The First PCBs: Open-source Industrial Robot Controller Part 3

Discord: / discord
Patreon: / excessiveoverkill
Paypal: www.paypal.com/donate/?busine...

[@peejay1981]

Just a suggestion - consider using AMP Superseal Connectors instead of the old parallel ports. They can be a bit pricey but much more rugged for industrial applications. There is a PCB mount option are they are commonly used for car ECUs

[@ExcessiveOverkill]

Ah that’s what those connectors are called. We did spend quite a while looking for different connectors. Since it needed ~40 pins, most of the nicer connectors ended up almost doubling the cost of the whole assembly. I would prefer something a bit nicer than the DB connectors though. I also still see them used in industrial applications for connections inside of panels, I just wouldn’t want them exposed to the environment.

[@benargee]

​@@ExcessiveOverkillI think you're doing good. I don't think your industrial robot will be used in an industrial environment. It's always good to keep an eye out for what can make it better though.

[@zyxwvutsrqponmlkh]

​@@benargee Starting with a minimally viable product is often the best. You can always pivot to ruggedize connectors later but it's a waste of money for a self funded basement prototype.

[@HansolControl]

Why not just jump to an actual industrial connector at that point? The man could spend 10k on harting connectors alone. Unsealed Molex connectors are good enough for Fanuc.

[@kca698]

@@HansolControlyup, Molex, or Amphenol in Fanuc competitor OEM

[@max_eley]

Nice update!, No one has done something quite like this before, I am very excited to see where you take it.

[@PoisonGryphon]

10:44 LMAO, followed by, "It's 4:00 AM?"

[@searchpartyguy]

So we get to smash stuff with the robot, cool. Can’t wait to see the open source controller in action. Good job. Went back and watched part 2 of this process, reminded me why this is so cool.

[@vaughnvarma8447]

regarding the line reactor: I don't think I've heard it by that name in industry before, but from a quick google, it looks like it is functionally similar to a line filter, which is common (moreso in Europe than in the US) in industrial motion systems, although typically it is an add-on feature external to the drive, since not everyone wants or needs it. The drive will typically just have the rectification stage, and maybe some smoothing downstream of that to get a cleaner DC bus, still allowing for the drive, as a single unit, to be powered by AC. The most common configuration is for 3-phase electricity, at the voltages typically seen in industrial buildings, 230V or 460V in the US, although single phase configurations down to typical household voltages are not uncommon. In my professional experience, what is actually more uncommon for commercially available industrial motor drives is to have the DC bus externally exposed. If the expectation is that the drive will be powered by, say, 460V 3-phase AC, then the DC bus doesn't need to be exposed, and if it is exposed, you need to make sure the DC bus generated off of the AC power can't make it out of the board with some (probably pretty beefy) protection circuitry. I've seen DC buses spike up past 900V during normal operation, and I expect (as I've never worked in R&D in this industry) VFD/servo drive manufacturers don't see much value in adding the protection circuitry and ports to allow the drives to be powered by an external DC bus, when direct access to 3-phase AC power is more common amongst their customers. This isn't universal; there are some use cases, such as multi-axis systems, where the DC buses of multiple drives are linked together with a busbar, but I think that standalone drives with only an AC power option are somewhat more common.

[@ExcessiveOverkill]

Mostly everything I want to do with the drives requires multiple, sharing a DC bus is great since it helps with the total bus capacitance and gives lower ripple. Instead of a normal full bridge rectifier to go AC-DC, I wanted to feed AC into the output terminals of the drives to use them as active power factor correction. That way the power factor isn’t horrendous like with full bridge rectifiers that only pull power off of the peaks of the ac sine wave. The line reactor/filter (basically just inductor) is needed since by default the ac line has almost no inductance and would result in the drive not being able to regulate the current correctly. The drive should also be able to boost the bus voltage up above what a normal rectifier could do and also support bidirectional power flow so you could feed regen energy back to the grid like some of the newest Fanuc controllers

[@cwbh10]

Some quality engineering there man, love to see it!

[@noelswedzinski4498]

Super cool! nice to see it is coming along nicely, cant wait to see where this goes. I don't know circuit board design but its nice to see the back side of development for once that is more than surface level information! would like to see more in-depth videos about design considerations. I'm an ME so electricity is magic to me.

