Axon II: release and design info, got suggestions?

[Admin]

Yeap, I am designing an Axon II. The design is almost complete, but wanted some input before I finish it up and make some prototypes. Feel free to suggest here!

Major changes made from the original Axon:
2x more PWM
2x more external interrupts
much faster, more stable regulated power distribution
four additional I/O
four fewer servos (no one needed more than 20 so far)
green number display
external memory support
lots of various little improvements

With the exception of pinout, the source code is 100% compatible with the previous Axon. Dimensions are exactly the same, too. Price might go down $5, but no promises yet until I get some quotes.

The specs:

# 58 I/O Total
# 16 ADC
# 25+ Servos
# I2C, SPI
# 3 UART + USB
# Up to 8 external interrupts
# 12 PWM Channels
# 64KB Flash, 4KB EEPROM, 8KB SRAM
# 16 MIPS throughput at 16 MHz
# 6 Timers (four 16-bit, two 8-bit)
# pre-programmed with a bootloader - no programmer required
# numerical LED display
# built in 3.3V, 5V, and unregulated power buses
# external memory support (port A)
# all software is free
# 100% open source, large support community
# Windows, Mac, and Linux compatible
# and more!

Expected Release Date: October 1st

I'll update this picture as I incorporate any ideas.

[Ro-Bot-X]

It looks beautifull! Congrats!

[Razor Concepts]

Is there a particular reason why the 1280 or 2560 is not going to be used? Is the 640 already overkill for most projects?

[cooldog]

good work

i like the number display idea

will the original still be available

[Admin]

Is there a particular reason why the 1280 or 2560 is not going to be used? Is the 640 already overkill for most projects?

The 640 has 64KB, but my largest Axon program didn't exceed ~40kb with optimization turned off. No one has reported running out of memory. The 1280 is about the same price of the 2560. The 2560 costs about $5 more each, so lets say I make 1000 units, it'll cost me $5k for memory no one will even use

will the original still be available

The original will no longer be manufactured, and will be sold out around the time the Axon II is released.

[Ryltar]

four fewer servos (no one needed more than 20 so far)

My current (ambitious) design is looking to use at least 33 servos, 13 sensors, and 1 camera >.>
Guess I'll need sub micro-controllers to distribute servo load, though I was looking at doing that already for power distribution reasons.

[BANE]

Man, the Axon II!  I planning to purchase the original Axon this summer too.   Is there any chance of the Axon II coming out earlier?

Bane

[frank26080115]

 I think you should make an Axon-like board with a PIC32MX460F512L instead. The current Axon is pretty much a massive MCU board designed for servos and sensors with a USB to serial converter and power regulation, so why not upgrade it with a better processor that costs less? You'll save even more money by not having that USB to serial converter since the PIC32 has USB OTG.

Have a look at UBW32 http://www.sparkfun.com/commerce/product_info.php?products_id=8971
It should be easy to make an Axon design with the same chip. I'm surprised people don't just solder that to another board with some headers for a robotics board, sure it won't look nice that way, but it'll save $90, minus maybe $20 for a perf-board and power supply parts. You'll still have enough money to get a PICKit2 if you didn't already own one.

Sure, it's overkill, but you are charging $130 for the Axon already. I'm sure most of that isn't associated with part costs, and the ATmega640 is about $12 while the PIC32MX460F512L costs about $8 according to Octopart, so why not? Your manufacturing costs should be the same and your part costs should be lower.

I'm working on an ATmega644 based quadrotor helicopter right now and an major concern has been processor speed due to the trigonometry and filtering functions, the trig has been sped up by using a look-up table in flash, but the filtering is still slow due to the floating point operations. I think a speed boost would be great, and the 32 bit processor running at 80 MHz would really be great for math intensive robotics projects like mine.

You might argue that AVR has more open source tools, but as I work into more complicated stuff, the more I feel the limitations of the free AVR-GCC compiler. The student editions of Microchip's C compilers are free and have great code support too. The PICKit2 is better than an AVR Dragon for about half the cost, although you can always ship the PIC32 with the USB bootloader.

You'll have one less UART and less PWM, and have 5x 16 bit timers instead of 4x 16 bit and 2x 8 bit, but there's two I2C busses, and the UART has autobaud support

[Razor Concepts]

Also the cost for Admin to write whole new source code, and he has to pay for the time for testing the new MCU. He said that the main cost of the Axon was the time put in to design and testing to get it to work just right.

[Admin]

Is there any chance of the Axon II coming out earlier?

Depends on Atmel's stock and how fast I sell the original. Probably won't come out sooner.


frank26080115, appears its better by the faster processor and built in USB. The disadvantage is that I'd have to relearn a new architecture, rewrite all my tutorials and software, make a new bootloader, and make an extra prototype or two until I figure it out. With the Axon you aren't just buying hardware, but also ready to use software and tutorials and forum support to get a beginner started within hours. That sparkfun link only offers datasheets and schematics, its not plug-n-play at the least . . . That being said, Atmel better get some USB in their ATmegas or they'll lose customers . . .


hmmm Razor just beat me to it!

[galannthegreat]

Wow, it looks great!

[frank26080115]

You are targeting a $130 product towards beginners? I thought your target market is competent builders who simply need a powerful prototyping platform. I feel somebody who needs all those IOs and peripherals is going to find the slow speed a real bottleneck someday.

How about sticking a microSD card holder on it? Using Tiny-FatFs is really easy.

[offy]

what is the point of the number LCD? Will it show how much space is used up?

