1 10 11 12 13 14 Previous Next 188 Replies Latest reply: Jul 15, 2014 8:34 PM by wallarug Go to original post RSS
  • 165. Re: Interesting "Competitors" for the Raspberry Pi
    morgaine Level 15

    Freescale released their new i.MX6 range of application processors last week, and the release announcement contains a paragraph that directly relates to the subject we were discussing earlier, continuity of support:

     

    "Freescale’s i.MX 6 series processors are included in Freescale’s product longevity program, with assured supply for a minimum of 10 years. For Terms and Conditions and to obtain a list of available products, visit freescale.com/productlongevity."

     

    That seems pretty clear.  It's certainly reassuring to read a totally unambiguous statement of intent.

     

    The NDAs  have now been lifted, in theory at least, although in practice the reference manual is still hidden beyond my meagre surfing abilities.

  • 166. Re: Interesting "Competitors" for the Raspberry Pi
    John Beetem Level 15

    Morgaine Dinova wrote:

     

    Freescale released their new i.MX6 range of application processors last week, and the release announcement contains a paragraph that directly relates to the subject we were discussing earlier, continuity of support:

     

    "Freescale’s i.MX 6 series processors are included in Freescale’s product longevity program, with assured supply for a minimum of 10 years. For Terms and Conditions and to obtain a list of available products, visit freescale.com/productlongevity."

     

    That seems pretty clear.  It's certainly reassuring to read a totally unambiguous statement of intent.

     

    The NDAs  have now been lifted, in theory at least, although in practice the reference manual is still hidden beyond my meagre surfing abilities.

    I sure hope Freescale stays around at least that long

     

    I've been surfing Freescale web site for a long time.  The structure used to change all the time which was a PITA, but it's more stable nowadays.  Here's the document page for iMX6 Dual: http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=i.MX6D&fpsp=1&tab=Documentation_Tab

  • 167. Re: Interesting "Competitors" for the Raspberry Pi
    John Beetem Level 15

    John Beetem wrote:

     

    Morgaine Dinova wrote:

     

    The NDAs  have now been lifted, in theory at least, although in practice the reference manual is still hidden beyond my meagre surfing abilities.

    ... Here's the document page for iMX6 Dual: http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=i.MX6D&fpsp=1&tab=Documentation_Tab

    51 MB .pdf for the Dual/Quad RM

     

    An interesting iMX6 feature: in addition to the OpenVG 1.1 2D engine which has the usual GPU lack of  documentation, the iMX6 has a "R2D" raster 2D graphics core for DirectFB (Linux) and GDI/DirectDraw (Windows CE).  This has quite a lot of description, but not enough in the RM to program directly.  However, if DirectFB will support R2D then the instruction formats will need to be exposed at some point.

  • 168. Re: Interesting "Competitors" for the Raspberry Pi
    Level 15

    51 MB .pdf for the Dual/Quad RM

    makes for some interesting reading..  I'm quite amazed by how much they've packed into it, although at 5700+ pages the ref manual isn't exactly light reading..

     

    A couple of interesting bits I noticed:

     

    If a 24 MHz input clock is used (required for USB), the maximum SoC speed is limited to 996 MHz.

     

    1. The theoretical maximum performance of 1 Gbps ENET is limited to 470 Mbps (total for Tx and Rx) due to internal bus

    throughput limitations. The actual measured performance in optimized environment is up to 400 Mbps. For details, see the

    ERR004512 erratum in the i.MX 6Dual/6Quad errata document (IMX6DQCE).

     

    and there's bound to be more like those..

     

    The ENET errata is quite amusing when coupled with:

     

    23.1 Introduction

    The MAC-NET core, in conjunction with a 10/100/1000 MAC, implements layer 3

    network acceleration functions. These functions are designed to accelerate the processing

    of various common networking protocols, such as IP, TCP, UDP, and ICMP, providing

    wire speed services to client applications.

     

    hmm... "wire speed" 

     

    It's also interesting to see it has a seemingly decent hardware RNG and crypto acceleration module - something sadly lacking on lots of other platforms today.

     

    Overall impression is that it's going to be a very interesting platform, assuming the costs work out good enough to give it a chance.

    The obvious amount of silicon space and effort that's been put into likely unuseable GPU/VPU is disappointing, but not unusual. I can't help thinking that if they targeted the real OpenGL instead of the cut down ES there'd be more of a chance of getting useful graphics support.

    So I'm still waiting to see an Arm SoC that doesn't have phone/tablet as it's primary target, maybe then some of the peculiar design constraints can be discarded in favour of a better overall design.

     

    Still can't decide if a CubieBoard / A10 OlinuXino or Wand(a)board will end up being the better system.

