Maple vs. Arduino-Due « LeafLabs Garden

LeafLabs Garden » Support » Maple Hardware Support

Maple vs. Arduino-Due

(27 posts) (9 voices)

Started 2 years ago by randomvibe
Latest reply from uzi18

12 Next »

randomvibe
Member

The Maple dominated the 32-bit board market for some time. Now with the Arduino Due, and huge worldwide support, and very active forum and blog (unlike leaflabs), the Maple has been marginalized.

What's the big advantage for the Maple over Due?

I was a big fan of the Maple. I must say I've been turned off by Leaflabs appearance of indifference and lack of enthusiasm.

I think Aruduino Due will steam roll and marginalize most 32-bit boards. The good news is that the Leaflabs team is very quick and inventive when they want to be. If they want to compete, they'll need to bring discriminating features and improvements and game changers. Do not reinvent the wheel. Here's a suggestion: a single-board-computer like the R-Pi but with the GPIOs and peripherals (12-bit ADCs, USART, SPI, I2C, etc.) like the Maple, but programmable in Octave or Matlab, for $45 bucks. A tall order. It's been kicked around at adafruit:

http://forums.adafruit.com/viewtopic.php?f=8&t=35959

To the Leaflabs team, I wish you well.

Posted 2 years ago #
bnewbold
Ex-Leaflabs

Hi randomvibe,

Reading your original post over at adafruit, the Rascal Micro seems pretty similar to what you're asking for:

http://rascalmicro.com/
http://rascalmicro.com/docs/

The device runs linux and is usually programmed in python. It is based on a Atmel ARM microprocessor, and the board seems to be decently documented and costs $175. If you are committed to the lower price-point you might be able to raise about a million dollars on your own and do a decent sized production run with volunteered time to get the price down?

Alternatively, Olimex sells a variety of i.mx* and A10/A13 embedded linux development boards at a lower price but with significantly less documentation and "out of the box" python integration:

https://www.olimex.com/Products/

If you are looking for other ways to get Octave-like syntax on an embedded device, and didn't like the Eigen C++ libraries linked in the other thread, you could find or write linear algebra packages for the eLua embedded runtime:

http://www.eluaproject.net/

As an aside, I was quite hopeful about the Fortress programming language (science-oriented, from Java founders, supposed to be a high level replacement for FORTRAN, LaTeX-style formatted source code) for doing DSP on embedded devices, but that project shut down a while ago:

http://java.net/projects/projectfortress/pages/Home

Posted 2 years ago #
randomvibe
Member

I'm not interested in developing boards or kickstarting a project. I'm just a Maker voicing feedback as to what I'd like to see in the future as a customer.

The Rascal is very interesting. But with that price, $175, it will be squashed by the R-Pi when R-Pi adds native ADCs and other Maple-like peripherals.

Back to the main question: What's the advantage of the Maple over the Arduino-Due? How are you going to keep customers like myself kept?

Posted 2 years ago #
bubulindo
Member

If you read the latest posts, you'll see not a lot is going to happen on the Maple world any time soon.

I wouldn't say the problem with the Maple is lack of support. If you check the number of people actually making good comments in the ARduino forums and here, the number is very similar (percentage wise, here is extraordinary though).

Maple could come back if leaflabs developed a bigger board with native support for Ethernet, SD cards, for example... that would come in handy to several projects and is something that the STM32 line of chips has (I'm sure the ATMEL ones also do).

Pricewise... I think it'll take a bit of time for the Due to catch up. The Maple mini is quite an extraordinary device for its size and the original Maple, or Olimex clones are far cheaper than the Due is at the moment. One can also argue that the Olimexino brings a lot more on the board than the original Arduino with it's SD card slot and a nice power regulator. :)

Posted 2 years ago #
gbulmer
Moderator

bnewbold - I am happy to believe Brandon Stafford is a wonderful person, but I'd also like to know why anyone would buy a $175 Rascal, when I could buy a $35 Raspberry-Pi which appears to be superior?

AFAICT the only benefits of Rascal are: lower-power, Arduino shield pin-out and trippy lingo on the web site: "Reference documents are the crack cocaine of documentation-- pure facts flooding directly to your brain". Wow! Wonderful! Documentation!