[SmAsH]

what is the point of the number LCD? Will it show how much space is used up?

it only has one digit, could it just be programmable to display a programmed number?

[Razor Concepts]

Looks like the 7 segment LED replaces the status LED from the original Axon. Kind of a waste to use 7 pins on it but I doubt anyone will miss those.

[Admin]

You are targeting a $130 product towards beginners? I thought your target market is competent builders who simply need a powerful prototyping platform. I feel somebody who needs all those IOs and peripherals is going to find the slow speed a real bottleneck someday.

How about sticking a microSD card holder on it? Using Tiny-FatFs is really easy.

Yeap, and it sells great too
I think your quadrotor is an exception, have you considered not using floating point? I multiple everything by 10 or 100 to avoid it. I agree with you that the PIC32MX460F512L appears better (on glance), but it'd require me to learn a new architecture and ditch the current Axon community.

You are the second person to mention the SSD card option, but I decided too small a percentage of people would find a use for it.

Oh and it's only a little more than $1 cheaper (at the quantity that I purchase).

Looks like the 7 segment LED replaces the status LED from the original Axon. Kind of a waste to use 7 pins on it but I doubt anyone will miss those.

Yea I don't think anyone needs more than 58 I/O, and it only costs me ~$1.60 to add it. It uses pins that people wouldn't normally want for anything. It was actually Dunks idea.

offy, SmAsH, it can display any number you want, to represent anything you want. For example, it can display a sensor value, or represent what code it's running at the time. Or perhaps be used as a timer. Or use it to count the number of times you push a button (to specify some value).

[frank26080115]

I have moved to integer math, everything is multiplied by 10000 since the filter needs delta time and that can differ by 1/10th of a millisecond. Each loop took around 3-6ms using floating point and now it takes 1.1 to 2.3ms.

I think I can agree now that it is probably a better decision for you to keep using AVR, but I'd still rather see an inexpensive and powerful 32 bit board designed with focus towards robotics.

With a massive chip like that, somebody might want to do some on-chip debugging, have you thought about making JTAG easily accessible? It will be a pain in the butt to wire up the interface manually.

[Admin]

frank, I just sent an email to Atmel telling them we need more USB supported chips . . .

With a massive chip like that, somebody might want to do some on-chip debugging, have you thought about making JTAG easily accessible? It will be a pain in the butt to wire up the interface manually.

hmmmm ok I might be able to do this. I'll look into it further. Anyone else require JTAG? If USB is available, I find USB and printing out data to be the easiest debugging method.

[frank26080115]

I've had cases when memory locations were accessed where they are not allocated, sometimes even into the stack, or the stack went too deep, and caused resets and other weird behaviour. Your hyperterminal will not let you analyze things like that in detail.

[Gertlex]

Progress is great

I'd suggest trying to move the reset button back to the edge of the board though, to make it easier to access.  But ya, I can tell you've got a mightily crammed board already.

[TrickyNekro]

Nice looking...
I agree with that JTAG think... I know many don't have a JTAG programmer and that's it's more expensive than
ISP... But the quicker downloading (FLASH) and the on chip debugging is nice!
If the JTAG pins are connected to normal IO then you could just ship a wire connector along to get a JTAG pad!!!
Remember that JTAG doesn't need reset to operate... so it's a go go!

[Admin]

I've had cases when memory locations were accessed where they are not allocated, sometimes even into the stack, or the stack went too deep, and caused resets and other weird behaviour. Your hyperterminal will not let you analyze things like that in detail.

point taken

I'd suggest trying to move the reset button back to the edge of the board though, to make it easier to access.

I didn't like it where it was on the original - hard to push with the male headers in the way.

[dannytemmerman]

What will be the cost of axon 2?

Danny

[SmAsH]

he said the cost will be around the same but maybe like $5 cheaper.

[Webbot]

Looks like the 7 segment LED replaces the status LED from the original Axon. Kind of a waste to use 7 pins on it but I doubt anyone will miss those.

Dont forget that the digital LED is made up of 8 independent LEDs (7 + 1 for the decimal point). So you effectively have 8 x status LEDs - obviously if you choose to make them look like a number then thats up to you. So its 'a bonus' - 8 leds that can also look like characters/digits.

'The glass is more than half full'.

[Admin]

I looked into trying to fit JTAG . . . there is just no way I can get a 10 pin header onto the Axon board:
http://support.atmel.no/knowledgebase/avrstudiohelp/mergedProjects/AVRDragon/AVRDragon_Board_Description.htm

Considering the Dragon has expand header holes it won't be hard to wire up. Just six wires to the Axon: 5V, ground, and the 4 JTAG pins.

[daved]

I would like to see the jtag header, not sure where you would put it, but it would be nice.

[paulstreats]

2 x 8 segment displays would let you view 8 bit binary as hex for better debugging (00 to FF). You would have some trouble distinguising between 8 and B on the display but I think if there was an option for 2 then it definately be more useful than 1 8segment.

Maybe you could create an LCD standard rather than an inbuilt display. So you could purchase an optional display unit that is plug and play with supported software.

I know that 1 x 8segment display is better than none, but its really neither here nor there. Look at some graphical LCD's rather than just text ones (you might get some people creating a graphical api type software). with graphical displays you can see your maps live.

If you go down this route, and remove the 8 segment display, you might have room for the JTAG connector and also have additional sales through an add on unit.

Combine this with an addon unit for external memory; you just need an input device add on and your close to having a full user able system

[MrWizard]

I think support for micro sd .

[galannthegreat]

How about an integrated bluetooth/rf setup on it.