[@r0sal3sr]

Might want to check out the TI AM2634 as well. TI has a set of RISC processor accelerator cores with tightly coupled (very low latency) memory to the primary RF5 cores. These are called the Programmable Realtime Unit (PRU). Same basic idea: offload low-latency, realtime communication, or control laws out to the PRU cores. Each PRU has its own GPIO and a small set of its own peripherals (UART, SDFM, and Ethernet).

[@devrim-oguz]

Everything looks really nice and professional. We should also put them together in an enclosure for easier connection with different types of robots.

[@Victor-ze3sd]

It would be cool to have a channel that teaches your from start to end how to build/fix these robot arms and run them.

[@peterbuckles2711]

Sweet video, I was excited for this one! I’m almost getting to the point in university where I can understand what your saying 😂

[@CAPTINKING]

very detailed very nice

[@AdaptivePhenix]

Great content! I have been designing/building motion-control systems for decades and my latest development is based on the Parallax-P2 MCU. This device is so perfect for my needs that I often wonder if they designed it just for me 😁 Eight cores ("cogs"), all identical and 64 "smart-pins". Each pin is like a mini-processor in its own right and is capable of many things. Set-and-forget PWM, any pair of pins can decode quadrature (zero impact on any cog) @ sysclock/4, 16bit DAC/ADC, etc., etc. No need to ever use an interrupt. Can be programmed in SPIN (byte-code interpreter), compiled SPIN, compiled C, compiled BASIC (really nice), PASM (assembly). No FPU but I do all the time critical stuff with integers-only (there is a 64bit INT/UINT). The quad-decode is cool because there is the option of reading the absolute count _or_ the delta between reads. It can also report axis velocity. Parallax is not mainstream so very few are aware of this awesome device.

[@rickgreer7203]

The Propeller chip and what came after is a brilliant design. But does require a bit different thinking. The Basic Stamp got so many started with microtrollers back in the day…wish Parallax was better known.

[@hugoboyce9648]

Nice work!

[@piotrmikoajczyk5152]

Great job!!!

[@sabercruiser.7053]

Great work keep it up 👍👍👏👏🙌🙌👑👑

[@22lambo]

I really wish i knew how to design boards like that, where did you learn everything you know? Im assuming you use your knowledge in your every day job?

[@-r-495]

reminds me of the GESPAC system interacting with a Baldor servo driver but with current technology.

[@shaunybonny688]

Idk why I’m even watching this, I have no programming or IC circuitry or electronics background but it’s still cool. 🤯

[@mrraimundo130]

Well done!

[@Citronitroify]

Definetely cool! 👍

[@gottbuilds9363]

I would love to see a rundown of some of the programming involved. Maybe a diagram of what controls what. I've done some embedded stuff but FPGA has always been intimidating

[@EngineeringNibbles]

Looking forward to this!!

[@joshuamouw14]

Man this is exciting. Nice work. I have a 1994 Staubli Rx130 with a complete cs7 controller that has a defective floppy drive keeping me from loading robot parameters. I have been thinking about building a new control to replace the give cs7. I'm following your progress😊

[@SuperHaptics]

Great work. I had same ideas for the mains dc but i have a feeling that direct rectified dc for the motor bus may be avoided for safety and power factor issues. I was wondering what high voltage industrial drivers do. Do they have a proper switching dc power supply to provide all those kw?

[@ExcessiveOverkill]

Most robots have a huge isolation transformer then probably some active PFC, not really sure though, they definitely aren’t a full switch mode supply but I would think they’d be a bit more than just a rectifier.

[@jb5631]

​@@ExcessiveOverkillit you have the connections and or time, it would be very interesting to find out more about it

[@jessebedard2070]

Hahahah adhd at its best my friend 😅 i feel that including the wheel scroll.. like "ummm... hmmm..." 😂 just found youre channel love content definitely going to watch some more intresting stuff. And i like that you explain most everything. Not going to lie and say i understand half of it but i get enough of it

[@F45CNC]

Hi EO, very interesting that you have succeeded in talking with a Fanuc motor encoder and that you are making a servo driver that can control a Fanuc motor. I work with Fanuc Robots and Fanuc CNC controls on a daily basis. but I am very interested in being able to control a Fanuc motor from another CNC or PLC controller, I would like to be able to reuse the Fanuc motor on some of the CentroidCNC NC controllers. their CNC controller has an open interface for controlling their CNC controller, something you might want to look into.

