32 bit 3D printing - What it means and when you need it
ฝัง
- เผยแพร่เมื่อ 26 เม.ย. 2024
- In the next video, I'll be doing a step by step guide to fitting an SKR V1.3 32 bit board to an Ender 3, but before that, I wanted to cover some background. This video explains what 32 bit means and why it is an upgrade over 8 bit Arduino compatible microcontrollers.
The concepts are simplified as much as possible and demonstrated with examples and diagrams. Hopefully this can help you understand why you would go through the hassle of setting up the software chain required for Marlin 2.0.
Previous Marlin explained video: • Marlin firmware explai...
If you want to learn more about Marlin 2.0 and its hardware abstraCTION LAYER, START HERE: Marlin 2.0 Github: github.com/MarlinFirmware/Mar...
Guide for fitting a MKS Gen L if you're not ready for 32 bit: • Easy and cheap Ender 3...
Buy quality and affordable filament from X3D. Buy 3, get 1 free and a free sample pack with every order: www.x3d.com.au
Take a look around and if you like what you see, please subscribe.
Support me on Patreon: / teachingtech
#3dprinting #32bit #skr1.3
![รุ่นใหญ่ EP.2 บุกบ้าน เสรีพิศุทธ์ ระหว่าง บิ๊กต่อ กับ บิ๊กโจ๊ก อา รักใครมากกว่า ?? l [Nickynachat]](http://i.ytimg.com/vi/j2inUqhN__0/mqdefault.jpg)
Hey Michael, I have been a software developer for 43 years. Usually whenever I see someone explain bit depth I see someone struggle, use confusing verbiage, or wanting to move on to another topic. On a few very rare occasions you see someone explaining it very well. You did it very well.
Except he didn't. Old 8 bit processors used in consoles had 8 bit registers and 16 bit addressing so they were not actually completely 8-bit CPU. And he comparing CPU speed 100 MHz vs 16 MHz and deciding "32 bit better". Use 8-bit at 100 MHz and you will have more than enough CPU speed for printing.
@@vladimirseven777 This is not the case for all old 8 bit processors; for example the 6502 did not have 16 bit addressing. What you are referring to are extensions to 8 bit memory spaces using bank switching in order to save costs or implement features ahead of their time. That there are exceptions and additions to an evolving design does not make the explanation of the basic design faulty.
@@louis-ericsimard7659 Not faulty, but very incomplete. Like modern AT-mega or Arduino are the same as old 8-bit processors. Bank switching in 16 bit processors was implemented to address memory above 65535 while pure 8-bit can address only 256 bytes.
@@StavrosKor It depends from application. For example if used for temperature measurement or digital to analog conversion 8 bit data bus or data width used in CPU it limited to max 8 bit capacity - from 0 to 255 (256 states). 16 bit (65536) or 32 bit (somewhere 4 billion +) - you can use better precision, for example work with 0.001 of temperature. But right now 1C degrees (for example from 20 to 275C) or stepper motor drivers (max 256 microsteps) seems enough.
@@vladimirseven777 That isn't true.
You can basicly calculate everything in an 8bit processor to the same precision a 32bit processor can by concernating the calculations using things like carry-flags. The only problem is, that it takes a lot more clock cycles for your calculation.
Modern 32bit CPU, like stm32, also come with an FPU (floating point unit) which handles floating point operation in much less cycles that the standart ALU of the Arduinos ATMega2560.
Problem of the stepper driving is that the Software on a 8bit processors in use aren't able to drive more that 10.000 steps/s due to lack of ressources with the workaround of sending multiple steps at once to reach up to 40.000steps/s. This method is called double and quadstepping which you can find in configuration.h in the marlin files and produces an absolute horrible sound, vibrations and possibly steplosses, as seen in the video. More than 40.000 isn't possible at all without disturbing the whole software loop comprising things like closed loop temperature regulation, watching for endstops being hit, safety loops (temperature runaway)etc.
Having 80steps/mm for X/Y axis at 16µSteps limits your movement to 125mm/s (=10000/80). If you increase the steps per mm by increasing the µStep-Number to let's say 256µSteps like the TMC allows you to do, you would be limited to ~8mm/s.
