Shop OBEX P1 Docs P2 Docs Learn Events
P2D2 - An open hardware reference design for the P2 CPU - Page 5 — Parallax Forums

P2D2 - An open hardware reference design for the P2 CPU

1235738

Comments

  • Day 1 :)
    jlcpcb.png
    896 x 189 - 64K
  • Cluso99Cluso99 Posts: 18,066
    Board looks good Peter. Sorry didnt get a chance to call yesterday. You will have the pcb back soon anyway so you can check it out.
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-07-01 07:57
    As I suspected, the pie chart didn't update but the order has already been shipped and I now have tracking!
    jlcpcb2.png

    First off though I was a little confused since I noticed the "quality complaint" and thought they had sent one, but it turned out to be a feedback link. So far, so good..... I will let you know how they turn out.
    1076 x 211 - 69K
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-07-02 13:04
    Well don't trust the JLCPCB job status updates but despite having zipped and clicked my order on a Thursday, this job was completed and sent for shipment over the weekend, then seemed to sit around for a couple of days and is now on it's way with delivery on Wednesday. In the past I've paid a high three figures for this kind of service, for less, and not as fast, but all this for 35 bucks. Go figure.

    If anyone wants one (gratis) then please let me know. It will be easy enough locally but if anyone in the US wants any I'd prefer to ship it as a "document" to one person to look after. I will see how many I can fit into a B5 envelope.

    What am I going to do with mine? Well I want to check the quality and also the "sliceability" where I chop off the headers on some boards to see which way gives me the best edge for castellated pads. I will also try a few different kinds of header connectors too.

    Then I intend to make up one unit and test the power supplies with dummy loads, and then hook up an FPGA board to it and make sure the board works correctly. If it looks good I will also order in some XCL220 1.8V switcher modules and run a 1A load off it over the center pad so I can check out the thermal profiles. There's plenty of time for me to make necessary changes but since they are so quick to get made and delivered, I won't need to order any more boards until I know more about P2 chip availability. Now that will be an agonizing wait.....


    jlcpcb-2.png
    746 x 154 - 23K
  • If anyone wants one (gratis) then please let me know. It will be easy enough locally but if anyone in the US wants any I'd prefer to ship it as a "document" to one person to look after. I will see how many I can fit into a B5 envelope.

    Peter,

    I can take on the US distribution. I'll start a new thread for US members as to not pollute this one.
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-07-04 00:55
    The pcbs arrived while I was having breakfast, so that's less than 6 days since I sent the files!

    Check out some unboxing photos.

    I noticed up close that some of the finer outline overlay wasn't printed around the caps, but I think I just made those track widths a bit too fine. In fact it looks like their software trims the overlay when it gets too close to a pad so the pin header numbers are unreadable. That's a little cosmetic to fix next time.

    Next step is to make up a board and I will order the switchers from Mouser today so I should have them on Monday.

    In the meantime I will make up a couple of packages to send off :)
  • I can't believe you got 50 for $16. Why did I just pay over $80 for 4?
  • David Betz wrote: »
    I can't believe you got 50 for $16. Why did I just pay over $80 for 4?

    Well it was more like $35 with DHL Express which translates to about $50 AUD but still really cheap. I think if I only wanted 10 boards I might have saved a few bucks and then again, for a few bucks more I could have 100.

    The thing to watch with my pcb next time is the clearance from the id overlay to any opening in the solder mask as in the past I could print an overlay over pads if I wanted.

  • jmgjmg Posts: 15,140
    I noticed up close that some of the finer outline overlay wasn't printed around the caps, but I think I just made those track widths a bit too fine. In fact it looks like their software trims the overlay when it gets too close to a pad so the pin header numbers are unreadable. That's a little cosmetic to fix next time.
    The thing to watch with my pcb next time is the clearance from the id overlay to any opening in the solder mask as in the past I could print an overlay over pads if I wanted.

    Yes, overlay over pads is generally frowned on, and these days it is common to print the PADS white/clear on the silk screen to make sure there is zero overlay encroaching onto the solder areas.

    KiCad even has a plot time options called 'Subtract Solder mask from silkscreen' that does this.