I can't even discover how much RAM is on the board! Maybe they need to talk to their "documentation pusher", and ask them to "stop cutting their reference documents with hyperbole". A triumph of form over function?

There are already expansion boards for a R-Pi which negate some of the advantage of using Arduino shields, and R-Pi itself negates other shields directly (Ethernet, USB host, Audio, SD memory). I personally know someone who seems to be working on solving ADC input, so there are probably hundreds of folks working on it.

randomvibe - I think the hardware feature part of your spec is doable within 8-12 weeks if you remove the single-board constraint; Raspberry-Pi + STM32F daughterboard. Doesn't octave work on R-Pi already?

I have a different view wrt Arduino Due.

The arrival of Arduino Due has significantly increased the interest in ARM within the Arduino and Maker community. The Arduino Due isn't 'perfect'. There is still room for alternatives (I actually did a talk this evening at a local Linux group, and folks were interested in R-Pi and STM32FxDiscovery boards).

My interest is the sub $15, breadboard friendly, DIY-able ARM, especially in robots, audio etc. IMHO Arduino Due isn't as good as STM32F.

Further, there are lots of other directions embedded-ARM can go because there are so many manufacturers with funky, interesting and useful devices. IMHO, ARM is a significantly different environment to the previous Arduino alternatives, limited to making AVR-based alternatives to Arduino Uno (and predecessors). Part of the attraction is all ARM Cortex can leverage the same toolchain.

So, I think Arduino Due has done a great service to ARM and the wider maker community, and I hope for much more interesting variety spurred on by Arduino Due. (Example Freescale Arduino-shield-shaped Cortex for $12.95 https://www.element14.com/community/videos/7820)

Posted 2 years ago #
crenn
Moderator

There is something that the Arduino Due can't compete with compared with the Leaflabs Maple and Olimexino; Size. It could be possible in the future for the Arduino team to release something that could actually compete with Maple and Maple Mini in terms of size as there are chips with small enough pin counts to do so, but I haven't heard anything regarding that.

There is some things the Arduino Due can't boast, and it's actually quite limiting for some projects but it sort of becomes mostly a Atmel SAM3X8E vs ST Micro STM32F103 comparison. Things like 5V pin tolerance, more than 1 SPI (although the Atmel chip's USARTs do support SPI mode, just like the other chips, including the STM32F103), more PWM pins (although I can't tell whether this is a chip limitation or a design/marketing decision), multiple ADCs (which mean you can sample more than 2 pins at the same time with some programming, the Atmel chip only has a single ADC with a analog multiplexer) and a single USB port (this is both an advantage and disadvantage, it depends on how you view it).

Then there is the design decisions which annoy me to an extent. The ICSP header is still there just for SPI (and only for SPI since the SAM line doesn't need ICSP) which includes the 'master reset signal' which also happens to erase the backup regions. The 'master reset signal' also annoys me as there are times where I can see it would be useful, but the fact they don't even breakout the normal reset signal which resets the microcontroller only (not the backup regions) is something I can't understand.

I'll finish off with this, there are some things I like about the Arduino Due, but some things I don't like. I don't doubt the Due will be more popular than the Maple, but there is a lot of advantages I feel the Maple has over the Due. I'll probably be getting a Due in the future, but it's not exactly high on my priority list as I don't have a huge use for it. If leaflabs released the Maple II, I'd be getting that over the Due as there are improvements in that compared to the Due. The only thing I'd wish for is for Leaflabs to release something which has the layout of the Arduino Mega/Due, but that's not likely to occur. If I recall a conversation a while ago, they were more interested in the Maple Native.

@gbulmer, it's got 512Mb (64MB), the model number for the SDRAM is most likely (this is from the schematics which is missing some numbers) MT48LC16M16A2, and there are 2 chips. The fun thing is, I can't see why it's 'worth' as much as it is, especially as I looked up the cost of the chip (Atmel AT91SAM9G20 which runs at max 400MHz) and in 100 quantities it's USD$8.4936 each (digikey). I'm guessing they're producing low quantities. There is only 2 advantages to it compared with the RPi, it does have an ADC, although it's slow and only 10-bits, and it does have more GPIO technically. Although quite a bit of the GPIO is used else where or not broken out and the analog input (Arduino Shield header) actually only has 4 channels, not the usual 6.