Very nicely done, Michael! This level of tutorial, first generic, then specific to 3D printers is just what the community needed. The level of information is perfect.
Just bought the same board for my Ender 3. Been putting off swapping it in, but there's no excuse now. Thanks for your hard work! Your channel is awesome.
4:50 if you already know what a bit is and want to skip to the part concerning 3D printers.
Thanks 👌
Thanks man
This is the best description of how a microprocessor works I have ever seen!
Nice job! It is nice to see positive evaluations and learning for new folks!
Perfectly explained as usual! Very interesting and very handy to know the differences! Cheers!
I would love to see a comparison between a 8bit system with klipper and a RPI and a 32bit board. To see how the 8bit board compares when the calculations are done on the rpi.
And as always a good and clear video. Keep it up
I love the way you explain things
very clear, concise and easy to understand
This is why your channel is called "Teching Tech" I learned a lot. Thank you so much for this outstanding Content. Greets from Germany!
As always your explanations are great and simple to understand.
Great topic and I really appreciate you guiding us into this new technology!
Definitely helpful information for a lot of us! And now I know enough to know that I don't really need to know any more, for what I do. Thanks!
I'm commenting only for thank you for such a good job doing videos, not only this one but every other videos. Nice work! Really helpful for newbies like me. I appreciate it a lot.
great video you are truly a great teacher in every definition of the word.
stay warm
Alex
Ordered mine last week, looking forward to the upgrade comparaison
Perfect timing, I just bought one for my ender 3 :)
Very easy to follow explanation! Easy to understand, I gues could be used even in schools. Great!
Your description of binary was amazing
YEEEEEES ! I bought a SKR and i'm waiting to arrive! Your video is going to be super helpfull!
Just when I start doing my own research on the SKR.
Thank you!
me also lol
I agree with Louis on this. you explained a very complex subject so easily and clearly.
Another awesome video, going online now to order my SK board so I am ready to follow you and upgrade my Ender 3.
my friend you teach something i was not expecting, by the first time i really learn binare codification
Perfect !!! Just what i was waiting, how to install the SKR.
This has been explained so well, thank you.
Nice intro
Looking forward to part 2
Thanks for sharing :-)
Another terrific video. You explain topics very well.
you sure did teach the hell outta that tech, cheers!
Good timing, I bought the same board a couple of weeks ago and am just finishing installing it into my printer.
We need 3d printer boards with 7 nm ZEN2 architecture, 64 cores and 8 ghz!
Max Lindner what would be awesome but total over kill but take a 1st gen core i7 980-990 extreme and turn it into a 3d printer all in one board. , octo-print , arduino ide, a slice of your choice fusion 360 and or what ever else you can throw at it. Even a high end 1st gen xeon. X58 chip set board or what ever. For the lower cost but way to much process power lol man that would be sick. I wonder if it could be done. I assume so.
@@mitchh6471 I have seen a similar thing done before by a friend. Take a look at Klipper firmware. It's general idea is process the print via a PC and pipe it to the Arduino in the printer.
You can run Klipper on a RPI. Or a high-end multcore workstation if you have one spare lol.
@@BillyBobJimPatton I actually do have a couple of pc's standing around and collecting dust. William, I think this is one of the few occasions, where a youtube comment has given me the momentum to do something great!
@@maximilianlindner happy to have bucked the trend and give a useful YT comment. I should screenshot it and frame it as proof that it can happen haha
I'd be interested to hear how you get on with trying it.
Thank you for the clear explanation!
Thank you for sharing you’re knowledge
This is a great explanation 👍
Great video. Definite thumbs up for this one!
Very Good content! I really like your channel with this type of content rather than the printer review. Keep it up and I will give you a shout out in my next
Wow, you are smart about this stuff! Thank you!
Wow thanks for the review of my System Architecture class...
Nice video! I'm waiting for skr mini e3 in ender 3 tutorial :)
Thank you for the informative video.
Great descriptive upload.
Condensed too a T.
Quality content.. well done.
