Shop OBEX P1 Docs P2 Docs Learn Events
prop galore[512 32 bit processors - 10.2GIPs] - Page 4 — Parallax Forums

prop galore[512 32 bit processors - 10.2GIPs]

124678

Comments

  • jazzedjazzed Posts: 11,803
    edited 2009-04-19 23:34
    The most efficient pins for a high performance 10 bit bus would be 10-19 (or 0-9) as previously articulated.

    Added: never mind shakehead.gif

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


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

    Post Edited (jazzed) : 4/20/2009 12:20:36 AM GMT
  • MicrocontrolledMicrocontrolled Posts: 2,461
    edited 2009-04-19 23:49
    What is a "bus"? Is it a wire that connects to many pins (on other chips) at once and each chip's pin it is connected to has it's own code it has to wait for before recieving data??
    I'm just curious.
    (Man, I'm on a roll tonight with the posting! smile.gif )

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Toys are microcontroled.
    Robots are microcontroled.
    I am microcontrolled.
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-19 23:50
    i agree 10-19 would be better but I would have to go to a 4 layer board to do it that way. This would bring the price up to $30/board with me needing 20 boards minimum.

    or i could make the board physically bigger and bring the price to about $20 to $25 each and keep the 5 board minimum.

    4 layer would be better as I could sheald the traces better from noise and the assembly cost would be lower as they would be full panels instead of individual boards. Would people be willing to pay an extra $15 over the original quated price? And how many want one? don't think we hit 20.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 23:52
    bus pins connect to the same pin on every chip. So current design P5 of all 32 props is connected together. a 10 bit bus does not mean you need to use all 10 at the same time. you could dedicated 2 wires for an i2c interface between prop 1,2,3 another 2 for props .... or use 3 wires for faster sda or how ever you want.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-20 00:07
    Oh. Thank you for clearing that up. I was wondering how you would control so many props with only 32 pins.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Toys are microcontroled.
    Robots are microcontroled.
    I am microcontrolled.
  • jazzedjazzed Posts: 11,803
    edited 2009-04-20 00:14
    Gee ... where is my head? There are only 9 bits available for MOVS, MOVD, MOVI ... never mind about that 10 bit bus optimization.

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


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230
  • WNedWNed Posts: 157
    edited 2009-04-20 00:38
    @Steve - You have really shaken my faith in you as the "Great and Powerful Oz!"
    So we're down to pitching bits out various pins? Keeping them contiguous would still be better for set-and-send... unless it's just not possible, then it would not be better.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 00:47
    there will be 2 halves that are contiguous for sure. will see what I can do.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-20 01:00
    I guess when it comes right down to it, if you predefine a key-mask, one bit pattern is as good as another.

    Ned

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 01:10
    movd is 9 to 17 which would not work to well. it would require me starting over but I could make it 1 contiguous bus from 9 to 19 which is what I think I will do. Back to the drawing boards.


    I wish eagle had a step and repeat option in the pcb. you can cut and paste on the schematic side but if you do that on the pcb side then the schematic is not usable anymore.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-20 01:18
    The bit that is a problem with EAGLE. I have found that out as I am trying to create a board with it.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Toys are microcontroled.
    Robots are microcontroled.
    I am microcontrolled.
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 01:24
    it is a great program but some funny things.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-20 02:34
    To err is human. To forgive is ... well, you know ....

    At least I didn't try to propogate some hoax out of honor or other motivation.
    Even the great wizard was called out. Socrates would make us all look foolish ....

    Eagle does have some weaknesses ... very tedious.
    Can't figure out to save my life how to pour a ground plane.

    @mac, having contiguous bits for a parallel bus "data" is important; less so for control bits.

    On the primary (or master cpu), with 9 bits properly aligned, you can set 8 data and 1 ATN/!DATA
    state in 1 instruction. (Doing the same with arbitrary bits requires 3 instructions. You obviously
    understand this.) Then you can monitor a "ready" ACK/!BUSY condition in some other bit for synchronous
    operations including command phase. Writes can be asynchronous; any read would be synchronous.
    I believe that if ACK is in a low state (!BUSY) read by the primary, the primary can not issue a new transaction.
    On a secondary cpu, all bits are inputs except for the ACK which is 1 bit.

