Shop OBEX P1 Docs P2 Docs Learn Events
prop galore[512 32 bit processors - 10.2GIPs] - Page 3 — Parallax Forums

prop galore[512 32 bit processors - 10.2GIPs]

135678

Comments

  • mctriviamctrivia Posts: 3,772
    edited 2009-04-17 14:45
    exactly what I was thinking.

    for the busses I will use 10k pull ups so communication can be safely done by altering pin direction like in i2c

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • James LongJames Long Posts: 1,181
    edited 2009-04-17 17:05
    WNed said...
    James - The neat thing about having individual EEPROMs is that with them the Props don't have to be the same, code-wise. By making them hardware identical, you allow them to be code flexible. If you wanted to, you could run them like 32 completely separate Props that just happen to be sitting on one board, or you could have them all sitting there doing the exact same thing. You have the choice.
    Would you be able to assemble individual boards or do you need to batch them?

    We can do individual boards, and panels of boards. Panels are cheaper to do though, less labor involved.

    James L

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    James L
    Partner/Designer

    Lil Brother SMT Assembly Services
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-17 17:17
    for panels can you cut apart? they will not do internal routing so normally I cut apart with band saw or demel. to do a full panel just this would make the minimum 20 boards. I 4 per panel 5 panel minimum

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • jazzedjazzed Posts: 11,803
    edited 2009-04-17 17:22
    Ned,

    The model I described would provide the ability to build and test without having to burn 32 eeproms and would require on SD card or prefereably one ethernet connection ... thus easing the so called "non-recurring" development pain. As you mentioned though, having each resource come up with it's preprogrammed ability would have great advantages specifically in recurring startup time.

    I realized this morning that pid for prop5-cog1 process would not be $51, but instead $29 ... oh well shakehead.gif

    Hey McT, what is a standard panel size?

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    --Steve


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230
  • James LongJames Long Posts: 1,181
    edited 2009-04-17 17:22
    mctrivia said...
    for panels can you cut apart? they will not do internal routing so normally I cut apart with band saw or demel. to do a full panel just this would make the minimum 20 boards. I 4 per panel 5 panel minimum

    Typically we used v-scored panels and "snap" them apart. We can cut them if needed. We typically find parts are too close to the edge to safely cut them.

    V-scored panels are great, a sharp edge of a table, and two hands (sometimes three) and you have individual boards. We even snap them assembled, just have to be a little more careful.

    James L

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    James L
    Partner/Designer

    Lil Brother SMT Assembly Services
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-17 17:41
    my standard panel size is 60 square inch


    they won't v score either.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • James LongJames Long Posts: 1,181
    edited 2009-04-17 17:52
    mctrivia said...
    my standard panel size is 60 square inch


    they won't v score either.

    Wow, you better be getting a fabulous deal if the pcb house doesn't even v-score.

    For $89 bucks (100 sq. in.), even Gold Phoenix in China will v-score (well up to 20 times).

    60 square inches doesn't help much. What are your typical dimensions of the panel?

    James L

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    James L
    Partner/Designer

    Lil Brother SMT Assembly Services
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-17 18:17
    I am getting a great deal. as for the typical there isn't any. I was not planning on panalyzing these by themselves like that because I figure there would only be a few interested. if 20 are interested then I will make as best I can for you.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • James LongJames Long Posts: 1,181
    edited 2009-04-17 18:31
    mctrivia said...
    I am getting a great deal. as for the typical there isn't any. I was not planning on panalyzing these by themselves like that because I figure there would only be a few interested. if 20 are interested then I will make as best I can for you.

    Do these have any PTH parts on them?

    We typically do no through hole parts, but do make exceptions on occasions.

    James L

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    James L
    Partner/Designer

    Lil Brother SMT Assembly Services
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-17 18:51
    only headers are through hole and they will not be populated so people can do what they need

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • James LongJames Long Posts: 1,181
    edited 2009-04-17 18:59
    mctrivia said...
    only headers are through hole and they will not be populated so people can do what they need

    Cool,

    We build a lot of propeller based products for ourselves, so assembling these should be a snap.

    We are working on a project right now that is scheduled to go into full production in the 4th quarter of this year. Testing prototypes should be finished in about a month. If we can prevent another tsunami from our customers. We did more work in 4 weeks (Feb-March) than all of last year.

    As a side note, we typically use lead-free solder. I figure you do not have a problem with RoHS solder. If you have problems with PB-free, let me know. We also do limited runs of 63/37. These are usually from temp restrictions, or customers who can not work with lead-free(lacking the tools to do so).