Also, the RPi won't add ADCs or some of the other things you'd find with a STM32F chip. But you can add external ADCs/DACs easily enough, people do. I also know it's possible to pair a Maple with a RPi, since it's what I've been doing with my Raspbian Rover (I need to draw up a schematic and do documentation).

Posted 2 years ago #
bubulindo
Member

@crenn, loved your comment and insight about the Maple vs Due. I hadn't looked into the schematic/datasheet and wrongly assumed it would be in line with the STM32 chip used in the Maple.
Having noticed that, for example, on the Arduino Mega they decided not to break out the TOSC pins and not to put a nice 32kHz on those, I'm not surprised to see similar things on the Due.

Also, since you mentioned the ICSP with the SPI connection on it, that is most likely to maintain compatibility with the original shields (notably the Ethernet one), and that is probably a place where the Maple is not stuck. On the other hand, having a different form factor than the biggest platform isn't very helpful without some expansion solutions or functionality provided by the shields.

Maybe that is a point where Leaflabs could improve... one other solution (and this may be stretching it), could be a joint venture between leaflabs and Olimex for example. Some of their STM32 boards are pretty cool (with expansion possibilities), and with software support to add these to the Maple IDE, the whole platform could see some growth and interest.

But, this is just my point of view.

Posted 2 years ago #
crenn
Moderator

The 'ICSP' port is to retain compatibility with the shields, it's actually something that bugs me quite a bit. One thing I've thought about is making an 'arduino' compatibility shield for various boards to make it compatible with almost all shields, there is always going to be trade offs though.

The Mega doesn't have an RTC last I checked (just checked, it has a "Real Time Counter with Separate Oscillator"), but the Due does have an RTC and it actually has a 32.768kHz crystal on board for it.

Posted 2 years ago #
gbulmer
Moderator

crenn - yes I looked at the schematics and noted the memory chips had the actual part numbers missing.

It seems to me that it is easy and straightforward to list its resources.
Simple, helpful and what almost everything else has.

64MB compared to the R-Pi 512MB seems a bit miserly for a $140, 5x, price premium.
Even if the R-Pi didn't exist, Rascal hardware can be beaten for a lot less cash.
So other than Python, which is also one of the R-Pi's strong areas, I really am struggling.

I am actually willing to believe Rascal is worth $175, but right now I don't get it.

Edit: Octave runs on Raspberry-Pi http://www.raspberrypi.org/phpBB3/viewtopic.php?f=27&t=24687
It appears (and randomvibe is on that thread) that GPIO is available.
So a good C/C++ developer could add access to a daughterboard carrying extra capabilities like ADC. Probably already done.

As for Arduino Due vs Maple; technically it is mostly STM32F vs SAM3X8E. For my interests, STM32F wins. Looking ahead, a STM32F3 in a 48pin package on a Maple Mini size board is a much better fit than an Arduino Due; the 100pin SAM3X8E isn't going to compete there. It'll be a struggle to make a breadboard friendly PCB with SAM3X8E unless the Ball-Grid-Array (BGA) part is used. BGA takes it beyond my current maker abilities.

The breadboard/veroboard-unfriendly Arduino header format is a big point against the Due for me. I'll likely buy a Due to support the team, and make it easier to port stuff to STM32F.

However, I think there is so much useful variety in ARM-based MCUs, that people will look beyond the familiar. Maple is well positioned to support those folks, as the base libraries and IDE are already in place.

I think wireless connectivity will become more critical. I am interested in mesh networks, where ST have STM32W products.

(Full disclosure: I am not LeafLabs' staff)

Posted 2 years ago #
randomvibe
Member

gbulmer - you are correct, I confirmed that Octave runs on the R-Pi, but inversion for small matrices is slower than I'd like, and there is no way to communicate with the input/output pins (GPIO). "Andy1978" claims to have done simple GPIO operations, but I have not confirmed it - although I tried very hard (I lack Linux skills).

