MatchboxARM

Guys,

There's an interesting kickstarter called MatchboxARM:
[www.kickstarter.com]

In essence, it's ARM Cortex-M3 chip running 72MHz with 64Kb flash on a small board cheap enough not only for prototyping, but also for embedding it into the final project. The board would have USB port and would run on its own without any extra components. One needs to pledge 15 GBP (it's UK-based) to get a single unit, which is cheaper than Arduino Mega.

Couple of things spring into my mind. Firstly, it's a great project, so please back it, and, secondly, looking at the spec, it seems MatchboxARM could be used to build simpler and cheaper electronics for a RepRap. I think Gen7 could be easily modified to host MatchboxARM instead of AVR micro.

What are your thoughts?

Traumflug is already working on a Cortex M0 version of Gen7. Great minds think alike I guess.

Looks extremely similar to a Teensy3: [www.pjrc.com]

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

That looks like a nice little board, but the Teensy 3.0 has been around for a while and is even slightly cheaper [www.pjrc.com]

But I think the problem is really not the hardware, it's about the software support and the "mindshare". AVR works, and people know and understand it, so there is little incentive for change.

Traumflug has done a Cortex M0 version, and I have ported some Teacup firmware for it, but I think there has been zero interest in the firmware. Not even to say, "that was a load of rubbish"

The fact is that all firmwares require some level of porting to run on an ARM, and then you leave the friendly comfort zone of the Arduino IDE (with the exception of Due). In addition, all ARM chips have different peripheral registers, so a port for STM32 will not run on an LPC chip. Basically you will need a lot of active developers to support the platform.

The only ARM board getting any traction appears to be the Smoothieboard, with quite a lot of development activity. The Smoothieboard is quite cheap, it's open hardware, and there are variants being planned. Smoothieware is rather large, but a LPC1769 is only $5-$6 in small quantities. The LPC1769 is also one of the mbed platforms, which has decent software support.

Thanks guys, interesting perspective on the limitations, but what about benefits? Can ARM bring anything to the table in terms of cheaper/better electronics for RepRap? Or it's just for variety sake?

Personally I think ARM is all round better, faster and cheaper, but I have been working with ARM chips since 1999 so they are well inside my comfort zone. I think it's amazing how many functions have been crammed into AVR firmwares, but if you want to do things like ethernet, run a web server, run non-cartesian geometries, ARM chips provide a lot more scope for expansion. You could also do things like run a 320x240 touch panel TFT display instead of 20x4 LCD.

I have never run an AVR based controller, I have only used ARM based ones, so I am fully committed to the idea

Good point, @bobc. Not cheaper, though, since an ATmega isn't exactly expensive either. And ARMs have only 3.3V signals which isn't sufficient to drive commodity MOSFETs or TB6560 or THB7128 stepper drivers directly, requiring more components.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

That's a fair point.

It's a difficult to compare costs, since it's a bit apples and pears, but the cheapest easily available ARM is probably the Kinetis KL25, 128Kb Flash for £2.00. ATMEGA128 is £5.00 on same web site (http://uk.farnell.com/). It's true that you can't drop a 3.3V part into a 5V circuit, it either needs extra components or different component selection.

I guess that overall it comes out pretty even I think I would still argue you get better value with ARM, 32 bits at 48Mhz vs 8 bits/20MHz.

@bobc,
Having recently gone through microcontroller selection at 1k quantities, I found Atmel was still coming out on top as far as low cost goes. ATSAM, that is.

When I added requirements for: USB, PWM, at least 12ADC channels, and 64kB of Flash, the cheapest ones to pop up were ATSAM3S1C at $2.44ea @ 1k (us.digikey pricing. Actual should be lower). And those are 64MHz Cortex M3.