    James L

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    James L
    Partner/Designer

    Lil Brother SMT Assembly Services
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-17 19:05
    I am tooled for both prefer 63/37 for my own stuff but for sale should be lead free

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • localrogerlocalroger Posts: 3,451
    edited 2009-04-18 03:33
    @mctrivia: seems I misunderstood. But then, if every Prop has its own EE which could identify it by being unique then the ID pins really are kind of redundant. They could at least go to widening the interprop comm bus.
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-18 03:52
    localroger yes you could write seperate code for each prop and would not need need id pins. This would be a bit of pain though if you wanted all with the same code. though i guess if no one is likely to mass produce the pain may be worth the extra bus pins.

    The pole system does not seem to work so here is my poll:

    What is better:
    1) Id pins so same code can be written to each prop no changes necessary
    2) only 3 id pins so same code can be written to each row with only an increment of 1 constant for next row
    3) no id pins you need to program in the id if necessary as a constant in your code.

    I should mention that 3 may not be possible because with 2 layer board there is only so much space to run traces. Also it should be relatively simple to make a programmer that could write the same code to 8 different props at once.

    Thinking about it a little more if you use a 512k eeprom then you could write the id in the top half and the prop plug would not change this so the same code could be used for every one.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.

    Post Edited (mctrivia) : 4/18/2009 3:57:37 AM GMT
  • WNedWNed Posts: 157
    edited 2009-04-18 04:26
    For any of the projects I have in mind, soft address would be best. No id pins required.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno
  • tonyp12tonyp12 Posts: 1,950
    edited 2009-04-18 05:04
    Not that I have much IC design experience (if any), this is how I would do it.

    There will be one start board, that comes with·5 props.

    You can cascade more boards (these boards comes with·4 props)
    For up to 65 props

    There will be One master Prop (on start board), ANY communication would go through this one.

    one 512kb eeprom, all the props get the same program loaded at boot.

    Two Quad 2:1 Multiplexer/Demultiplexer Bus Switch (FST3257) for each slave prop.
    To establish a 8 bit bi-directional communication between master prop and any ONE slave prop.

    8 bit (2 pins with serial to parallel with latch) Prop ID selector.

    At boot, the master prop figures out that he is in charge.
    First he start with a $00 on prop ID pins and sends a handshake value on the data line.
    Wait for a replay,·Tells it so start the program it have in RAM or sends it·new code to run.

    And move on to prop id $01 and so on, up to $3F

    You design the id selector lines so that each cascade moves it up·4 in value.
    So no jumpers to set etc.

    Post Edited (tonyp12) : 4/18/2009 5:40:22 AM GMT
  • jazzedjazzed Posts: 11,803
    edited 2009-04-18 05:12
    Having the individual EEPROMs contain the ID would work. A default COG management program could do primitive ping responses and potentially handle any device open/ioctl/read/write/close service dispatch.

    Looking back at your pinout I see 4 pins 4-7 dedicated to the "COG-BUS" ... it seems one should at least have 8 pins for data transfer, 1 for CMD/DATA phase, and 1 for RDY (did I forget something?). Pins 0..9 are accessible in software from MOVS instruction so that data and signalling can be done with 1 instruction and all bits can be accessed in immediate mode. If you're doing serial inter-prop communications and don't ever expect to need high bandwith (2.5MB/s or more), 4 pins would be fine.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    --Steve


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-18 05:21
    well so far i have 2 votes for get rid of id pins and make the bus as wide as possible(without using up accessable pins)


    because of the chip pin layout P5-9 and P21-P25 are only really usable for dedicated functions. If there is physical space to do so using those pins could give 10 bus lines. Would be nice to use 0 to 9 for easier high speed bus but pins on the top and bottom of the ic are all i can get out to headers.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
    483 x 289 - 27K
  • jazzedjazzed Posts: 11,803
    edited 2009-04-18 05:27
    If 10 pins are on a bus you would only have one 2x10 2x5 or 1x10 header on the end of the bus.
    Pins 10-19 could be used with the MOVD instruction instead of pins 0-9.
    Maybe a .png of your layout is due?


    Never mind shakehead.gif

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    --Steve


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230

    Post Edited (jazzed) : 4/20/2009 12:18:56 AM GMT
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-18 05:32
    will post a png of layout later tomorow(probably around this time). been working on writing random numbers to a sd card.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • localrogerlocalroger Posts: 3,451
    edited 2009-04-18 13:28
    I like putting the ID in the large EE. Has the set it once and forget it advantage of ID pins, but frees up the pins. (Even if you don't run them far, you could run them to vias for jumpering with wire or just use them for intercog comms on the prop.)
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-18 13:37
    true I can definitely get them to vias. sounds like everyone agrees kill the id pins in exchange for soft id.

    a note. if you want to repurpose bus pins for io you could carefully cut the traces on the back side of the board. care is needed as I only run 6 mil clearance

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • WNedWNed Posts: 157
    edited 2009-04-19 03:29
    Just as a gee whiz, by the way... I used to be a simulator technician for Flight Safety, Intl (got my first Stamp, a BS1 while I was working there). Every single light, switch, bell and whistle in a simulated aircraft cockpit goes through the host computer in one way or another. If you really, really need to think of a "practical" application for a board like this... I sure as hell wish this were the compact type of IO board we'd had to work with. Could have even used one board to do all the sound effects, and context sensitive "radio chatter" for Pete's sake.
    Hey McT, there may be more of a demand for this board than you thought...

