gen 4/5 electronics

Ryan Mulligan Wrote:
-------------------------------------------------------
> Hi, this is Ryan from Pololu.
>
> The heat issues you mention with the A4983 are not
> an issue with the board, but an issue with the way
> people are driving it. Some people were stepping
> the board with a very low duty cycle with too
> little current limiting. This leads to current
> spikes of over the 2 A per coil spec. of the
> A4983. If you stay within the specifications there
> will be no heat problems, and if you want to
> exceed them you can use a heat sink.
>
> - Ryan Mulligan

Hello :-)

Thank you! Is always a pleasure to have help from the seller/producer

So I guess we should say instead that board/ic is not good for our extruder application.

Ryan, maybe you can put some mention to this question on driver page?

Thanks.

---

---
New cutting edge RepRap electronics, ARM 32 bits @ 100MHz runs RepRap @ 725mm/s:

[www.3dprinting-r2c2.com]

Rubbish, look at the datasheet for the amount of heatsinking required for 2A.

---

[www.hydraraptor.blogspot.com]

Yep. The way that current ratings are worked out is whichever current can flow when the device's thermal pad is connected to an active heatsink holding it at 25 celsius while the silicon junction is at maximum temperature (150 or 175 celsius) due to the thermal resistance between junction and heatsink.

Given the thermal resistance from junction to ambient with no heatsink, and the Rds(on) of the internal fets, I calculated that with no heatsink the chip can pass about half an amp, which aligns perfectly with my experimental results.

the upshot of this is that expecting the chip to pass its maximum rated current with no heatsink in still air is extremely naive

---

-----------------------------------------------
Wooden Mendel
Teacup Firmware

Hello,

This is Jan from Pololu. I was just alerted of this thread, and Ryan's post is indeed incorrect in that 2A continuous without any heatsinking is unrealistic. However, I think some of the claims about our carrier underestimate the capability of this little driver. With the total internal resistance per output pair on the order of 0.75 ohms, at a 1A current per coil, the power dissipated would be on the order of 1.5W, which is feasible on this board (and both coils are typically not simultaneously on at full current). Also, because the input voltage to the driver can be a lot higher than the coil voltage, the measured current on the power supply can be lower than the coil current (the driver and coil basically act like a switching step-down power supply).

- Jan

Jan,
The fact that the supply current falls when you increase the voltage does not reduce the heat in the chip. If you are driving say 1A then for some period of time 1A is coming from the supply, the rest of the time it is re-circulating via the transistors in the chip, so the chip is always passing 1A with the associated I^2R losses.

---

[www.hydraraptor.blogspot.com]

The point is that while that current is recirculating, power is not being drawn from the supply, so the power supply current is lower than the actual coil current. Most of the customers I have talked to about this have measured power by looking at their power supply, which leads them to report overheating at lower currents than they are actually achieving. (If the voltage is very high relative to what the motor needs to get the set current, the duty cycle will be quite low, which will also lead to significant differences between average and RMS currents.)

- Jan

Oh I see. No you can't set the motor current by monitoring the supply. You have to work it out by measuring vRef and dividing it by 8 times sense resistor value.

I think Triffid_Hunter reported only 0.5A per coil without a heatsink and I would think he would know how to measure it.

---

[www.hydraraptor.blogspot.com]

Does anyone know the reason for keeping the extruder/motherboard split for the Gen4 electronics? I understand the original reason for the split (limited pin outs, memory,etc) but it seems to me that the extruder board does very little work, and as the Mega has so many pins available, it could run the extruder directly. It would also remove the timing/synchronisation issues between the two controllers?

Or am I missing something fundamental?

DaveR (Renoir)

It is so you can have multiple extruders.

---

[www.hydraraptor.blogspot.com]

You can drive at least 2 extruders with the Arduino Mega alone.

I think we should stop calling the electronics "Generation n". There are currently a couple of working electronic derivations that are going in different directions. It doesn't matter if it utilizes multiple micro-controllers, I2C, RS485 or none at all.

All that matters is that it gets the job done.

---

-------------
FiveD Firmware fork (supports Arduino Mega): [github.com]
reprapsource.com

Several things to note about stepper motors verses gear motors.
Stepper motors:
Advantages

[1] Precise to 1/8 of rotation a little better with super micro stepping
[2] Allegro chips makes them easy to program and run

Disadvantages

[1] Moves in digital chunks leads to jagged edges
[2] Cannot control the current very well it is either all or something less
[3] They get hot if you run them wrong which ruins the magnets

Gear motors
Advantages

[1] Cheap and reliable
[2] With an encoder can be precise to 1/360° or even better
[3] Smooth response no jaggies
[4] Easy to control and monitor the current. Only direction and voltage.

Disadvantages

[1] Need an encoder and they are hard to build and mount
[2] Backlash have to be run in one direction only or have to account for the backlash

For me the jaggies seen on edges run on anything using a stepper is too much from me in precision needed output. The Mendel is such a requirement as I want smooth edges when I define a circle.

Backlash can be easily handled by simply running in one direction. Machines know of this and compensate. The encoders if one can deal with them like the AS5040 are wonderful devices capable of 1° of accuracy. The even have a analog output capability that could get even tighter response.