  • 169. Re: Interesting "Competitors" for the Raspberry Pi
    GeorgeIoak Level 4

    I've started my discussions with Freescale and with any luck I can all my questions answered before Thanksgiving. Now that CubieBoard successfully launched their campaign it will be interesting to see how they handle their success and what their board can really do. If they can handle their success and the board is a solid performer it might be tough to compete against them although I still have hope.

     

    My goal is to hopefully get a good enough price on the i.MX6 Solo so the board can be designed at an entry level cost to be competative with RPi and CubieBoard but then also be able to offer the option of using any of the higher performance parts in the i.MX6 line. If I have to design a board just for the iSoloLite then it better be a pretty darn good board at a really low price in order to justify the effort to design and produce it.

     

    I already feel like it will be tough since CubieBoard and Wand(a)board have already produced prototypes but we'll see.

     

    Anything you guys can help with would be by looking through the documentation would be great!

  • 170. Re: Interesting "Competitors" for the Raspberry Pi
    Level 15

    George,

     

    A couple of hopefully obvious suggestions:

     

    Keep it simple, try not to get suckered in by 'shiny' stuff very few people will use. Or put another way, don't spend lots of time and resources on MIPI, CSI, DSI, MLB, LVDS, FLEXCAN, WiFi, bluetooth.

     

    Solid USB implementation and power is much more useful as it allows people to do their own thing.

     

    Do take care of the basics - battery backed RTC for example

     

    If possible, allow for the option to get the board up to the full 4G

     

    Flexible boot configuration - allow the user to choose to boot from nand / sd / sata / usb, don't force a boot method with OTP fuses

     

    Try not to force strange configuration options - like having 1GB ram means you MUST have an on-board camera..

     

    Keep the experimenter / tinkerer in mind. Plenty of GPIO's available and not on some expensive 0.01mm pitch smt header that's effectively not able to be hand soldered by most people.

     

    Oh, and try to keep obviously external connectors (usb/lan/audio/hdmi/vga etc) on just one side of the board.  If there's a barrel connector for power, at least allow an internal power connector as well. (there's probably also a use case for an internal usb connector too)

     

    No micro usb charger connectors tho

  • 171. Re: Interesting "Competitors" for the Raspberry Pi
    GeorgeIoak Level 4

    selsinork wrote:

     

    Keep it simple, try not to get suckered in by 'shiny' stuff very few people will use. Or put another way, don't spend lots of time and resources on MIPI, CSI, DSI, MLB, LVDS, FLEXCAN, WiFi, bluetooth.

     

    That's probably going to be the hardest to do in the initial design phase, especially if I'm able to get the Solo part at an attractive enough price to use on the initial board. If the pricing on the Solo works then trying to make a generic enough board to handle the full line which still keeping the entry price low is going to take some brain power!

    Solid USB implementation and power is much more useful as it allows people to do their own thing.

    This is an absolute must. If the USB core is crippled I won't even bother putting my time into this project.

    Do take care of the basics - battery backed RTC for example

    For the entry level board this will be tough. I haven't researched low cost option but the cheapest Dallas RTC (Maxim) is the DS1375 ( http://www.maximintegrated.com/datasheet/index.mvp/id/3815) and that alone adds ~$0.50.

    Flexible boot configuration - allow the user to choose to boot from nand / sd / sata / usb, don't force a boot method with OTP fuses

    IMHO, the board has have the option to boot from an onboard NAND part. That's going to have to be a cost adder though even though I can get 8GBit MLC NAND parts for under $3. If there's anyway I can squeeze that into the entry level board it definitely will go down but it absolutely will be designed in.

    Try not to force strange configuration options - like having 1GB ram means you MUST have an on-board camera..

    This one is going to be a little tricky only because they offer a 64-bit bus which opens up all sorts of options. I haven't gone through the spec sheets in detail but my first thought is putting down a 64Mbit x 16 (or 128Mbit x 16) givng the board 1GB (or 2GB) of DDR3. Then I could have the option of adding a SO-DIMM connector and using standard DDR-3 SO-DIMM memory modules.

    Keep the experimenter / tinkerer in mind. Plenty of GPIO's available and not on some expensive 0.01mm pitch smt header that's effectively not able to be hand soldered by most people.

    A couple of off the top of my head options on this one is to provide a limited amount of GPIO's on a 100mil header and then have a small Hirose type header and immediately offer a breakout board that plugs into this header. Producing 1k of these small "simple" break out boards will keep the cost down and immediately give people the option of getting to anything they want as well as allowing people to design their own adapter boards for specific use cases.