I think Maple vs. Arduino-Due is about the system as a whole, not just STM32F vs. SAM3X8E. By system, I mean the processor, board, layout, software/IDE, forums & blogs, and compatibility with peripherals. The forum/blog is very important for me.

I'm not a microprocessor expert, but I don't think the STM32F and SAM3X8E differences are as dramatic like a transistor vs. a vacuum tube. I think they're close enough for the majority of makers like myself.

For now, I'm moving on with the Arduino-Due. The extensive blogs & forums and extra libraries are extremely important for my robotics projects involving feedback control systems. I even have my own libraries for PWM (setup multiple PWM pins with different PWM frequencies!) and Octave-like matrix programming library using Eigen.

I think the leaflabs brand suffered a bit due to lack of communication, but I'm still a big fan because of their impressive 32-bit board clearly ahead of it's time in the "early" days. Hopefully we'll see something new and game-changing again in the future. An R-Pi plus a Maple sounds very interesting if the "getADC" function is fast. A new single-board-computer with native ADCs (and other Maple-like peripherals) on a breadboard friendly board sounds even better.

Posted 2 years ago #
ala42
Member

From the CPU performance the SAM3X8E is at about the level of a some years old STM32F1. Compared to a one year old STM32F4 it has 1/2 of the integer speed and 1/10-1/20 of the floating point speed, as it has no FPU. It also lacks 5V level compatibility and has fewer PWM channels. The SAM3X8E is simply a very bad choice when a STM32F4 is available. As the Arduino guys are linked to Atmel, they had to chose a chip which was already outdated at their first announcement about a year ago. This was a bad decision for the end user, especially as they kept the mechanical pin design of the older Arduino boards, while they are not electrically compatible due to missing 5V tolerance.

Posted 2 years ago #
rascalmicro
Member

Hi all,

I'm the guy that built the Rascal, and I am delighted to learn that someone around here believes I'm a wonderful person!

I get questions like, "Arrr! Why would I buy a Rascal for $175 when I could buy <long list of specs> for less!" pretty frequently. In general, I would say that if you just want to play with electronics (rather than build some specific device) and your embedded Linux skills are strong, by all means, please buy the cheapest, commodity Linux board you can find.

The people who buy Rascals are mostly people who are trying to get something done and can afford to pay a few hours wages to skip a week or two of messing around with kernel configs, compiler settings, SSH, and the rest of the command line stuff that drives new Linux users crazy.

I get a lot of orders from scientists who use Numpy and for data-crunching or interactive art folks who use Javascript stuff like D3 for visualizations. The Rascal's web interface, Python libraries like Numpy, and Javascript libs like D3 and jQuery, make life much easier. In theory, all the open-source software could on the Rascal could be ported to some cheaper Linux board, but it took me around a year to do make it all play nicely together. Also, the Arduino shield compatibility is a big plus, at least to me and my customers.

Anyway, I'm happy to answer questions about the Rascal, either via email (brandon at rascalmicro dot com) or in the Rascal forums (http://rascalmicro.com/forum).

Brandon Stafford
Wonderful Person

Posted 2 years ago #
gbulmer
Moderator

randomvibe - "I confirmed that Octave runs on the R-Pi,"
My reading of the thread is confirmation had already happened.
You might not have believed it until you did it yourself, but that's different.

"... and there is no way to communicate with the input/output pins (GPIO)"
Maybe "I haven't been able to get gpio to work" might be fair and more accurate.

""Andy1978" claims to have done simple GPIO operations, but I have not confirmed it - although I tried very hard (I lack Linux skills)."

Please try to be objective and fair. Please don't project your own lack of skill, knowledge or success to become an assertion like "and there is no way to ...".
People who don't have the time to read the Raspberry-Pi Ocatave thread might be mislead by you, and that would be a waste of everyone's time. Someone might come across this thread, and mistakenly believe you, and hence not go on to do a critical bit of work which would complete the very thing that you are asking for.

