminate the occasional bugs I've had.

There are lots of free tolerance tests on thingiverse or printables.com too

Sounds good. I think a potential problem with IPA is that it dissolves the fat and then leaves it on the bed as it evaporates, so it's not the ideal bed cleaner.

From what I've read, some people use Windex to intentionally reduce adhesion, primarily when printing PETG on glass to reduce the risk of it fusing with the bed. Never had to do it myself, but supposedly it leaves a film on the surface. Is this anything you've noticed?

Ah, nice.

Btw. perhaps you'd like to add:

build: .

to docker-compose.yml so you can just write "docker-compose build" instead of having to do it with a separate docker command. I would submit a PR for it but I have made a bunch of other changes to that file so it's probably faster if you do it.

Awesome work! Going to try out koboldcpp right away. Currently running llama.cpp in docker on my workstation because it would be such a mess to get cuda toolkit installed natively..

Out of curiosity, isn't conda a bit redundant in docker since it already is an isolated environment?

Not sure I'd describe any printer Creality manufactured after 2019 as "solid" :) They have really shitty quality assurance, so it's a bit of a gamble when buying anything from them.

I'd also take the "5x faster" claims from Anycubic's marketing with a massive grain of salt, and I'm fairly sure you'll get the same speeds on any printer that isn't a just another clone of 5 year old designs. The Kobra 2 has a volcano style hotend, but if you buy a different printer and the hotends melt capacity is the limiting factor you can just put a CHT nozzle in it, either genuine from Bondtech or a high quality Chinese clone like Mellow or Trianglelab.

Not entirely sure how the LeviQ thingy on the Kobra 2 works, but if it is what I think it is, then it'll automate the nozzle-probe distance (a.k.a. probe Z offset) calibration which is very nice. Having a probe automates the bed level compensation, but I've always found it somewhat annoying having to recalibrate the nozzle-probe distance every now and then.

Finding some mixed info regarding the type of hotend the Kobra 2 has.. One review claiming that it is all-metal, while another saying it's PTFE lined. If you're only going to print PLA it doesn't matter, but if you want to print filaments that require higher temps you might need to upgrade it.

Apart from the LeviQ I think these two printers and Sovol SV06 are very similar feature wise. The SV06 is one of the few cheap printers that come with all-metal hotend, but that's usually a pretty easy upgrade on others.

I guess a few hours until I had memorised basic stuff like moving around, copy-paste, etc. Then maybe a week or two before I really felt comfortable with it. There's some point where you actually understand vim and start using it more like a programming language and less like a traditional text editor. For example I love the "repeat last command" (default bound to . key), but to use it effectively you need to create a command that will be usable everywhere you want it to, and there are many different commands that do the same thing.

Here's a random example:

Let's say you have the following text in a document you're editing with vim, and you want to delete the word "dance". Vim is currently in normal mode, and with the cursor placed on the highlighted "c":

Monkey dance party

If your brain is still in Notepad.exe mode, you might achieve this by pressing "i" to go into insert mode and using a combination of backspace and delete to edit the line.

A vim beginner might know that "x" and "X" in normal mode works like delete and backspace, without going to insert mode, so "XXXxx" does the same.

Someone who has learned basic movement could instead solve this by combining two commands, first "b" to move the cursor back to the beginning of the word, followed by "dw" (delete to next word).

But there's also a single command that deletes the current word, regardless of where the cursor is in it, which is "daw". And since this is a single command, you can repeat it with the "." key to delete a different word.

Now here comes the vim magic: If you instead want to replace the word "dance" with "pool" you just modify the above commands by replacing "d" (delete) with "c" (change). So in our example you could type "caw" in normal mode, which deletes the word under the cursor and places you in insert mode. There you write "pool" and press Escape to go back to normal mode. Now you have a single command that replaces any word under the cursor with the word pool, which you can repeat anywhere you like.