    Ned

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno

    Post Edited (WNed) : 4/19/2009 3:54:06 AM GMT
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-19 07:14
    Well it is always nice if lots of people want something.


    Pictures were asked for. I am still relatively early in my layout but here is one of the 32 props.

    the whole board would be a step and repeat in an 8x4 grid with power supply and driver placed on one of the edges.

    At this point I have 16 io lines to headers. prop plug reset line is not yet hooked up but i know how i can get it done. so far 5 bus lines have been hooked up but that will be expanded.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
    278 x 564 - 20K
  • jazzedjazzed Posts: 11,803
    edited 2009-04-19 19:51
    Nice layout. Your intentions are clear now.

    It might be time to revisit exactly what the goal is of this project.
    Is it to provide a ton of I/O or a ton of processing power (as your title suggests) ?


    Looks like you have chosen I/O as the priority since many I/O are brought to headers.
    One could use 2 of the interconnect pins as a SDA/SCL interface with one master (primary)
    or a 3-wire serial packet interface RX/TX/GND with a target address protocol ... unicast/broadcast.

    Inter Process Communications (IPC) is a major bottle neck in multi-processor systems.
    The faster the IPC transport the better if you want processing power.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    --Steve


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230

    Post Edited (jazzed) : 4/20/2009 12:20:10 AM GMT
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-19 20:29
    the goal is a happy medium. 16 io and 10 bus lines. still not done wiggling in all the bus lines.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-19 22:27
    you can always wire up the io pins together to make a bigger bus.

    now I need to try and find a serial ram chip I can squeez in and wire up the rest of the bus pins

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Need to make your prop design easier or secure? Get a PropMod has crystal, eeprom, and programing header in a 40 pin dip 0.7" pitch module with uSD reader, and RTC options.
  • MicrocontrolledMicrocontrolled Posts: 2,461
    edited 2009-04-19 22:52
    32 Props?? That's a big board!! Imagine the processing power!! I do agree, however, that you should make a main board with about 4 props on it and a lot of connectors around the edge of the board to fit in even more props! If I was getting the board done I would do this and have one prop for the CPU, one for graphics, one for sound, and one for input. Then you could have a virtual Prop Computer!!! As you added more props you could add more screens, input devices, and sound channels. You could even make a sensor controller one and one for the controlling of servos. Along with all the things parallax offers; the possibilities would be endless!! You could build your own custom 64 prop supercomputer for under $1,000!! If you made each module seperate from the 4 prop main board, than you could just keep adding on modules with sensors and/or sensor plugs in them. Or build a wireless robot with a wireless and servo dedicated prop!! hop.gifsmilewinkgrin.gif

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Toys are microcontroled.
    Robots are microcontroled.
    I am microcontrolled.
  • WNedWNed Posts: 157
    edited 2009-04-19 23:20
    @McT - I am, admittedly, an idiot 2/3 of the time, and only other people get to choose when that is, but I think you've done a terrific job keeping this as general purpose a platform as is possible. If I want to, I can use it for neural simulations by wiring all the header pins together... well, leaving something for user IO or it would be really, really boring... or I can wire it out to every electronic device in my house (I don't think there are more than 512), and begin my evil plan to control the world! Either way, damn, I'm already out of pocket for two boards from you, and I don't even know what they'll cost yet...

    @Steve - the 10 bit bus was your suggestion, why do you sound so disappointed?

    @McT, again, RAM?... Now you're just toying with us...

    Ned

    ...Smile, I'm going to need at least three if I want to do cortical layering...

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno
  • MicrocontrolledMicrocontrolled Posts: 2,461
    edited 2009-04-19 23:32
    Wow, I added it up, and this is going to cost you a fortune!! shocked.gif

    This is why I posted my suggestion of smaller interconnecting boards. I do, though, disagree that you should have one EEPROM that all the props run off of. Each prop should have it's own individual EEPROM that can be programed by it's own header.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Toys are microcontroled.
    Robots are microcontroled.
    I am microcontrolled.
Sign In or Register to comment.