I'm tired of the anti-3.3V logic arguments. What does commodity even mean? Does it mean standard? What does standard even mean in MOSFETs? It's not like there are any standards... there may be long-life products, but no standards. If you mean "cheap" by commodity, here are two mosfets you can drive direct from 3.3V that are under $1

nxp
infineon

what would be nice is if board manufacturers would either include 5v shift level voltage buffers to i/o or the hardware that attaches can output 3.3v logic signals, which by the way is detectable and outside the grey switching of logic ttl? meaning 3.3v output can be detected and managed by 5v logic just fine. only the 5v output voltage is the issue.

Edited 1 time(s). Last edit at 08/07/2013 08:44PM by jamesdanielv.

Hello.

I want to tell you some more things the MatchboxARM.
The micro used is STM32F103C8T6, it has 72MHz, 64Kb Flash, 20Kb RAM, also on the same board can be used
the STM32F103CBT6 128Kb Flash!
The price of this micro is lower than any ATMEL or Freescale micros with the same speed/flash ammount.
The board works at 3.3V and has a 5V to 3.3V LDO.
Most of the pins are 5V enabled.
The bootloader helps updating the software in an easy maner.
We are oppened to help you porting the software to our board.
We just waiting to be near to our goal on the kickstarter project, and we will make a web page for the board.
We will publish free the Eagle files for the board, and the bootloader source. It's a must now to still keep these closed until
kickstarter project ends.
Best regards!

Bogdan from MatchboxARM team.

Quote

I'm tired of the anti-3.3V logic arguments. What does commodity even mean?

It means, many electronic components work only with 5V signals. Especially ones for handling high currents.

Quote

What does standard even mean in MOSFETs?

The voltage required to switch the MOSFET. That's individual and can be extracted from the data sheet. There are only few which do at 3.3V and even fewer of them can handle 50A and up, which is required to use it for switching the 15A of a heated bed without additional cooling.

All this can be solved, of course. But it makes electronics more complex and more difficult to assemble. RepRap is about DIYers, ya know?

If complexity doesn't matter to you, feel free to pick an Intel i7. Waay faster than any ARM.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

Fascinating thread, especially since MatchboxARM folks have joined us

Just want to repeat earlier question, is there a benefit of using micro faster than 16/20 MHz when controlling a RepRap?

Quote

It means, many electronic components work only with 5V signals. Especially ones for handling high currents.

That was true in the '80s. It might have been true in the '90s. It wasn't true in the '00s. It's not true now.

A tiny fraction of electronic components only work with 5V. In fact, 3.3V is the industry standard for GPIO and parallel, non-differential buses. Even USB is 3.3V for low and full speeds. Most logic inputs (going all the way back to LS TTL) will accept a 3.3V CMOS output as a TTL signal. Most (no, not all, but enough to be relevant) of the ARM micros have 5V tolerant inputs. So 5V logic isn't a problem. So what is? MOSFETs? Maybe, but that's a much more complex picture.

Quote

The voltage required to switch the MOSFET. That's individual and can be extracted from the data sheet. There are only few which do at 3.3V and even fewer of them can handle 50A and up, which is required to use it for switching the 15A of a heated bed without additional cooling.

MOSFETs, even "logic level" ones, are spec'd at 10V Vgs. They (almost) all accept gate drives up to 20V. The IRLB8743 is pretty good as a logic level FET, but you're still not getting great performance out of it. It dumps 3W as heat if you drive it at 30A, Vgs = 4.5. If you were to drive that same FET at 12V, you could probably cut that down to 2W or even 1.5W; I'd be more specific, but IRF's curves are cut off below 0.1V. If you were going to do that, you could probably use a lower-cost FET.

So what does a level translator actually cost, both in complexity and in price? PWM on an extruder or a heated bed doesn't need to be run faster than 1kHz and, at that frequency, a level shifter is a $0.10 mosfet and a $0.01 resistor. And you could easily make up the cost difference on a lower cost MOSFET.

Are we quibling over $0.11? Or is two discretes too complex?