[@NicksStuff]

This is fascinating. I don't understand any of it (FPGA means black magic to me) but I love watching this series

[@CollectiveSoftware]

This is all super cool

[@Ender_Wiggin]

Why do these boards have FRC 971 energy.

[@Scrogan]

I’ve just ordered some TMC4671s and TMC6200s myself, because programming FOC sounds like a pain. I’ll also try to shove some ZXGD3006 totem poles between the TMC6200 and the FETs, though I’m not sure if I’ll end up with oscillation on my gate pins. TVS diodes from gate to source, maybe? I guess to minimise ringing I should design the PCB in order to reduce the gate-source-driver loop inductance.

[@ExcessiveOverkill]

I think the TMC6200 is a pretty powerful gate driver itself. I spent quite a lot of effort to get the gate drive loop area very small, the traces are directly above and below each other to minimize ringing, most of my issue I think is from the DC bus having too much inductance between it and the fets.

[@nin1ten1do]

nice.. perfect-- GG you hear that KUKA?? ABB?? MOTOMAN?? you dont give it to us.,. we make OWN"!

[@bobroberts2371]

Line reactor not commonly used. Could this have something to do with causing problems or needed for regenerative braking of motors ( Power back to the grid ) ? I'm aware that some sort of circuity would be needed to convert variable AC back to stable grid. I know that with a standard inverter running a motor for variable speed with a high inertia load, an external resistor bank is used to burn off excess DC bus power. For example, Haas milling machines use standard home stove top heating elements as an external resistor bank.

[@ExcessiveOverkill]

Brake resistors are common, but it should be possible to use the servo drives in reverse with the addition of a line reactor (inductor) to convert power to/from the AC grid. That way regen energy isn’t burnt off and is instead fed back

[@joels7605]

Beauty! I'll buy one!

[@nanoconfuso7727]

Bro looks majestic af

[@chronokoks]

Imagine going through your local scrapyard and finding a servo (happened many times to me) and convincing a the guy to sell it to you. Then you get home, you wire the encoder carefully, hook the driver to the pc, set up the encoder type/variant, 24V brake on the servo, dick around with the settings so the follow error is good enough and voila you'll be able to power your coffee blender with it :D

[@404Anymouse]

7:41 POWER OFF BEFORE SEVICING!

[@bobroberts2371]

FULL BRIDGE RECTIFIER !!!

[@anterprites]

8:14 that PCB needs some rounded edges

[@paulw2791]

Why would you need rounded edges? Just cuious

[@anterprites]

@@paulw2791 to protect fingers for example, the PCB is pretty sharp. Also to a lesser extent it distributes forces from one point, so it may chip/damage less. Same reason why you make chamfers on mechanical parts, be it from metal or 3D printed. Sharp corners bad.

[@paulw2791]

@@anterprites Oh ok, makes sense.

[@ExcessiveOverkill]

Added to the revisions list🙂, can’t believe I overlooked that.

[@ZOBOFFICAL]

do you have the robot at the end running at 330v or 480v and is it a fanuc robot?

[@ExcessiveOverkill]

It’s running off 48v with Odrives, I went over it in the first video in the series

[@infestus5657]

Nice

[@dharmendrachauhan7884]

How can i learn all embedded from KZbin and how can start learn embedded engineering?

[@gedr7664]

what does entropy source validation do again?