    Oh, and try to keep obviously external connectors (usb/lan/audio/hdmi/vga etc) on just one side of the board.  If there's a barrel connector for power, at least allow an internal power connector as well. (there's probably also a use case for an internal usb connector too)

    I'm not sure if this is a good idea but again my intial thought is just to provide the mimimum; USB, LAN, HDMI; all together. Then have something like on a computer motherboard where you can plug into a header for "front panel" connectors.

     

    This is going to take some up front thinking but I think it makes the most sense. Provide a "kick ass" base board that does the basics very well and then provide building blocks for options/features. If these are designed in the beginning then a standard case could also be designed and anybody can mix n match what they want to create their own version of the board.

    No micro usb charger connectors tho

    This is a tough one. I personally hate the micro USB connector but it has become the defacto standard and there's no escaping it. I haven't thought through this or the power connections but for power there will be a connector as well as a header.

  • 172. Re: Interesting "Competitors" for the Raspberry Pi
    Level 15

    George Ioakimedes wrote:

     

    That's probably going to be the hardest to do in the initial design phase, especially if I'm able to get the Solo part at an attractive enough price to use on the initial board. If the pricing on the Solo works then trying to make a generic enough board to handle the full line which still keeping the entry price low is going to take some brain power!

    Nonody said it was going to be easy   Haven't checked, but I was under the impression that all the different parts were at least pin compatible, so hopefully you don't need a redesign when going from solo->dual->quad...

     

    Do take care of the basics - battery backed RTC for example

    For the entry level board this will be tough. I haven't researched low cost option but the cheapest Dallas RTC (Maxim) is the DS1375 ( http://www.maximintegrated.com/datasheet/index.mvp/id/3815) and that alone adds ~$0.50.

    According to the ref manual there's something called an SRTC or Secure Real Time Clock in the SoC, but I'm currently drawing a blank on any more documentation.. regardless, not having an RTC is a fairly big problem for just about anything - witness the inability to login to the early fedora builds for the Pi unless networked and the clock set by NTP.. So, even if populating an external one proves too expensive, at least having the footprint for a common RTC + battery would allow people to solder one on themselves.

     

    Try not to force strange configuration options - like having 1GB ram means you MUST have an on-board camera..

    This one is going to be a little tricky only because they offer a 64-bit bus which opens up all sorts of options. I haven't gone through the spec sheets in detail but my first thought is putting down a 64Mbit x 16 (or 128Mbit x 16) givng the board 1GB (or 2GB) of DDR3. Then I could have the option of adding a SO-DIMM connector and using standard DDR-3 SO-DIMM memory modules.

    Sounds good... In case you didn't realise, I was having a mild dig at the wand(a)board where to get 1GB ram you have to take a dual core + wifi + bluetooth...

     

    A couple of off the top of my head options on this one is to provide a limited amount of GPIO's on a 100mil header and then have a small Hirose type header and immediately offer a breakout board that plugs into this header. Producing 1k of these small "simple" break out boards will keep the cost down and immediately give people the option of getting to anything they want as well as allowing people to design their own adapter boards for specific use cases.

    that sort of thing is fine, if the board is available immediately. One of the other dev boards I have doesn't have a breakout board available leaving you to build one yourself and significantly reducing the ability to easily tinker..

     

    Oh, and try to keep obviously external connectors (usb/lan/audio/hdmi/vga etc) on just one side of the board. 

    I'm not sure if this is a good idea but again my intial thought is just to provide the mimimum; USB, LAN, HDMI; all together. Then have something like on a computer motherboard where you can plug into a header for "front panel" connectors.

    That sort of thing would be perfect. Makes it significantly easier to put in a random case. Lots of the current boards have connectors on too many sides and no thought for the screw post that's in the corner of a lot of common cases or for people who want a one box device with internal power supply.

     

    No micro usb charger connectors tho

    This is a tough one. I personally hate the micro USB connector but it has become the defacto standard and there's no escaping it. I haven't thought through this or the power connections but for power there will be a connector as well as a header.

    I don't particularly have any gripe with the concept of a usb charger, but the Pi has proven that the implementation leaves a lot to be desired. Mainly down to crap cables and really cheap&nasty psu design, but you really do have to take that crapness into consideration and ideally come up with a better way to supply power..

  • 173. Re: Interesting "Competitors" for the Raspberry Pi
    GeorgeIoak Level 4

    I don't particularly have any gripe with the concept of a usb charger, but the Pi has proven that the implementation leaves a lot to be desired. Mainly down to crap cables and really cheap&nasty psu design, but you really do have to take that crapness into consideration and ideally come up with a better way to supply power..

    I recently finished a module where I was forced to use the micro USB connector because the module was so small and that was the onlu USB connector which would fit. If cost permits, one thing that is high on my personal priority, is for the power to be a high current buck/boost switcher. If you're going to design a board for the masses why not let them be free and grab any old power supply and plug 'er in! The last module I used a nice little switcher that had 2.5V-12V input with up to 2.5A capability, I used that on the main power input. One I also really like about this switcher was that you needed at least 2.5V to get the switcher started but once rolling it would operate down to ~1.9V which made perfect for battery usage.