So with all of that diatribe, if I use simple outputs with simple motors with encoders, I do not need much computing headroom. As such I am happy with the Mega chips from Arduino and do not desire to learn yet another language. If these suppliers would standardize on say Java that does not require special licensing to commercialize the product then I would say go for the highest need. I like simple USB inputs, I also do not use and abhor microchip based stepper motor drives. This is the only place I see you need ram space for the tables. They do not nor will they ever work. The Allegra chips in hardware are the last word on this for me, but still it is the aspect of the jagged edges that leads me to use gear motors like a Dayton 12vdc 1.3A 270:1 ready to go with a mounted AS5040 hall effect encoder. Simple and done.

Reprapped objects don't have jagged edges. The steps are 0.1mm (or 0.025mm if you use microstepping). The discrete steps get smoothed out by mechanical inertia and the fact that the filament width is about five times bigger and has a minimum bend radius.

The reason circles may have stepped edges is because the STL file models them with triangles. As long as you set the STL conversion accuracy correctly you can get smooth circles if your machine can keep up with lots of small line segments.

If you have backlash you cannot fix it in software like you can on a milling machine. The time you spend moving over the dead band will lead to an excess of filament in that area.

DC motors maybe cheap but they are not reliable, the brushes wear out. Steppers last a lot longer as the only parts that have any wear are the ball bearings.

---

[www.hydraraptor.blogspot.com]

Actually the Dayton DC gear motors are permanent magnet models the 12vdc ones do not have brushes. Brushes are dead in the water exciter types. Everyone now uses rare earth magnet motors that are brushless at least at Dayton that are 12 volt models. You are thinking of the 90 VDC models which are terrible.

As far as jaggies, talk to any machinist and they do not run stepper motors. Stepper motors are the cheap alternative to DC motors running an encoder. But with the AS5040 and a cheap rare earth magnet we now have the capability of matching the expensive encoder technology. Lasers run stepper motors and until you run one you will not know this. The circles are terrible approximations of vector driven circles. The problem is the machine code standard. All circles are based on XY grids. But a vector is a line and all that has to happen is to ignore the machine code that is old and archaic and define a circle based on PI. If you drive two DC permanent magnet brushless motors properly based on PI you get perfect circles. You cannot do that with a stepper. But that takes mathematics maybe that too is the problem. The same issue is with Mendel and yes triangles but again it takes math to define a circle. All the triangles have done is add a third dimension to the XY grid and you still approximate a circle. A pascal bet if there ever was one. Triangles are fine for representing computer games. It does not work except for crude parts with machining. If you want to build a gear, it is first machining on an index mill, then it goes to laser milling then it goes to 3D printing. The mill gives the best shape, the laser next and last is the 3D printer. Laser milling is good but not great under the scope, the 3D printer is worse.

DC motors are very reliable I do not share your opinion at all and yes you can buy them with ball bearings that truly is no big deal at all. Expensive milling machines use standard DC motors that are connected to encoders, not stepper motors for good reason.

But what about the backlash compensation causing a build-up of filament?

Interestingly, permanent magnet brushless DC motors are more or less the same thing as stepper motors, they just have far fewer poles. So you can get similar results with a 3 pole motor BLDC (which is really a synchronous AC motor with some fancy drive electronics that take DC) and a 30:1 gear reduction, or a simple 200 step stepper and simple open loop electronics.

Another factor in rapid prototyping is speed - I can get 50 mm/s out of some $13 steppers with 0.1 mm resolution; I'd go faster, but my cheap steppers start to overheat. Plus, in a reprap, you can't just go one direction to compensate for backlash - we need to be able to move in all directions, without any pauses in motion, which would be problematic with a gear train.

Switching to DC motors and encoders is probably the way to go - look at any modern inkjet printer. However it adds another order of magnitude of complexity to the electronics and firmware. I'm not convinced we'll see a big cost reduction or accuracy improvement just yet.

On the topic of processor, would I be out of line to suggest the mbed board? It has an LPC1768 (Cortex M3) at 100MHz for $60. Pretty inexpensive for such a beast if you ask me. Yes, some people see the online IDE as a negative - but you can easily develop in the usual offline way (gcc, et cetera) if you wish. Plus, it may actually make firmware development more accessible since folks wouldn't have to set up a toolchain if they didn't want to. The community stuff is pretty neat also with the code-sharing features and such.

It's also a standard platform that will be around for some time to come, which might otherwise be a concern with "insert random cut-rate dev board here". The board is also set up as a standard DIP layout which makes it easy for prototyping. A carrier board could easily be designed with all the drivers and such, into which plugs the mbed module.

I'm obviously new to reprap development (hi, btw!), but I thought that I'd throw it out there...

Edited 2 time(s). Last edit at 04/26/2010 03:26PM by plasmator.

plasmastor

A couple of us are working on an LPC1768 based solution.

[forums.reprap.org]

We have decided to start with an MBED module (even though it has some limitations) for an initial prototype. We are in the process of building the prototype hardware that goes around the MBED module so there isn't any significant progress to report yet.

I found the Gen. 4 Makerbot Electronics kit on ebay for 270 USD, that will save you 100 bucks.
And for that price it is very interesting.