Hopefully you can mention the TMC2209's that are coming embedded on some ender 3 direct replacements. Not even sure if it's fully supported but it's the cheapest/easiest way to push into 32bit boards with the ender series
Muchas gracias por compartir tu conocimiento!
nicely explained, thank you. This would be the first video I have watched that clearly explains the differences, drawbacks and complications around 8 bit vs 32 bit boards. Love your work, keep them coming. cheers..
Thank you for this video, I literally just asked a question about the difference between the boards and whether they are necessary on the facebook ender 3 groups.
omg, Alex kid! You gave me a nostalgic heartache😢
You should do a test of showing print quality differences as well, especially when using some of the smoother drivers such as 2208
Your channel is great, thanks for the content. You are one of the only ones reviewing the A10M on TH-cam that I trust, because of that I have a couple questions.
What are your thoughts on the Geeetech A10M after a couple months with it?
Is it useful for single filament use too ?
How is the maintenance ?
Would you still recommend it?
Wow! I just posted a request for this on Michael's Patreon Page. Now that is responsive!
He was probably already working on it.
Excellent video! Subbed!
Would your next video for the ender 3 also work for the ender 5?
Great work btw, newly subbed, and new to 3D printing, you've help enormously
Great explanation Micheal, Can't wait for the next video, it's the board that I plan to put in the Borg. Do you plan to cover using the E1 for the second Z motor? I haven't even touched Marlin 2.0 yet, but will be having a practice soon ... after your video
Awesome video. Would you recommend a 32-bit board, such as the SKR 1.3 as an upgrade for an Ender 3 (Pro) ? Purely looking at it from a perspective with whether it will improve print quality or not?
can u do a review of the FLSUN QQ 3d Printer? it also uses a 32bit controller
I really enjoyed this video, I think it's a great tutorial on why we might need 8 bit versus 32 bit, and I was pleased to see you mention how stepper drivers with higher resolution can aggravate this need. I do have one criticism though, the H-bot style such as the Ender 5 is still a cartesian kinematic, the Y axis just moves the X gantry instead of the bed, and isn't harder to calculate. Spot on about the CoreXY and Delta style tho.
I don't own any h bot machines but I was under the impression two stepper inputs were needed for most movements. I may in fact be wrong, no problem admitting that.
Teaching Tech hey so I think I am in the wrong - I just saw you seeming to refer to the Ender 5 as H bot which it is not. So all good! Good video thank you!
Nice video as usual.
When can we expect the video on the SKR 1.3?
Thanks!
Monday morning Aus time.
What about an accurate pinout for this board? Please include some info on that in the next video. The information for this board is so hard to find! Thanks for the great content as always!
Those leftover 8 bit controllers also make really sweet general purpose Arduino compatible dev boards.
@Teaching Tech
Actually 8/32 bit is not really the precision of the ALU but the width of address map. For a controller it is called x bit only because the address map is x bit width - more the 'x' means more memory it can address.
Page 17 of the datasheet - datasheet.octopart.com/LPC1768FBD100,551-NXP-datasheet-8326490.pdf
This video is great with lots of information. Would you know which bit control board will work for mixing extrusion commands (concurrent feeding of 2 filaments in 1 hot end)? I am using Creality CR-X Pro printer with Marlin 2.0 version firmware. Printer has Creality V2.2 ATmega2560 processor control board. I assume it is a 8 bit board, would it require 32 bit board to execute mixing commands?
I am waiting for a board featuring an ATSAMD51, preferably done as drop-in replacement for the Ender-3.
Hi @Teaching Tech, did you heard about klipper? Great way to re-purpose 8-bit hardware as kinematics are run in another device like a raspberry pi.
My favourite analogy is to say that going from 8-bit/16MHz to 32-bit/72MHz is like going from one milk-float at 15mph on a country lane, to four minivans at 70mph on a four-lane-wide highway.
The bigtreetech SKR-mini-E3 looks interesting btw - $30 drop-in replacement for Ender-3/5, using a 72MHz ARM Cortex-M3 processor and soldered-on TMC2209 drivers - a real winner, I think.
good job man
[Being pedantic for fun, not to be a pest]
There are actually analog computers that aren’t base-2, but I don’t know if they’re still used anymore.
