Please hear my request oh hardware designer god's!

[Flateric]

I have come to believe that the pace of controllers and reprap electronics hardware is not keeping pace with the printers themselves.


What I find more alarming is that rather than newer, better faster, compatible controller boards are not the primary focus of the hardware developers. Rather they seem to be focusing on budget and cheaper and simpler but little improved controller electronics.

I do not think I am that small a minority here when I ask for and seek not cheaper, but better. I am willing to, actually desperately wanting to spend a little more on a quality controller electronics package that has improved speed, features and connectivity and yet still offer modularity and subsystem replaceability. Blow a new high speed stepper driver, no problem just unplug the old one and plug in a new one (ramps style), want to easily and effortlessly hook up all your fans and extruders with connectors and not have to mess with pin assignments (Rambo style), fast cpu and communications with PC and larger buffer (arm style), wireless module (not sure who style), repetier and marlin firmware compatible (ok I know stretching it here due to many conflicting reasons but it is listed only because of my next point here >>>>>>), Well implemented Delta support in firmware (Ardruno style).

I would gladly pay double the price of a Rambo for such hardware, if it were to exist and was robust. Hell I would pay upwards of $450 for such hardware if it were trouble free and never a concern during printer use and operation.

Smooth fast and expensive? I'll take it. But it seems, cheap, problematic and none modular with a primary focus on cheap seems to be the driving direction of the controllers on market now and upcoming.

Make me a an expensive fast reliable modular controller.

I bet I am far from the only one who would buy one!

[geneb]

There's actually a number of people working on getting controller designs based on ARM based computers up and running. I know of at least one that's based on the Beagle Bone Black.

g.

[Flateric]

Please point me in the correct direction to follow their efforts. And we are talking delta's here correct?

[geneb]

Yes. I'd join the "deltabot" list on google groups, that's where I first saw it discussed. That's also where Johann first posed video of his auto-level feature on his Rostock (or Kossel, not sure which).

g.

[Polygonhell]

I have one I'm working on based on the LPC1768, but it's not really intended to be a commercial venture.
There is the Smoothie Board based on the LPC1769, there is a shield for the Beagle Bone, there are moves to port various firmwares and shields to the Arm based arduinos.

When you step up from the existing Atmel controllers, you end up having to make a choice, either stick with faster uControllers in the 100-400Mhz range with limited RAM, but raw hardware access, or go to something like a Beagle bone which is much faster, and have a large RAM pool, but where half the features are undocumented and unavailable unless you use binary linux drivers. The issue is that Linux is very far from ideal for use as a real time controller.

The reason I'm messing with it is that I have a fundamental objection to using USB as the interface, it's just a bad idea for several reasons, and the way most firmware implements comms over USB makes it worse. I'm also unconvinced that a host sending GCode to a controller is a good idea, but I can see the value there.

[Flateric]

I have met with resistance from other users with regards to this suggestion I am about to make, but after running a CNC gantry router that implemented my suggestion I am convinced it holds merit especially for application to 3D printers.

My suggestion is this.

Why have we not seen a servo based reprap 3D printer yet. Servo's are more expensive in both controllers and hardware implementation I'll give ya that. But servos have a number of VERY big pluses for use in the 3D printer design.

They are smooth as it gets. A stepper system can't hope to ever come close to this level of silky smooth. The main reason any quality vinyl cutter uses servos is partially because of how SMOOTH and also because of the SPEED and finally because of how RESPONSIVE and ACCURATE servos behave. Especially when moving relatively low weighted loads such as in reprap's and delta printers.

So the pro's are far more than the con's IMO.

SMOOTH, SPEED, RESPONSIVE, ACCURATE <VS> COSTLY.

I have a chinese made vinyl cutter that is relatively new, 3 years tops, stepper driven, and in it's cuts if you look closely it is apparent that the steppers create a "machine" pattern that immediately reminds me of the exact same pattern we see with our reprap's when printing any type of smooth arc, cylinder or even some flat linear surfaces offset from the printers primary movement access.

I also have a 14 year old Roland Camm-1 vinyl cutter that is 100% servo based. It cuts easily 10x faster than the chinese unit (no exaggeration I assure you, perhaps faster than 10x) It also cuts perfect arcs that have absolutely zero stepping or hint that a machine made the cut. Long sweeping arcs are ARC's not little glued together line segments. Even though however, the software feeding the printer (any HPGL program) can and does often feed thousands of little lines to create the arcs. Small tight cylinders have no moire "machine" pattern, you cannot see the effect of the servo drive system visually anywhere. The same print with the chinese stepper driven printer, they are not always apparent until you look closely. It prints acceptably, but side by side with the servo system it then looks awful.

