what uC + touch-screen dev-kit?

[awallin]

Hi all,
I'm looking for opinions on a user-interface + controller solution for various lab-instrument ideas I have.

Desirable features:
- touch-screen 4" or larger for user input and display
- open-source toolchain (compilers, programmers, etc). programming/debugging over USB or some other common standard.
- existing real-time OS + graphics/touch-screen library
- application programming in C or similar high-level language
- SPI bus for communicating with one or more DAC/ADC chips (I'd want probably up to 5, maybe 8 SPI-chips connected for some projects)
- enough processing power to run PID-loops, digital filters, and datalogging on 24-bit or more (32-bit) numbers at 1 Hz to 1+kHz loop frequency. integer math is probably enough (i.e. floating-point not a strict requirement)
- datalogging onto e.g. SD-card
- possible ethernet connectivity for publishing data, pushing it to a database-server, or remote operation/monitoring of device
- possible battery/LiPo operation (charger circuit would be nice also)
- small, affordable

I've been looking at a standard PC (ITX-motherboard) + FPGA-card based solution, but this is overkill in terms of size and performance for e.g. simple temperature-controllers or data-logger applications.

Then there is e.g. Olinuxino
https://www.olimex.com/Products/OLinuXino/
but again a whole linux-distribution, a possible xenomai real-time kernel, etc. seems a bit overkill.
I'm not sure how easy/hard it is to do SPI on the olinuxino.
Or indeed the quite similar BeagleBoard and its derivatives.

So any opinions on a microcontroller-based approach? Any suitable dev-kits out there?
Any opinions on how to choose between PIC / AVR / ARM / other?

I'm more constrained by time than money, I need something that "just works", preferably open-source, easy to program, where I can develop the analog IO cards myself and just plug them via SPI into the UI/controller board - at least that's my plan now... 

thanks for your input,
Anders

[ddavidebor]

Raspberry pi

[awallin]

Raspberry pi

is there an affordable and available 4" or 7" touch-screen solution for the Pi? last I checked it was so-so.
Also, does the screen use the dedicated lcd-connector, or the GPIO/SPI which I would want to leave free for my application to use?

[senso]

PID loops in a RPi that is running a full blown OS with undetermined response time seems like a great idea 

[ddavidebor]

Well you must use another micro for this

[Harvs]

I've spent a bit of time on this subject.  My conclusion was, for display and touch screen around the 6-10" size nothing can touch (no pun intended) the tablet market.  Android tablets with 800x480 resolution displays and dual core processors are available from a local shop for the same price as I can buy the bare TFT LCD off ebay.  Seriously I've got one of these LCDs sitting here, and that's all it does, it's just not worth the significant time investment required.

This is an really interesting trend in the low volume user interface market too now, as you can go to any shopping center here and see IPads mounted in acrylic holders where a custom build UI would have once lived.

The dev tools for these devices rock (I've only used android eclipse but I here apples is just as good.) 

Then you just need to pick your interface into the device.  32bit ARM micros that have USB hosts are a relatively easy tethered option for droid based tablets (examples are shipped with pretty much every dev board.) Or you can pick a bluetooth or Wifi solution depending on the requirements.

You will no doubt spend extra time understanding whatever communications stack you're going to use to get data in and out.  But, you get massive time savings from using apple/droid dev tools for the tablets.

[ddavidebor]

But you'll get a massive decrease in quality and reliability

[Bored@Work]

I've spent a bit of time on this subject.

But not on his subject. Read the specs, Luke.

- existing real-time OS + graphics/touch-screen library
...
- SPI bus for communicating with one or more DAC/ADC chips (I'd want probably up to 5, maybe 8 SPI-chips connected for some projects)
- enough processing power to run PID-loops, digital filters, and datalogging on 24-bit or more (32-bit) numbers at 1 Hz to 1+kHz loop frequency. integer math is probably enough (i.e. floating-point not a strict requirement)

[awallin]

PID loops in a RPi that is running a full blown OS with undetermined response time seems like a great idea 

I've worked with LinuxCNC(formerly EMC2) in the past. It uses a real-time linux kernel (previously RTAI, but more recently Xenomai), to achieve roughly 10-20 us timing jitter on standard PC hardware. This is fine for PID loops at 1 ms.
There are xenomai ports for ARM, but they are fairly new and in development/testing.

Maybe a full linux system with a xenomai-kernel is the way to go? It's a bit overkill and heavyweight, but at least I don't have to learn any new vendor/processor-specific RT-OS and programming techniques.