In addition, many SSDs are actually base-4 or even base-8, meaning each memory cell stores two or three bits at a time. Explaining Computers has a very good video about this discussing the new Samsung QVO 3-bit SSDs.
Love your work, keep it coming. 🙂
I know I’m commenting on an old video, it’s just how I roll. ;)
Are but what about the multiplexers or the DACs ? Good work on the basic electronics.
Your review of 8bit vs 32bit is nice. Thre are limitations of 8bit it has been known for a while. Expecially with matrix bed leveling and linear advance adding a lot of overhead into the movement.
Most 8bit setups are running a RPI 3B+ with octoprint to make life easier transfering files wirelessly. If we use the RPI processing power for running Klipper so all data processing is made by the RPI we can print to a very high microstepping. The limitation is made by the transfer speed of the arduino USB. Its a little lower than the processing speed of your 32bit but a lot higher than what 8bit can handle.
I have two klipper videos coming up. Thanks for sharing.
@@TeachingTech from memory a 8bit system with 1/16 steppers (a4988) tops out around 160-180mm/s. With a ABL and linear advance its around 120mm/s which is still a lot higher than most users print at.
Using klipper a arduino has 115200bps baud rate. Since we are only sending move data and sensor information we can use a ABL and LA and still print upwards of ~200mm/s.
Is there any tutorial for building that awesome LED light in the background?
Thanks for the video, great job! could you do a video of marlin setup for Delta style printer such as Monoprice mini delta. My MPMD controller died lost the hot end temperature control. and the controller board is currently out of stock. I know these videos are a lot of work but they are of great worth for those of us who don't have the expertise with firmware builds. There are a lot of problems & fixes that this would help solve on the MPMD as well! thanks in advance!!!!
Would be very interesting to see a full comparison on a Tarantula Pro or an Ender 3... Stock vs 32bits vs Klipper
Have you made any video about microstepping and what you can gain/cons?. Now with access to Skr 1.3 32bit and 2209's maybe that's a path to take? What would going all 1/32 stepping do for example?
Awesome & timely video Michael. I am considering using Marlin 2.0 to help implement BLTouch. But I am afraid the ENDER-5's Sanguino board might be incapable of supporting Marlin 2.0. Is the MKS Gen L board a better choice; or something else. Maybe I should stick with Marlin 1.1.9 on my Ender-5; and look at Marlin 2.0 for my next printer, down the road ..... I look forward to your advise.
I haven't tested marlin 2.0 on an Ender board but on paper it should be fine. The Gen L is superior in every way to the Melzi, but there isn't any need to upgrade unless you want to use more IO pins or change stepper drivers.
Good overview of "bits". Just minor corrections: 8 bit signed has a range of -128 to +127 (not the -127 to +127 as in the video). Also, the 32 bit unsigned has a range of 4.29 billion values or a signed range of -2147483648 to + 2147483647. it may seem funny that there are more "negative" numbers than "positive". In the case of -128 to +127, this comes out to 128 negative values PLUS the value 0 PLUS 127 positive values (128 + 1 + 127 = 256 possible values an eight-bit number can represent).
question any advice on how to reduce that vertical line that shows up on prints. because when printing circles it's a noticeable factor that effects tolerances just enough to not get smooth fits. I think it's from extrustion where the printer stops and starts. Also i use cura, because ya.
Change Z Seam in cura to random to get rid of that line
@@tiaancordier3798 This does not solve the problem, just mask it, because the stopping point of each layer will be distributed randomly, leaving small dots all over the print surface. I think the better way is to sand that vertical line to get rid of it.
when using octoprint does the processing still happen on the mainboard or the pi? if its all done on the pi would 32bit vs 8bit mainboard be irrelevant? you make awesome vids that help a lot.. thank you..