I can't help but think a servo based print in 3D next to a stepper based print in 3D would have a similar result.

I wonder how difficult it would be to design and make a 3D reprap servo printer that is run through Mach3. I have never run a 3D printer with mach3, only subtractive machines like routers and lathes.

Is it a nightmare to get a reprap controlled through Mach3? Mach3 is not difficult to configure to run servo based machines. And the printer gcode fed into mach3 would not have a clue that it is going to servos or steppers. Perhaps this justifies a test of some kind?

Can a delta be run through mach3?

[Eric]

Just so we don't confuse people lets define terms.

In a stepper motor solution, you tell the motor how far to move and make the assumption that it actually did so. Your design must run the motor well under it's limits for fear of missing steps. Missed steps in an axis motor will usually ruin a print, although it might go unnoticed on the extruder.

As for servo's, it's an abused word...you'll often find stepper motors being called servos, especially in the hobby industry. In a true servo motor solution, there's an encoder on the motor that gives feedback on the exact position of the motor shaft. That means you run the motor, up to its limits if you want, in the right direction until it gets to the right position. It's also very modular, although the modules aren't what you call cheap.

Certainly you can get servo-based solutions off the shelf...they're already there for the CNC industry. But the pricing for the whole package (motors, controllers, power supplies) is at least an order of magnitude higher, so far as I can tell. Which would be fine for commercial production...but it's a killer in the hobbyist market. How many Rostock kits do you think would sell if the price was more than doubled?

[Flateric]

But we are not talking about rostock max kits here. And I am very clear on the increase in cost as I state, repeatedly it my posts.

I think it well within reason to say that a servo based solution would run under 2.5 times (perhaps just under) the cost of a stepper based system. I am not referring to stepper with encoders in a closed loop system. I am talking about brushed DC servo motors in a closed loop full feedback system. Such as something from Maxom Motor or a mish mash of Gecko servo drivers with matching motor encoders. So when I refer to servo's I mean servos, such as the referenced Roland camm-1 cutter. I have a servo driven CNC metal cutting mill in my garage and a stepper driven cnc gantry in my basement, both my own design, the basement machine is my own design, right down to the stepper drivers, controller/computer interface. Granted it's old school PRT port driven antique hardware. But it is what I knew at the time and so I went with it.

I'm not by any means any kind of expert in the reprap field, very new to it in fact, learning quick enjoying myself and meeting alot of brgith and friendly chaps such as yourself, Poly and the rest of the guys here in the forums.

My frustration is that I have my machine dialed in at such a point that it is capable of outperforming nthe controller cards currently on the market. Also I have hit a brick wall with the speed and performance VS quality of print situation. I find this frustrating. As soon as I am able to go a little further in speed and quality I am let down by over heating, buffer under-runs, stepper skipping, stepper resolution limits VS microstepping VS stepper stalls VS stepper curent limits that lead back to lower end controllers that are not modular, burn out 1 component and replace the entire board which ends up being expensive anyways when you are on your ### (Higher than I care to admit) LOL GASP.......

See where I am coming from now?

If I can wrap my head around the servo driven reprap situation through mach 3 with a delta I will let you know here in the forums if they have any merit in the reprap world.

I'll let you in on a little secret. I more than likely have everything I need already to do this little project with what I have salvaged from high end cnc dental mills that are made by Cercon. This is where I have gotten the MAXON servo drivers, controllers, and motors along with bonus items like extra precision linear slides etc. These dental mills are rated at accuracy down to 4 microns. Granted they originally sold for hundreds of thousands of dollors, but that explains why the dentist is always such an expensive visit, HA!

My goal is not to build or sell a new Rostock max, I seriously doubt I could even come close to the price point and performance that the seeme guys have got the product too, I also have no interest in doing this. I do however wish to up the standards that I print at for customers (and myself) while doing so at perhaps unheard of speed. Although I appear to have hit the limit that can be extruded with current plastics and plastic shear when the nozzle is at speed.