Igor Lobanov Wrote:
-------------------------------------------------------
> Fascinating thread, especially since MatchboxARM
> folks have joined us
>
> Just want to repeat earlier question, is there a
> benefit of using micro faster than 16/20 MHz when
> controlling a RepRap?

I thought I already answered that, did you find my answer lacking?

It's a bit chicken and egg really. People see no need for new electronics because there are no printers that need them, but people don't design new printer features because the standard electronics won't support them!

There is interesting stuff going on with polar and delta printers, web interfaces etc which need more than the standard electronics.

bobc Wrote:
> I thought I already answered that, did you find my
> answer lacking?

I read your answer as mostly addressing cost side of my question, but I see what you mean

On the subject of cost, a complete RAMPS kit goes for about €116 on ebay, Smoothieboard when it was on sale was about €110 and that includes 120Mhz ARM with 512KB Flash and Ethernet. So I would say there is a pretty close argument on cost, and the Smoothieboard is easily more powerful than a Mega2560.

I have backed the MatchboxARM, I hope they make their target. I ported Teacup to an STM32 chip for my Sango-bc project, I might redesign that board so a MatchboxARM can plug in. I could have some firmware running pretty quick, since Chibios has good support for STM32 chips.

The STM32 chips are good value and I found them quite easy to work with.

Annirak Wrote:
-------------------------------------------------------
>

Quote

The voltage required to switch the MOSFET.
> That's individual and can be extracted from the
> data sheet. There are only few which do at 3.3V
> and even fewer of them can handle 50A and up,
> which is required to use it for switching the 15A
> of a heated bed without additional
> cooling.

>
> MOSFETs, even "logic level" ones, are spec'd at
> 10V Vgs. They (almost) all accept gate drives up
> to 20V. The IRLB8743 is pretty good as a logic
> level FET, but you're still not getting great
> performance out of it. It dumps 3W as heat if you
> drive it at 30A, Vgs = 4.5. If you were to drive
> that same FET at 12V, you could probably cut that
> down to 2W or even 1.5W; I'd be more specific, but
> IRF's curves are cut off below 0.1V. If you were
> going to do that, you could probably use a
> lower-cost FET.
>
> So what does a level translator actually cost,
> both in complexity and in price? PWM on an
> extruder or a heated bed doesn't need to be run
> faster than 1kHz and, at that frequency, a level
> shifter is a $0.10 mosfet and a $0.01 resistor.
> And you could easily make up the cost difference
> on a lower cost MOSFET.
>
> Are we quibling over $0.11? Or is two discretes
> too complex?

First off, that's 2 discretes per output, which if you have a board with 2 fans and 2 extruders and 1 HB out means 10 discretes.

The other thing is this expands the board size.

Both of these are not insurmountable, and they're not an issue IMO, but they do need to be accounted for.

FWIW: One of the issues with using 12V to drive the gate on a FET is that you then assume that we're using 12V to drive other parts of the printer, which for some of us is definitely not the case.

My printer runs entirely off 24V, and I know a lot of people who use 15V to get their 12V heated beds to heat properly.

On something RAMPS-like you could simply use a 9V regulator to get an intermediate voltage to feed the FET drivers and the voltage input for Arduino board (eg: a Due).

On other designs it means you either have to:
- Limit it to 12V only (annoying)
- Provide an intermediate voltage that otherwise would not be necessary (bumps up the price if populated)
- Use components that have a greater voltage range (which may bump up their price)

As it is, I see lots of RAMPS boards with 16V capacitors on them (instead of 35V or 50V caps) because they're cheaper. I also see lots of cheap FET's that make it onto RAMPS boards because once again they're significantly cheaper than either the original or the preferred alternatives. This leads to all sorts of annoying issues when people do things that these people don't account for, or just leads to all sorts of random failures.

Anyway, as I said, I'm definitely NOT against the idea of using level shifters to drive the FET's, we just need to get it done right first time to avoid issues down the track.