3:02 Surprised no one's pointed this out. N digits of binary can represent up to 2^N values, and thus 8 bits can represent up to 256 values (0-255 unsigned). Early computers represented negative integers as described here, called signed magnitude representation-- one bit is reserved for positive/negative, and the rest is used for the number's value. However, modern computers represent negative integers differently, and 8 bits can represent 256 values (-128 to 127). This happened partly because of the cost of transistors in the early days of computing, and partly because it was kinda silly to have both a positive 0 and a negative 0. en.wikipedia.org/wiki/Signed_number_representations#Two's_complement
Signed magnitude representation didn't catch on for integers, but it lives on in modern floating point (IEEE 754). Floating point does have a positive 0 and a negative 0, and they're used in cases where, say, a mathematical function has a discontinuity at 0. 1 divided by x, for example-- approaching zero from the positive side of the X axis, it shoots up to positive infinity, whereas approaching it from the other side, it falls off to negative infinity.
Morio, I got to 3:15 and went nope, 2s compliment. I'm an embedded designer.... Apart from that, great explanation
Just small notice. Download speed is measured at Bits/s not bytes/s (B - bits, b - bytes)
@ 3:18 you reference the max decimal value that can be represented in a 32 bit binary number as 2,147,483,648. That is for a signed number or two's complement. That effectively makes it a 31 bit number. A 32 bit binary number is 4,294,967,295 in decimal format.
My primary reason for a motherboard upgrade is for quiet operation and having 5 servo outputs for a future dual z-axis upgrade with auto leveling.
Hi, have you tried using klipper with octoprint and ender 3, I managed to get faster prints but not so clean do you have any comments or advice??, congrats on your channel by the way
I was thinking a long time to upgrade to the MKS gen L but up until recently i wont. i will use the SKR 1.3 because 2 killer features: SD card and 32 Bit proccesor
MKS Gen L can have sd card. It requires to buy SPI card reader for $0.5, connecting it to AUX3 and some small change in marlin.
Hey i have an Ender 3 Pro with the old Board. Will it improve Print Speed and Quality if i upgrade to this Boards or is it only Noise REduction and more Functions like Auto Bed Level and other stuff? Thx 4 Answer and nice Videos man
Wow, I think you covered it all, every painful detail.
So is the controller in my Anycubic Linear Kossel 32 bit? I loaded Marlin 1.1.8 onto it about 8 months ago and now it's dialed in its completely dependable and delivers high quality prints. Same with my Ender 3. I know it's got a cheap-ass 8 bit board but it works well. And there is the issue. My interest is not in building the best printer but in having reliable printers to output my projects. They are both dialed in such that I rarely have to level them and I'm not sure screwing around with the controller will give me additional value. Maybe I need to build a delta from scratch.
If it's reliable and printing to your satisfaction I would leave it.
As written above, I think the rule "If it ain't broke, don't fix it" applies in your case.
-You're already satisfied with the prints, which means it might not be worth all the trouble and frustrations upgrading.
But at a later point, when you think your Kossel is no longer an interesting piece of technology, you could upgrade it to a 32-bit board, which should then give it 'new life'.
My CNC mill runs on an FPGA (smoothstepper) because it is more able to kick out the steps at regular timing with the chip set up specially for the purpose. Are there 3D printer boards that use FPGA?
Great video. I'm wondering which drivers, TMC2209 or TMC2208?
Not a ton of differences between the two as far as usability goes for 99% of the people/setups. Get whichever one is cheaper.
The last 32 bit upgrade I did from an 8 bit board was about a 25% speed improvement using the same gcode file with the 8 and 32 bit boards. One of the things I noticed was the 8 bit board had a very short delay between gcode commands and that disappeared on the 32 bit board.
Very interesting, this is something I have to test.
In the future, when slicers and printing algorithms become more sophisticated, expect a much larger speed difference.
In particular, I'm talking about smooth surfaces formed by arcs rather than lines.
Cortex-M microcontrollers spend only one clock-cycle for multiplying two 32-bit numbers (with a 32-bit result), whereas an 8-bit MCU would need at least 19 clock cycles, so a single multiplication would thus be 19 times faster if it's 32-bit. For things like arcs, you'd use multiplications a lot. Cortex-M4 and M7 have a floating point unit in addition, which means one can take advantage of a built-in square-root instruction (which can be used to calculate distances between two points - this will probably not be used in the near future, though).