Here is a fairly old video with some early tests I was doing. Keeping in mind that the purpose of this test was not to produce superb quality of print, but to produce a functional print at a very high rate. This was also with stock arms, back before I saw the berry bot, tried to reproduce them on the rostock, failed, then Xnaron took berry`s inspiration, some of my ideas for implementation and finally hit success. This is exactly what I enjoy about this field. The collaboration, the inspiration and the furthering of the possibilities we can do with these printers. I could no doubt easily out perform the test from this video with magnetic arms we have refined now, new all metal hotend such as the E3D, my stepper damper experiments have improved the audio level and the steppers run smoother and faster with them too, combine all these things and we are really getting somewhere. It``s what drives me and it gives me a good feeling when I have had a small part in contributing to the now general knowledge on improving these machines and what can be done, the quality, etc.

http://www.youtube.com/watch?v=4QBdDmWD5kc

When I started my 3D and compoter graphics journey they didn`t even know what to do with 3D CGI, there was certainly no money in it, it was a bizzare basement hobby that unpopular geeks did. Slowly people found ways of improving and making and speeding up the graphics. The first 3D graphics I made were far from real time and the machine could only manage to juggle around 1000 polygons per scene, not in real time either.

The end of my work career was with Nintendo in senior roles making 3D games I can almost certain promise you have played, millions of polys juggled in real time, fully texted and lit beyond anything anyone dreamed possible for the cost of next to nothing, AND IT RUNS ON YOUR CELL PHONE! Anyone would have called you insane or told you not in this life time.

I look at the reprap community and the 3D printer development cycle to be in a comparable state to what a Commodore 64 was with a data tape drive. Looking back I can hardly believe that everyone thought it amazing and stunning that you could load a 64Kb game off a cassette tape to play (wizard of wor anyone, or Zork!) and it loaded so fast, only 20-25 mins to wait for loading times. Now we look back and think it stupid and lame. Back then IBM had machine with massive 5 Meg hard drives. Ya, ok they were expensive, many order of magnatudes more expense (I used to seel them as a kid).

This is where we are right now in the 3D printing world, at the commodore 64 generational phase of development, where your average joe like you and me and anyone can still contribute and advance the development of the scene with no multi-mega dollor investments. Just 5-8 years ago a 3D printer that output weak breakable wax objects at about the same resolution we print at now cost no less tha 100-200k and running such a printer was a huge investment simply to turn on the power to it and cycle it up for a print (I ran one of these in Germany for many years, turning a printer like those on cost around 200 bucks just to prime it and have it ready to run and go, thats why you tried not to turn them off either)

Can you imagine where we will be 5-8 to 15 years from now in this geeky makergeek niche market? Think about it, no really think about it. You, me, everyone here needs to look ahead and have some vision. Think about if you could go back to the commodore 64 stage of computers with what you know now........you could rule the world, or at the very least have a really good fun ride.

Mark my words, I'm old enough now to learn from past trends and I tell you we are in the cusp of a very large on right here right now so grab your cassette tape drive of today (a reprap printer) or your 5 meg hard drive of today (a delta type reprap) and hang on!

So I rambled a bit, my point is however valid I believe.

Now back to servo driven reprap printing.....LOL

[Polygonhell]

You can run a printer from Mach 3 easily enough, it was the primary selling point of the SeeMeCNC H1, you need external temperature control, to do it on a Rostock you'd have to write a preprocessor to subdivide the paths into short line segments which isn't actually very hard. You could then use any drivers or motors you want.

As to servos I remain largely unconvinced with their use on printers without significantly improving the mechanical designs, a single step while uStepping is in the 0.01mm range, there are many mechanical factors that dwarf that.
The positional feedback is overstated even in the CNC world, often the failure cases are because the move requires exceeding the motors maximum available torque, while servos will increase the power to the motor as the motor is slowed below the requested speed, once that maximum is reached it will stop just like the stepper.
Servos do have a significant speed advantage over steppers, but in most cases printers are limited by the extruder speeds.

My issues with the existing electronics are the use of USB as a connection, the lack of CPU cycles for the kinematics on more complex designs like the Rostock. I also think much of the software stack is just about "good enough", which is unfortunately typical of open source solutions.

[Flateric]

I have the external temp controllers still I ran when I had blown my rambo(s) thermistor ports.

I totally agree with you on the number of cycles. That is what kicked off this whole hardware designer gods rant thread of mine here.

But I thin I may have overcome the extruder issue to a small degree anyways. It's unconventional and I don't know anyone else doing it this way at all, but I'll PM ya tomorow, I am beat tonight.