[@TheKdcool]

Entropy source validation refers to the process of ensuring that a source of randomness (entropy) produces output that is sufficiently unpredictable and uniformly distributed, which is crucial for cryptographic applications. This validation is important because cryptographic systems often rely on random numbers for key generation, encryption algorithms, and other security measures. If the entropy source is predictable or biased, it can compromise the security of the system. In entropy source validation, various statistical tests and analyses are performed to assess the quality of the randomness. This includes checking for uniform distribution, unpredictability, and lack of bias in the output. The validation process helps in determining whether an entropy source meets the required standards for use in secure cryptographic systems. (From ChatGPT4)

[@dkraft]

In motion control systems, entropy source validation is crucial for ensuring the security and reliability of the system. Randomness is essential for generating unpredictable and secure cryptographic keys, which are necessary for secure communication and data protection in motion control systems. To validate an entropy source, several factors must be considered. These include the quality and randomness of the entropy source, its resistance to attacks, and its stability over time. A good entropy source should be unpredictable and produce high-quality randomness, even under varying conditions. The entropy source must also be validated to ensure that it meets the requirements of relevant standards, such as NIST SP 800-90B. This includes testing the entropy source under different scenarios and conditions to ensure that it consistently produces high-quality randomness. In motion control systems, it is particularly important to validate the entropy source to ensure that it is resistant to environmental factors, such as temperature changes, vibrations, and electromagnetic interference. The entropy source must also be protected against attacks, such as side-channel attacks and attempts to manipulate the source of randomness. In summary, entropy source validation in motion control systems is essential for ensuring the security and reliability of the system. The entropy source must be validated to ensure that it produces high-quality randomness, is resistant to environmental factors and attacks, and meets relevant standards.

[@JakubKraus0]

@@dkraft Did you generate that with AI? Why? Entropy source validation is a part of the chip he's using, and doesn't have anything to do with the motion control system

[@MuhammadAli-iy4qn]

could you pls share the schematic

[@IainMcClatchie]

Where is part I?

[@zyxwvutsrqponmlkh]

FOR SCOTTLAND 🦎

[@prayashbaniya6434]

can u use nvidia's controller

[@graykimchi]

Curious how your going to setup the Zync. I think this is going to be a massive software/pl effort if u roll ur own motor control loops. I think ur better off using a old pc and using EMC (linuxcnc) and a MESA card for 0 jitter step/dir since the gcode preprocessor in it is really good.

[@AdaptivePhenix]

Control loops are actually child's play and LinuxCNC need to drop the "real-time" from the PC and let a MCU handle it. I SMH when I read the LinuxCNC forums.

[@Yakroo108]

👍👍👍

[@ARJUNDASED]

why dont use ethercat solution ?

[@ExcessiveOverkill]

We did heavily consider Ethercat but due to it requiring a license or dedicated chip that isn’t that great to interface with we decided against it. It’s very much a locked down industrial thing, not very open source friendly

[@ARJUNDASED]

@@ExcessiveOverkill cant you buy leadshine controller and use it ? or beckoff ?

[@AdaptivePhenix]

@@ARJUNDASED Do you buy Beckhoff with your own money? They are a ripoff and a PITA to use. If you decide to _not_ use their IPC, you pay an inflated price for their license. REAL engineers understand that there is no rocket-science to motion control and multi-axis, closed-loop servo control can be achieved for almost pocket-change.

[@aurorajunior6328]

So Many PPl Are About To Flood Your Patreon Just So They Can F around With Your Big Robot. Thats A Brilliant Way To Make Cash NGl.

[@GregsStoneYard]

Great progress. I do think it has limited usefulness, given the relative affordability of modern used robots. Just watched two unused Kuka's, fully outfitted with 9kw (may have been 11kw) spindles, tool changer, enclosure... sell for $33k at auction. They were purchased new in 2021 or 2022 and never put into production. But I hope to proven wrong.

[@Citronitroify]

But where is the fun in that?

[@GregsStoneYard]

@@CitronitroifyTrue!

[@ExcessiveOverkill]

I don’t expect it to really be a replacement for where you could get a normal robot and controller for cheap. I want it mostly to fix the limitations of what the really old controllers can do. And also run stuff besides robots

[@GregsStoneYard]

@@ExcessiveOverkill Have you seen what Brian Brocken has done with an ABB SC4 controller? He basically made a drip feed for the program that's running on the controller. So you get to run long complex programs, but keep all the existing robot drives/controller/pendant. Your solution will require replacing all that will it not? While doable for someone with your knowledge, I'm thinking there may be only a handful of others who will take on the task. Anyway I'm not trying to be a jerk about it, I do hope you are successful. It's a monumental amount of work and impressive.