A question like "According to the link below, a C library providing access to the GPIO pins exists. Is it possible to link that library to Octave?" suggests you lack the understanding to estimate the technical feasibility of making Ocatave use GPIO.
Your further comment:
" I am at a disadvantage of not being a Linux expert or C/C++ programmer or installer, etc. I think C/C++ is as horrible as radioactivity." seems to confirm that you aren't equipped to asses feasibility or make it work.
Maybe it'd be helpful to avoid making assertions about feasibility? After all, it doesn't sound unlikely does it?

Raspberry-Pi already has device interfaces to GPIO, so that bit works. That implies there'll be a C interface: programs can get at Linux system resources, for example devices, and hence can be accessed by programs written in C/C++.

According to http://en.wikipedia.org/wiki/GNU_Octave, Ocatave is written in C++.
Further, wikipedia claims "Octave is extensible using dynamically loadable modules."
This is confirmed in the Ocatve documentation: http://www.gnu.org/software/octave/doc/interpreter/Dynamically-Linked-Functions.html#Dynamically-Linked-Functions
So we might reasonably expect to be able to write some C/C++ code that would enable an Ocatave script to get at Raspberry-Pi GPIO without having to modify the core of Ocatve. Of course, if one were willing to modify the core of Ocatave, it is even more likely that a raspberry-Pi version of Ocatave could use GPIO.

So IMHO it is extremely likely that a skilled and knowledgeable C/C++ developer can get it working.
My reading of the thread is that, at least, a proof-of-concept has already been done.

Posted 2 years ago #
gbulmer
Moderator

Brandon - Thanks for stopping by, and sharing some answers. I'm not sure if it would fit with your web site, but those points seem quite helpful, and might be useful there.

I can reasonably conclude that I am not your target audience :-)

"The people who buy Rascals are mostly people who are trying to get something done and can afford to pay a few hours wages to skip a week or two of messing around with kernel configs, compiler settings, SSH, and the rest of the command line stuff that drives new Linux users crazy."

Okay, I can buy that. Some Linux users seem to think a dauntingly steep learning curve is part of the fun! But not me. One of my interest is helping folks build their own devices. Frankly, many of them might struggle to afford $175. That might more available income than they have for a month or more. We aren't your target audience.

I would suggest that Raspberry-Pi is not simply a cheap, commodity Linux board. It has a significant community of developers working on diverse projects, including embedded Linux. As evidence their are R-Pi builds of Linux packages. A goal of R-Pi is to introduce new users to computing, so I expect the initial install and configure will become better and better.

R-Pi may even pass the one million mark in a year. (I haven't seen anything on Arduino sales, but it may be nibbling at their heels)

So if you are only one man year of effort ahead, might it be worth making an R-Pi daughterboard, and doing some porting?
I assume their is quite a lot of basic Linux hacking that might be usefully done by a wider R-Pi community, saving you effort so that your value-add can leverage lots of diverse work, not just your own.

Edit: I wrote "I am happy to believe Brandon Stafford is a wonderful person". However, that isn't as big an accolade as you might think. I am happy to believe the overwhelming majority of humanity is fundamentally wonderful, even when they are wrong (i.e disagree with me ;-)

Posted 2 years ago #
gbulmer
Moderator

ala42 - I agree. I don't think the Arduino team "knocked one out of the park"** with the Arduino Due.
I don't mind them using Atmel Cortex-M; I think that was extremely likely (but no conspiracy theories please:-)
Due was delayed for such a long time after announcement that I had expected Aduino Due to use a Cortex-M4, so I was disappointed. Maybe M4 is this years Arduino product?-)

However Arduino Due seemed like a good opportunity to introduce more friendly mechanical design of the header-pin interfaces, and get the community onto something more maker friendly. As you say, given the lack of 5V tolerance, sticking with the old interface didn't seem essential; IMHO it actually seems like using a different mechanical interface would have made more sense.

Though they may have judged their user base well - looking like an Arduino may be more important than almost anything else.

** I've been to pro-baseball games at Fenway, Toronto, and elsewhere, so even though I'm British, I think I'm allowed that metaphor.

Posted 2 years ago #

RSS feed for this topic

12 Next »

Reply »

You must log in to post.