This approach does require dedicated hardware for SPI-comminucation at ~MHz, and some buffering of the SPI-data before it can be read/written at 1ms update-rate by a real-time thread.

[Harvs]

I think you might be missing what I was trying to say.

Get something like a nexus 7 tablet for the UI aspects.  Then get as powerful a uC as you need that's got a USB host interface to communicate with the tablet.

PID aspects, SPI etc would all be done on the uC.  The tablet would then take care of any UI requirements.

I don't see how that doesn't meet the OP's requirements or would be any less reliable than other suggested solutions.

Further, it's pretty much off the shelf for the digital hardware aspects.  Just get an ARM dev board with the USB host connector, put whatever RTOS you want on it, use the example USB stack as an android accessory, then sort the rest of your design.

[ddavidebor]

Because this cheap tablet has an expected life of 2 year

So, 2 years fo me is not reliable

[Deagle]

I kinda glossed over the specs you were after, but I have something similar to this:
Kentec touchscreen with a kentec ribbon adapter plugged into a STM32F4DISCOVERY

Screen:
http://nz.element14.com/kentec-electronics/k430wqc-v3-ff/lcd-module-4-3-480x272-touch-screen/dp/1893551
adapter:
http://nz.element14.com/kentec-electronics/eb-stm32f4discovery/expansion-board-lcd-stm32fdiscovery/dp/2216779
Micro:
http://nz.element14.com/stmicroelectronics/stm32f4discovery/stm32f407-usb-otg-discovery-kit/dp/2009276

I went for the 3.5" screen with a stm32F0 (but the adapter didnt support the touchscreen, even though the screen has it ).

It also appears that the bigger screens can just be swapped for the smaller ones.

[Spikee]

I have such a screen and the stm32 dev board but the refresh rate is very slow unless you want to do some DMA stuff etc...
Even if you get it working it will influence the performance of the rest of the system (if use use it also to do PID stuff etc).

I have looked a the same problem for a while but a tablet seems the best solution for UI. And a tablet can handle a few years easy. And if it breaks you just need a random brand Android tablet. Another way would be a small ITX board or something like that but that will be more expensive and bigger.

[MacAttak]

Because this cheap tablet has an expected life of 2 year

Extraordinary claims require extraordinary evidence. Where do you get the 2-year life expectancy rating from?

I've not seen a tablet die in under 2 years that wasn't being seriously abused. I still have a first-generation iPad that works like a freeking champ (used several hours per day). In fact, I kind of wish it would break so I would have an excuse to buy a new one with one of those pretty Retina displays... but the thing just won't die. And all of the Android tablets I have seen were similarly robust.

[ddavidebor]

if you want to put a tablet inside something you cant buy a 700$ ipad..

[Rasz]

if you want to put a tablet inside something you cant buy a 700$ ipad..

no, if you want to put a tablet somewhere put a $43.80 tablet there
http://dx.com/p/jxd-s18-4-3-resistive-screen-android-4-0-mini-pad-tablet-pc-w-tf-wi-fi-g-sensor-black-161536

this has better specs than any dev kit and you can learn android on it

[nctnico]

PID loops in a RPi that is running a full blown OS with undetermined response time seems like a great idea 

No problem at all. Just use a timer interrupt like on a microcontroller.

[Rasz]

PID loops in a RPi that is running a full blown OS with undetermined response time seems like a great idea 

No problem at all. Just use a timer interrupt like on a microcontroller.

interrupts have non deterministic response time under linux (or any other preemptive os)

[nctnico]

If that where true then a lot of hardware wouldn't work. You shouldn't expect nanosecond accuracy on the execution of an interrupt (just like on a microcontroller) but sub millisecond shouldn't be a problem at all.

[ddavidebor]

Use a pc type processor to control this tipe of things is just complicated and unreliable.

I will never buy something done that way.

[nctnico]

And yet some industrial control systems work that way. Often the ethercat protocol is used to control I/O modules where the processing is done on a PC (and sometimes time critical things are done inside the interrupt service routine of the ethernet card!). If you are willing to pay then HP offers a wide range of very stable PCs.

Besides. I thought we where talking about some embedded (ARM) device.

[ddavidebor]

Why si much trouble when you can use a 5$ micro

[Rasz]

Why si much trouble when you can use a 5$ micro

5$ micros dont have fancy touchscreen UIs

[ddavidebor]

But 5$ micro can do pid!

And work, work without a boot, an os, a fu****g java virtual machine and a damn touchscreen!!!!!

Do you understand?

[SeanB]

So can a quad opamp and a few passives, and lasts generally decades in industrial service.