    Someone please, check me on a 10 bit parallel protocol:

    The basic idea of a communications over common 10 bit parallel bus would be an attention or command
    phase and data phase ATN/!DATA. The attention phase gets the secondary device attention by sending
    the destination address in the 8 bit data and ATN pin as high state. The secondary responds by
    asserting !BUSY momentarily ... primary waits for !BUSY to become ACK and sends the packet length
    request. Then the primary sends the packet ... sending a separate packet length from packet allows us
    to be synchronously asynchronous .... Hopefully the attachment illustrates what I mean.

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


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

    Post Edited (jazzed) : 4/20/2009 3:10:38 AM GMT
    569 x 600 - 15K
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 02:44
    to poor a ground use the polegon tool.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-20 04:27
    I managed to fit a 23K256-I/ST serial ram chip in.

    Do pin order mater? P0-SCLK,P1-SI,P2-SO is the easiest for me to layout.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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) : 4/20/2009 5:56:27 AM GMT
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 05:58
    Many people(or the same repetitvly have not checked) have asked for a more modular design. I could make these as strips of 8 props each with a header board to interconnect up to 8 strips. That way you only need to pay for 8 props at a time but can get up to 64 props total per layered level.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-20 06:45
    How would the RAM chip fit in to the modular scheme?
    You would probably be left with far fewer remainders using the more modular design. Because of the surface mount requirement, I'm not likely to do an incremental build of a 32 chip board; I'll order it fully populated or not at all. If not at all, I'd start kluging together that device I spoke about earlier, which would be penny wise and pound foolish, but would allow me a reasonable entry point.
    So, I guess I'm one vote for modularity.

    Ned

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno
  • jazzedjazzed Posts: 11,803
    edited 2009-04-20 07:09
    mctrivia said...
    I managed to fit a 23K256-I/ST serial ram chip in.

    Do pin order mater? P0-SCLK,P1-SI,P2-SO is the easiest for me to layout.
    You forgot Chip Select (CS) ... very important to have connected for this SPI device. If you put a resistor between SI/SO, you could use one pin for that. Order doesn't matter. Is this 1 SRAM per Propeller? Too bad the pinout is not the same as for the Atmel dataflash. I saw your switch but didn't know what it's for. I'm also not sure what value the serial SRAM offers.

    I'm more inclined to buy an 8x assembled FAB that is expandable by attachment even at a greater potential cost rather than a partially assembled 32x FAB mainly because of the chosen package land pattern. I can solder a QFP no problem, but I have doubts about my ability to solder the QFN package.

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


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230
  • dMajodMajo Posts: 855
    edited 2009-04-20 11:13
    It's impressive how this tread went on in two days
    mctrivia said...
    I managed to fit a 23K256-I/ST serial ram chip in.
    No need: look at this new products FM24V10 and FM24V05; they can replace the eeprom (unlimited and same speed as read write cycles + unique serial ID) BTW it will be nice if at least one can be FM31L278.
    Don't forget thak you have 1MB of ram around the 32 props and access through parallel bus is faster than SPI (the new 512/1024Mb ramtron products can go up to 3.4MHz clock rate)
    PostEdit: 3.4MHz is 185KB/s in random access and 375KB/s in sequential


    You must find a solution for the bus: the best is to use P0-P7 (this can improve the performance and also open doors to non traditional way of programming) control lines can be P24..P27. It is always good to leave entire bytes (P8..P15; P16..P23) to the external IO if someone wants to use vga/tv-out.

    Also re-thinking the power supply unit: on such board there is no need for it. For testing "everyone" have a lab adjustable power supply; for production is not so hard to develop the appropiate power supply for the targeted product. IMHO onboard you need only the decoupling caps and the power supply header.


    Post Edited (dMajo) : 4/20/2009 11:22:53 AM GMT
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 12:55
    IO Pins:
    P0-P7(full set available here)
    P8-P9??
    P20-P27
    P28-P29??