[@ExcessiveOverkill]

Yea I’ve seen the drip feeds, unfortunately by default most Fanuc controllers don’t support external program input so they’re kinda useless. Anyway it is pretty dang complicated rn, I’m hoping one it’s actually functional I can get it closer to something others could use without knowing all of the super low level stuff. It’s going to take years for that tho, but it’s a good side project for me.

[@johnsimons92]

I hate it when people re-use standard connectors for non-standard uses. Why not pick any of the other b2b connectors that can't be mistaken? There are so many card edge connectors and you picked the only one that has a standard use.

[@ExcessiveOverkill]

You pick standard ones because they’re cheap and a ton of manufacturers make them so they’re usually in stock. Lots of things use pcie connectors that aren’t pcie because they’re common. I did look at other more industrial connectors and they just cost 2-10x more and are harder to get

[@johnsimons92]

​@@ExcessiveOverkillThat's simply incorrect. There are many other edge connectors that are dirt cheap. E.g. G630H9811310EU or FX27-40S-0.8SV or any of the plethora other options. It would make for a much more professional board if you use purpose suiting connectors, aside from the risk somebody might plug something in a board it doesn't belong.

[@ExcessiveOverkill]

Theres no reason to complicate designs by using extra nonstandard connectors unless needed. They function perfectly fine. If I ever would switch it would be for something industrial like the DIN 48612 connectors which are way nicer than edge connectors. There’s not really a risk of someone plugging in the wrong thing with one of these connectors since the whole thing is supposed to be in a housing that only fits certain size boards. My actual concern was using rj45 connectors to run the drives since they are NOT Ethernet compatible. However both the connectors and the cables are so common and cheap it didn’t make sense switch to a custom solution solely to make it harder for someone to break something. Especially considering non-Ethernet rj45 connectors exist everywhere in the industrial world.

[@AdaptivePhenix]

@@johnsimons92 Obviously, you have never had to contend with a dead production line where there are dozens of hourly employees standing around with nothing to do. I can't remember the number of times that Radio Shack bailed me out on a Sunday morning in an automotive plant in the middle of nowhere. It's all well and good these engineers, designing product based on the latest/greatest, sexy components but I'd like to drag their a$$es to the plant-floor where going-home is simply not an option until that plant is up-and-running.

[@johnsimons92]

@@AdaptivePhenix Wrong on all accounts. And your designers not being able to properly source components doesn't mean you should just use whatever. We all laugh when we hear Russia has decided to give chips in washing machines a second life by putting them in their weapons. This is is the same, where available you use what is appropriate based on your industry standard. This in fact fixes the issue you described because everyone uses the same components so they will be produced at volume.

[@mvpolo2k3]

Galil DMC......just saying

[@AdaptivePhenix]

Command processing is dog slow. Their flagship ($$$) takes 11usec to process a single command. I was their first-ever machine-tool customer, back in the mid-80's.

[@adamruck]

I find the title to be a bit misleading since it says "industrial" , but you are making a home grown solution with no certifications and missing many design features that would all be required for actual industrial use. Is this a robot controller, absolutely. For industry.... not really.

[@ExcessiveOverkill]

Industrial is describing the robot, not the controller. It’s much more for super customized or developmental stuff, not a normal controller replacement. Sort of like what LinuxCNC is

[@AdaptivePhenix]

You see a plant floor recently? Windows-based PCs are everywhere. I keep coming across machinery that still run on my "home-grown" PC-based DOS controls from the early 90s before "industrial PCs" were a thing. Thanks to being a common architecture, they have been able to keep them running. Today, I replace Siemens controls that are

[@8starsAND]

Ok, who takes the responsibility if hardware fails and someone’s get killed??? You???

[@deweydecimal1446]

Seriously, go away idiot😂

[@AdaptivePhenix]

Real-world: You equip your machine with all kinds of safety-interlocks, laser-scanners, etc. Later, you discover that the production department decided that too much time was being lost and they instruct maintenance to bypass the safeties. "Safety officers" haven't a clue what they're looking at. If there are bright-yellow devices in-place and warning signs, it's "all good".