And if it's not just the current word you want to change, then replace the "w" in the command with something else, for example "ca)" will change a pair of parenthesis and everything inside them, which is very useful when programming C/C++/etc.. And if you have nested parentheses and want do something with two levels at the same time, the command becomes "c2a)" which might look complicated, but it's basically just simple rules that you string together.

https://vim-adventures.com/ might be a fun way to get started, then look up some nice vim cheat sheet and keep that nearby. I would also recommend looking up some example vim configs, because the editor is very bare bones with the defaults. I like relative line numbers for easy jumping between lines, so give that a try.

When you've mastered (?) vim you'll also see less (the program) keyboard shortcuts make a lot more sense. You can also change your shell to vi mode for faster editing of commands.

Like many others I also use NeoVim, but it was quite a bit of learning curve before you get comfortable with it. And you really have to go all in and learn at least the basics, if you try to use it like a normal text editor thinking you'll learn commands as you go along then you're going to hate it.

In addition to having to learn how to use vim, you also need a good configuration and probably some plugins if you want to use it as an IDE. Personally I use fzf, coc, vim-dirvish, lightline, lightline-bufferline and papercolor-theme.

Pre-assembled they're €1200 and €1000, kits cost €890 and €720 respectively.

CoreXY (i.e. the P1P) is definitely sexier than a bedslinger, and if I was going to get either of these two for home use I would definitely go with Bambu Labs. Was just thinking that for a business reliability and support might be worth more, but that totally depends on your use case.

Bambulab might depend on proprietary/custom parts if you need to repair it, but as far as I've seen these are fairly priced. There's of course no guarantee that prices won't go up in the future. Hopefully you won't have to fix it very often though. Unless you have a cheap 3d printer that self-destructs, I think the most common thing you have to "fix" is replacing the nozzle when those wear out. But it sounds like you're only going to use the printer occasionally so this shouldn't happen very often unless you're printing abrasive filaments. Based on what I've seen it appears to be somewhat of an advanced procedure to swap nozzles on the P1P so maybe it's more cost effective for you to replace the entire hotend when it's time for that.

Prusa has exceptionally good support, with 24/7 chat available in multiple languages, which might be extra valuable to a business. I believe the MK3S uses standard nozzles which are easy to change, though it is possible to screw it up and clog the printer so don't let untrained employees do it :)

Any particular reason you're looking at MK3S+ rather than an MK4 btw? And since I mentioned nozzles for the other options, the MK4 is using non-standard nozzles just like the P1P. The Prusa nozzles are almost as expensive but look easier to change, and unlike the standard nozzles the MK3 (and most other printers) use they should be close to idiot proof.

Check https://www.protondb.com/ for the games you play. If some doesn't work, ask yourself if you can live without them.

I've been full time Linux for quite a few years now, but I do have a dual boot mainly for VR. Other than that there haven't been many games that I want to play that don't work with Linux.

Good luck!

I like printing fast, so Klipper's resonance compensation helps maintaining good print quality at higher speeds. I've done manual calibration of it and that worked really well for me, so I don't think an accelerometer is mandatory. Pressure advance also improves print quality, but Marlin has linear advance which I think does something similar.

I also think it's very convenient to keep the whole configuration in a text file and just reload it with the press of a button. Marlin has some settings in EEPROM which you can change through printer menus while others require a serial connection to the printer, but for "advanced" settings you have to recompile the entire firmware and flash it. Most things can be circumvented by adding overrides to the slicer start gcode, so it's not really an issue, but over all working with Klipper feels more consistent and streamlined.

I also think having a full web based UI that I can control from any computer, like Mainsail, is really nice. But a disadvantage is that the display I have connected to the printer (the stock E3v2 display) does not work with Klipper so I have to use my phone to interact with the printer if I don't have a laptop in the same room.

First of all, I'm sorry to hear you've had a crappy experience with your printer, and I hope you have something that works better for you now. What you write does however align with what I've heard about this brand previously.. The printer buying guide that I like to refer to has put Geeetech in the "hall of shame" category, with the following remark:

No quality control, useless customer support, offers paypal refunds to customers who leave 5 star reviews on their Amazon pages, offers to pay to have bad reviews deleted. Pays youtubers to lie about the machines.