    Bus:
    P9-P19


    Yes i is 1 SRAM per chip. I have not forgotten that there is 1M on board between the props but it never hearts to have more options if it does not compromise others. Would not populate the IC if someone wanted they would need to hand solder on the tssop package thereself. If you can find a beter use of the space I can change.

    CS can I not just tie it to ground and leave it always selected.

    As for the power supply I would definitely leave headers after it so you could just use a lab supply. My reasoning is not everyone has big beaffy 10 Amp lab supply where we probably all have one strong enough some where between the 4 and 40V rating(big battery work but require being very careful) At 18V(drill battery) a 1A supply would be more then enough to provide the 5A the regulator can handle.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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.
  • dMajodMajo Posts: 855
    edited 2009-04-20 13:33
    What do you think of the 512Kb/1Mb I2C fram substitutes of the boot eeproms?
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-20 13:44
    I think it a great option. I am on a cell phone but will check package layout when I get off work in 10 hours

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-21 00:36
    if this chip was available in tssop pack it would be direct replacement. cool. though at $15 it is not cheap. I will see what i can do to swap my eeprom to a soic pack instead.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-21 02:15
    Requiring a dynamic CS is an inconvenient issue. CS marks the beginning and end of a cycle in the SPI interface. Various points in the serial SRAM spec mention change in CS terminating the cycle especially for the read and write transaction ... and one instance where high to low change brings the device to active state. Other SPI devices also require a dynamic CS such as the ENC Ethernet and the SD CARD ... so if you want another Ethernet port on one of Propeller matrix for example, you will need CS. I've tried just tying ENC CS to ground ... didn't work. This is a case where "think different" doesn't buy much except trouble.

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


    Propalyzer: Propeller PC Logic Analyzer
    http://forums.parallax.com/showthread.php?p=788230
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-21 02:41
    well I don't have room for the cs pin. so unfortunately I will have to remove the sram. the good news though is the board has lots of ram in the props and the eeprom will be replacable with a 1m fram

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-21 05:07
    Update:

    Headers
    P0-P7,P20-P27,P28-P29 - 18 IO Lines to header.
    P10-P19 - Bus
    P30,31 - Prop Plug Head

    Extra Harware per Prop
    512k EEPROM Default but pin compatible with FM24V10(1M FRAM)
    Optional CHS-01TA Power Switch

    Board Layout
    *8 props per board with 10bit bus
    *I will make a master board that can take up to 8 of these boards with power tap and pads for two 5A power supplies and vertical headers for attaching multiple stacked boards.


    These boards will be going to the manufacturer on May 11th. Will figure out assembly cost closer to the date.

    Will upload diagram of the new layout tomorrow.

    All bus pins are going to vias. if anyone wants to repurpose it is just a matter of cutting traces on bottom side of board.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-22 04:22
    As promised new pictures. Grid is 0.05"

    Final board is step and repeat sideways. 8 of these squares plus on the left a header with bus pins, XI, reset bus(not drawn yet), VSS, and 3.3V

    the prop plug is nestled in the bottom left above the row of 8 pins.

    On the bottom right you can see where the pins for the next prop over would be. There is no gap but there will be a divider line on the silk screen.


    Any questions, or suggestions of adaptations?

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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.
    593 x 533 - 42K
  • WNedWNed Posts: 157
    edited 2009-04-22 14:40
    Man, I almost snarfed my coffee! I counted 'em up, and it looks like about 20 X 0.05" on a side...
    Very nice!
    Have you thought of having a way to secure the whole strip? Looks like there may be space for something like a servo mounting sized screw in the upper right area (not in every inch, but every few inches). At one inch wide, wire ties may be just fine... The most annoying thing about the demo board is its lack of holes, or even decent edges to clip it down. For various reasons, it would be nice to be able to keep this thing still.

    Ned

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "They may have computers, and other weapons of mass destruction." - Janet Reno
  • mctriviamctrivia Posts: 3,772
    edited 2009-04-22 15:01
    I will fit at least 6 screw holes in some where. 2 at each end and 2 in the middle.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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 2009-04-22 15:10
    by the way it is only .8 long each the sides over lap a bit.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    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) : 4/22/2009 3:50:44 PM GMT
Sign In or Register to comment.