[626Pilot]

I have met with resistance from other users with regards to this suggestion I am about to make, but after running a CNC gantry router that implemented my suggestion I am convinced it holds merit especially for application to 3D printers.

My suggestion is this.

Why have we not seen a servo based reprap 3D printer yet.

http://www.kickstarter.com/projects/173 ... e?ref=live

It failed to secure funding (raised a little over 10K pounds out of 50K), but was servo-based.

Repetier will compile on Arduino Due, which is ARM-based. That's probably your shortest path to most of what you want, if there's a board for it. If you desire more accuracy, consider your steppers. 1.8 degrees per step, right? I've seen steppers all the way down to 0.25 degrees per step. I'm betting these are pretty expensive. However, I think they're still cheaper than the time+money it would take to convert the Rostock to servos... which could be done... but there's no firmware.

[Flateric]

I appreciate your suggestion, gonna look into that right now.

I have some .9 degree steppers on the way to experiment with. Perhaps the combo of these will satisfy me for now.

Do you have a link to the Due?

[Polygonhell]

The standard boards like RAMPS won't work on a DUE, the DUE runs on 3.3V logic, and RAMPS is designed for the older Arduinos which are all 5V logic. Alot of the board will work, but you'll have issue with at least the temperature control, both the thermistor inputs and the Mosfet outputs. The endstops should work if you use mechanical ones as should the motor drivers.
There was at least one person looking at building a bridge board for Ramps to allow it to work on the Due.

[626Pilot]

There are a few options.

If you want raw speed, I've got an ATMega644 sitting on my desk that's been running at 20MHz since sometime in May without crashing. The RAMBo and most other Arduino stuff runs at 16MHz. The 644 has been shown to sustain clock speeds up to 30MHz without losing stability. So, if almost twice as fast as stock is enough, that would be one option. However, I'm not sure if the delta version of Repetier will fit in the 644's 64K of flash memory. You would probably have to #define out all the LCD controller stuff to get it to fit. If that DOES work, you can build a Sanguinololu clone (ATMega644 + Pololu stepper drivers) with as many ports for controlling fans and LEDs and whatever else you care to include.

Another option would be to hot air the 16MHz crystal off your RAMBo and replace it with 20MHz or faster. This will involve compiling a bootloader with the correct speed selected and editing boards.txt under Arduino, then flashing the bootloader somehow (perhaps over the programming header on the board.) It's kind of a pain, but so are all the options. The CPU used by RAMBo is similar to the 644 so it's probably about the same, stability wise. If you're good with designing PCBs and not afraid of SMT parts, you could design a RAMBo with whatever extra ports you want. They've already done almost all the work in terms of diagrams, etc.

You can try your hand at RAMPS-FD, which is RAMPS for Due: http://reprap.org/wiki/RAMPS-FD - It's still experimental. So is Repetier's support for Due. The latest update to the wiki page on the interface board is from March. I don't know if anything else has happened since then.

If I was looking for another solution, I'd think about setting up an expansion board for a Raspberry Pi with a bunch of stepper drivers and MOSFETs and then run LinuxCNC on the Pi. According to a Youtube video, LinuxCNC can run delta robots. One of the LinuxCNC developers actually got a Rostock Max and is working on the code. https://plus.google.com/115008887121578 ... BsGCzSrPtv

[kbob]

Two comments.

1) We have digital calipers. They have linear position sensors that have a nominal resolution of 20 microns. (Opinions vary on whether they actually achieve that resolution. (-: ) The cost is low -- basically a single-sided PCB as long as the distance you want to track. The complete caliper is something like $15 for 150 mm, including the controller, display, and some machined steel pieces.

The biggest obstacle is that the current controller chips are very slow, because nobody needs sub-millisecond response on calipers. New controller hardware needed.

I bring it up because a linear position sensor would give a closed loop all the way to the carriage position, not just the motor shaft position as in a typical servo.

2) I agree with Polygon's dichotomy: you can have a hard real time microcontroller with limited memory or a proprietary, non-real-time Linux box. There is a huge opportunity for a chip manufacturer to exhume its head from its rectum and build a completely open source system that the geekiverse can put a real-time OS on. Unfortunately, that means not using any designs from Intel or ARM.