  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-07-04 04:02
    I've made up a board and updated the P2D2 album which had the unboxing.
    BTW, I dug out some really old solder paste just for an exercise, fitted a new nozzle, squeezed it out until it looked oily from flux and dabbed that around the pads rather sloppily. Placed the parts, popped it into the preheated toaster oven on max for 4 mins and..... perfect! (except I noticed I left off a couple of the decoupling caps on the corners of the P2). No tombstones or parts moved. The photos show how I used a few layers of foil and a second tray as this helps to stop the bottom of the board, which in this case is bare, from getting too hot. It also helps to cool the pcb when the foil mat is removed from the oven so it is ready to handle in around a minute.
  • Cluso99Cluso99 Posts: 18,066
    Excellent Peter :):):)
  • I may as well send some extra components such as the microSD socket and some Flash chips for those who want them and can use them. I should have a good supply of most parts, except for maybe the P2 :) plus I need to order the switching regulator chip. I ended up putting low profile socket headers on the top side since this still allows me to plug the board in upside down onto pin headers or insert the short end of standard pin headers so I can attach probe leads to this pins. I don't have any 12MHz TXOs yet but I do have plenty of the 10MHz SMD crystals but they are only useful with a P2.

    I just added the photo of a board with those low profile header sockets and header pins stuck into one side to the album. I prefer to have pins on main boards and sockets on modules since pin headers are very cheap plus modules pack better without pins and there's nothing to bend.

    Is there any sense in supplying a kit of parts? For myself I'm making up boards for testing the pcb and components etc as well as providing the breakout board with SD and Flash and Prop Plug header for an FPGA board but I may do a red production version with a few changes that I will get made up and assembled once I have tested a real P2.

    Once I get some of those switchers I will bond a dummy load to the center pad and let it draw 1A at 1.8V for 1.8W of heat but I may also split that dummy load up via MOSFETs driven by some I/O to simulate load transients.

    But if you are excited about getting hold of a real P2, get a P2D2 pcb but if you are not excited, then get a P2D2 pcb.

    Who here in Oz wants a board? I can send one to Cluso99 and some down to the Parallax Universe, Melbourne.
  • PublisonPublison Posts: 12,366
    edited 2018-07-04 13:52
    I have about a dozen request for US and Canada, (that includes one each for Chip and Ken). If you want to send available support chips and SD header in bulk, I'll kit them out for each order. I have a new oven on order, so this will be a good test.

    EDIT: Board is looking great so far!
  • Oh yes please, Peter. We're probably right for most parts so just the flat pcbs would be useful.

    I have a thermal camera to look into the heat distribution. I asked Chip earlier about the earlier bricked P2 chips, and whether we could use these to inject heat, but these don't have the exposed pad. I'm sure there is another way
  • Publison wrote: »
    I have about a dozen request for US and Canada, (that includes one each for Chip and Ken). If you want to send available support chips and SD header in bulk, I'll kit them out for each order. I have a new oven on order, so this will be a good test.

    EDIT: Board is looking great so far!
    I guess I'd like one. I'll have to hope someone offers a service for populating it once the P2 becomes available.

  • PublisonPublison Posts: 12,366
    edited 2018-07-04 14:46
    David, I'll try and get a couple of extra boards to test my new oven with the parts. I am hoping Peter will give a full BOM once he tests the regulator he has on order.
  • jmgjmg Posts: 15,140
    Tubular wrote: »
    Oh yes please, Peter. We're probably right for most parts so just the flat pcbs would be useful.

    I have a thermal camera to look into the heat distribution. I asked Chip earlier about the earlier bricked P2 chips, and whether we could use these to inject heat, but these don't have the exposed pad. I'm sure there is another way

    Interesting point, I wonder if OnSemi make anything at all in the large exposed pad TQFP100, that might be used to spread 2W ? or anyone ?
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-07-04 23:57
    In regards to having a dummy P2 to test the heat spread, I think that would be nice but I intend to use some smd resistors in parallel (maybe 22Rx12+2) soldered around and onto the ground pad to spread and conduct the heat and maybe cover this load in a suitable epoxy or some kind of lid. I don't have to be precise but besides checking the thermals I will also be testing the load transient response of the regulators. The extra couple of resistors there will allow me to overload the regulator for short periods to get closer to the 2W mark but this can also be handled by loads on the 3.3V side too so I might even just make it 22Rx12 + a couple of resistors on the 3.3V side.
  • cgraceycgracey Posts: 14,133
    jmg wrote: »
    Tubular wrote: »
    Oh yes please, Peter. We're probably right for most parts so just the flat pcbs would be useful.