A simple MCU like STM32F103C6 has 20KByte RAM and USB interface; this IC can be acquired for less than US$2.
An 8-bit Atmega usually cost more than $2, and most of the boards I've seen for controlling 3D printers and CNC-routers use some CH34x or PL2303 chip on the UART to get USB functionality. This requires more PCB space and also more ICs (which means extra shipping cost for the manufacturer); it all adds up.
So even though the STM32F103C6 is only 72 MHz and only has 20K RAM, it'd still beat the 8-bit AVR in every case I can think of. The STM series also have some sophisticated motor-control features (which have not really been taken advantage of in any PCBs I know, since all of them use standard stepper drivers).
5:33 Ok so 32bits is a must if the drivers have high resolution microstepping, nice.
A must have for the TMC drivers basically.
for my doot changer (corexy) I consider it a necessity TBH. it was so jerky on marlin with 8 bit. it kept jittering and pausing ever so slightly on corners. it was weirdly like I guess switching from 20fps to 60fps. The most different comes when it's printing text.
I get a lot less ghosting and a bit of a nicer print quality on it. there's far less kick for me in klipper than marling.
With the update of the MK3S+ to MK4 from Prusa, can you make an update on your notes?
I would love if you would do a comparison on the popular coreXY printers. Voron, D-bot, Hypercube, HEVO, V-king, etc. no one really has a complete comparison that I could find.
Possibly because no one owns all of them :)
What are the pros and cons of an skr 1.3 vs the new creality 1.1.5 silent board on an ender 3?
You can still use the Arduino IDE fior ciompiling Marlin 2+, however it takes an age. I've moved to Platform IO.
Can you get 2 identical printers but upgrade one with the 32bit board and one on 8bit then print something high speed? I would like to see the results at the end.
I'm going to try this on my CR10 S5, so will do the experiment. Any suggestions as to which model to use for the experiment?
@@originaltrilogy1 You asking me or did you mean to ask Teaching Tech? I am not sure what model would be good to print but would would guess anything with lots of angles, corners and dimensions within very short distance from each other... something where it has to process many movements on all axis very quickly. Maybe some kind of chain style objects or something with lots of circles and other shapes? im thinking something like this at high speed but i could be wrong. www.thingiverse.com/thing:1363023
@@originaltrilogy1 hmmm perhaps a little boat?
At 4:58 {80, 80, 400, 93 } Love the integer values for the steps/mm. The people who loaded values like 80.0178, or 400.17 and so on were not doing anybody any favors. I wanted my Z axis to stop on whole steps only to improve the accuracy and finally realized my firmware had some BS numbers there. Nice to see others are doing the same. Seriously with a number like 400.17 you are saying with a 400 mm motion the axis drive is off by 0.17 mm. Really. That's half the width of a typical .4 mm nozzle.
Hey Michael, A question for you. Have you any ideas on reducing bandwidth issues with printing over USB? Currently i run 5 printers all running off 1 linux server which runs 5 instances of octoprint using docker. All 10 usb ports on the machine are used, 5 printers, 5 webcams. I don't really notice much issues but 4 of the printers are identical, same firmware, same upgrades and often i find one or 2 of the printers will finish first by upwards of 30 mins running the exact same gcode and my only thought process is that somehow the usb overhead is slowing down the gcode processing to some of the printers? are there any boards that have usb 3 on them? is there any particular board upgrade that will limit this overhead? will a 32bit board solve this? thanks for any info mate.
other option is flash klipper onto the 8 bit board and offload all the maths onto the pi wich then just sends step pulse timing/dir to the controller
Most of those 32 & 64 bit game consoles were actually just 16bit as far as the actual instruction set they could register. The manufacturers used the size of the memory bus to the RAM to justify the higher 'bit' since thats all the public knew as a number being 'better'. But I get your point.
Ok quick question which would be better i have ender 3 pro and want to be able to print PETG and Nylon. So should i get the SKR or The duet. And i also want to throw a touch screen into the mix. Not sure if SKR can handle the add-ons for the high temp settings and LCD. Also all the duet 2's on Amazon say cloned is this ok or should i get right from the duet's website?