I have been working on a solution in my totally insufficient spare time that combines a Raspberry Pi and an Arduino compatible connected by USB using a protocol that is much lower level than G-code. The Arduino does all the real time work, and the Pi does all the heavy calculation. My target is a laser cutter, not a printer, and I don't have a complete system working yet, but all the calculations, simulations, and tests indicate that it will outperform a standalone Arduino. The version in my public GitHub repository is rather out of date, sorry. (Link is suitable for software geeks only.)

[Gimbal]

What about this.

http://smoothieware.org/smoothieboard

[626Pilot]

I think this might be the real deal. http://www.panucatt.com/Azteeg_X3_repra ... ax33dp.htm

It's got 1/32nd stepping, support for several extruders, and extra MOSFETs just for driving LEDs and fans. It is pin-compatible with a RAMPS board and will run Marlin and Repetier. On top of that, it's a lot cheaper: $109.

I think I want one.

[kbob]

I have an Azteeg X3, though Roy has updated it since I got mine. I use it in my laser cutter.

It is a fine board, but it is basically the same class as RAMPs, RAMBo, Rumba, etc. ATmega 2560, Pololu-compatible stepper drivers, Arduino compatible, runs Marlin, Repetier, etc.

It does not offer a quantum increase in processing power.

[Gimbal]

Well Smooth'i dos, STM32 is about 170 times faster than AVR, but whats the fuzz do we realy need that speed, i'm more conserned about the communication over usb, maby move over to usb 3.0

[626Pilot]

Well Smooth'i dos, STM32 is about 170 times faster than AVR, but whats the fuzz do we realy need that speed, i'm more conserned about the communication over usb, maby move over to usb 3.0

USB is capable of transmitting data far more quickly than a board with as much RAM as the Smoothi will ever be able to consume it. I run at 115200 and I've never seen the buffer empty. Other people say it happens to them, but consider the AVR chip has RAM limitations and the buffer is something like 20 entries long. With a "real" CPU you could make the buffer dozens of times larger without running out of memory. The slowest USB 1.0 speed is 1.5 megabits/sec, or about 160K/sec. A 30-character G-code instruction is less than 1,000th of that.

It also looks like that board works directly with Rostock machines, so it's probably a good choice.

I have an Azteeg X3, though Roy has updated it since I got mine. I use it in my laser cutter.

It is a fine board, but it is basically the same class as RAMPs, RAMBo, Rumba, etc. ATmega 2560, Pololu-compatible stepper drivers, Arduino compatible, runs Marlin, Repetier, etc.

It does not offer a quantum increase in processing power.

What kind of laser cutter do you have?

[Polygonhell]

The issues with USB are two fold, the biggest of the issues is the latency, USB transmits data in packets on a clock pulse, and for USB 1.0 that clock is only 8Hz (for USB 2 it's 128Hz), the existing firmwares write data to it in the worst possible way, basically they write a line and wait for an ack, so you wait on the clock at both ends. This is the primary reason copying files to the SD card over the USB interface takes so long.
The secondary issue is it's intolerance to noise, it's very susceptible to noise from the motors or high voltage sources, again the existing firmwares do the minimum to work around this, by using Serial over USB and appending a check sum, rather that solving the issue with the base protocol.
Ethernet by comparison is extremely noise tolerant, by virtue of the twisted pairs, has base level protocols that deal with error retransmission (TCP) and has no issue with latency.
In the CNC world, there is a popular motion control device called a smoothstepper, there were many complaints about noise causing issues until it moved from USB to Ethernet.

[kbob]

I started building a Buildlog.net 2.x laser cutter last year. It is currently waiting for me to write more firmware for it.

Site for the laser cutter.

My build log.

[kbob]

My frustration is that I have my machine dialed in at such a point that it is capable of outperforming the controller cards currently on the market.

What speeds are you able to print at now? Give me an idea what I could aim for someday. (-:

[Flateric]

[youtube]http://www.youtube.com/watch?v=4QBdDmWD5kc[/youtube]

This is an older clip of mine, before magnetic arms and my e3d hotend and I would say I am not only smoother now by a long ways, but also faster yet.

If I sustain a longer print at these speeds it starts causing me issues.

[Gimbal]

"How to improve your USB transfer speed by increasing the maximum transfer size of 64KB block to 2MB in the Usbstor.sys driver"
Well maby not 2MB but 255KB and expand ringbuffer size to 255 and force flush at 192KB

http://en.ocworkbench.com/tech/how-to-i ... ys-driver/

Haven't have the time to test it, i'm on vacation