     

    Yeah, it adds a little to the cost but the benefit in this case seems to far outweigh the cost.

  • 174. Re: Interesting "Competitors" for the Raspberry Pi
    John Beetem Level 15

    George Ioakimedes wrote:

     

    I don't particularly have any gripe with the concept of a usb charger, but the Pi has proven that the implementation leaves a lot to be desired. Mainly down to crap cables and really cheap&nasty psu design, but you really do have to take that crapness into consideration and ideally come up with a better way to supply power..

    I recently finished a module where I was forced to use the micro USB connector because the module was so small and that was the onlu USB connector which would fit. If cost permits, one thing that is high on my personal priority, is for the power to be a high current buck/boost switcher. If you're going to design a board for the masses why not let them be free and grab any old power supply and plug 'er in! The last module I used a nice little switcher that had 2.5V-12V input with up to 2.5A capability, I used that on the main power input. One I also really like about this switcher was that you needed at least 2.5V to get the switcher started but once rolling it would operate down to ~1.9V which made perfect for battery usage.

     

    Yeah, it adds a little to the cost but the benefit in this case seems to far outweigh the cost.

    Yes, I very much like the idea of a wide-range voltage input and a switching regulator.  So many RasPi issues are due to its needing a precision external power supply.  All you really need is a barrel connector and a protection diode (if it's not already built into the switcher).

     

    Regarding RTC, I like the idea of having a non-populated SOIC so customers or VARs can add them later.  My favorite is the Maxim/Dallas DS1338DS1338.  The DS1375 you listed is cheaper, but it needs to get a clock from somewhere and doesn't have battery back-up, kind of defeating the purpose of an RTC chip.  The SoC's RTC may be usable, but they usually draw a lot more battery current than a specialized part like the DS1338DS1338.

     

    Regarding GPIO headers, I suggest following Cubie's lead and using 2mm connectors.  They're almost as convenient as 0.1" pins but only use 60% of the area.  IMO 0.1" connectors are so 20th Century.  JMO/YMMV

     

    For JTAG, I think the Tag-Connect system is so cool.  I'd really like to try it some day.

  • 175. Re: Interesting "Competitors" for the Raspberry Pi
    GregC Level 7

    Hello Morgaine

    i.MX6 and Sabre-Lite have been officially launched during Electronica previous week ... Thank you for your notifications concerning the Freescale web-page updates.

    So could you finally download the Reference Manual and the Datasheet of the i.MX6 version, you are interested in ?

  • 176. Re: Interesting "Competitors" for the Raspberry Pi
    morgaine Level 15

    Yes thanks Greg, the Dual-Quad RM was at the link provided by John, and the Solo-Dual-Lite RM was similarly located.  Linking them both here for easy reference:

     

     

    Morgaine.

  • 177. Re: Interesting "Competitors" for the Raspberry Pi
    jamodio Level 13

    John Beetem wrote:

     

     

    For JTAG, I think the Tag-Connect system is so cool.  I'd really like to try it some day.

     

    I use TAG-Connect on several products we designed with Renesas and Microchip MCUs. It is good for quick programming/testing, not good for development and extensive debugging.

     

    From a product stand point good option with zero cost on the BOM.

     

    -J

  • 178. Re: Interesting "Competitors" for the Raspberry Pi
    liveandlearn Level 1

    Personally im wondering what others would use for power. 64 cores should increase that quite a lot.

    the pi uses what 2,5watts? ok haveing a mini supercomputer could be usefull of course but imho. For most of the uses maybe 1or 2x 1-1,2 ghz and 1gb of ram is the max i would need. i mean most of the learning aspects and retro stuff don't even ask that much. it already runs 1080 video. basicly to browse a little smoother that little more power would be usefull.

     

    i'd kinda was thinking of getting smaller and even less power consumeing. use microsd instead of sd. maybe microhd and micro usb ? but wondering if that would be more practical. but how about an onboard bluetooth. (usb bluetooth is already very tiny) and maybe a tiny wifi card?

     

    but im stil quite new in this. but this is what i was thinking.

     

    but that my opinion.

  • 179. Re: Interesting "Competitors" for the Raspberry Pi
    wallarug Level 15

    The Parrella Board is now available for sale I believe.  They sent me an email saying that they are taking pre-orders.  Here is the link: http://shop.adapteva.com/

     

    And you can follow the project here:  http://www.parallella.org/

     

    I think this could a very interesting project to follow over the next 6 - 12 months.