    I have a thermal camera to look into the heat distribution. I asked Chip earlier about the earlier bricked P2 chips, and whether we could use these to inject heat, but these don't have the exposed pad. I'm sure there is another way

    Interesting point, I wonder if OnSemi make anything at all in the large exposed pad TQFP100, that might be used to spread 2W ? or anyone ?

    I'm pretty sure ON Semi does not have anything. This was a special accommodation they made to use the exposed-pad Amkor package.
  • jmgjmg Posts: 15,140
    cgracey wrote: »
    I'm pretty sure ON Semi does not have anything. This was a special accommodation they made to use the exposed-pad Amkor package.

    Let's hope Amkor do not drop that package !
    Peter's approach to load the areas thermally with a number of resistors sounds good enough for now.

  • Cluso99Cluso99 Posts: 18,066
    While I personally dislike the exposed pad, it does have big advantages.
    1. Being the ground connection for the chip, it makes routing easier than if they were pins spaced around the chip.
    2. It also acts as a heat sink.
    Hey, guess I should change my opinion ;)
  • Cluso99 wrote: »
    While I personally dislike the exposed pad, it does have big advantages.
    1. Being the ground connection for the chip, it makes routing easier than if they were pins spaced around the chip.

    I did not realize that the only ground connection was the exposed pad until I read your post and then looked again at the chip pinout. Wow, I have never seen that done before on any exposed pad IC package. There are always ground connections among pins and it's more common that the exposed pad is nothing more than a heatsink so it doesn't matter what it connects to.

    As for the TQFP100 with exposed pad in general, it is very uncommon. I still feel that once production chips are qualified, a QFN package is a must have.

  • jmgjmg Posts: 15,140
    As for the TQFP100 with exposed pad in general, it is very uncommon. I still feel that once production chips are qualified, a QFN package is a must have.

    Does any QFN package come with the correct size (large) paddle area ?
    The difference between eTQFP100_0.5mm and QFN100_0.5mm is not that much, just some gull lead length.
    Looks like there may be a 100-VQFN (12x12) - Microchip have one with 8.12mm paddle and 0.4mm lead pitch

  • TubularTubular Posts: 4,620
    edited 2018-07-05 12:03
    Ti and NXP, eg TMS320F28069PZP

    The exposed pad is smaller on that, however.

    I've had luck looking for HTQFP-100
  • jmg wrote: »
    Does any QFN package come with the correct size (large) paddle area ?
    There may be a variety of QFNs with appropriate pad sizes as a QFN-100 for a straight across comparison.
    jmg wrote: »
    The difference between eTQFP100_0.5mm and QFN100_0.5mm is not that much, just some gull lead length.
    The largest advantage is that of assembly. The "lead pads" and thermal pad are in the exact same plane on a QFN, while the lead foot and thermal pad on a QFP are not in the same plane. This can present issues depending on restrictions of stencil thickness that may be required for other parts on a design.

    After taking another look, I would say that a BGA package with optional heatsink would be the best overall package. 144 ball, 12x12 BGA would be perfect. a 6x6 ball pattern in the center for thermal purposes still leaves 8 free balls in the package.

  • roglohrogloh Posts: 5,122
    edited 2018-07-07 03:33
    Who here in Oz wants a board? I can send one to Cluso99 and some down to the Parallax Universe, Melbourne.

    Hi Peter,
    I'm definitely up for a board or two if you have any remaining. They look great.
    Cheers,
    Roger.

    Ps. just noticed something pretty cool about your board. On its serial header there is a signal (V+) fed via the VCC rail through a 1 ohm resistor R6, probably designed to feed power to/from the P2D2 at this voltage. When assembling it should be easy for us to optionally omit this 1 ohm resistor default placement and instead connect it to a 3.3V rail nearby on the 0.05inch header. That will allow ESP-01 type boards to be provided power directly from the P2D2 board (assuming current limits are not exceeded with the regulator choices) and in time that could potentially support wireless debug/downloads and resets of the P2 just with a simple custom cable harness, software permitting of course. I like it!
  • roglohrogloh Posts: 5,122
    edited 2018-07-07 03:36
    Peter, just for interest, here's where your P2D2 board with a P2 would get first used by me - to replace a dual P1 system I have built for video/kbd/mouse control of an old Z80 machine (a Microbee). The P2 itself would be ideal with its generous internal hub RAM, and sufficient IO pins for the requirements. Your board is a perfect fit here really and I love the green PCB for the full retro look from the 80's I've been aiming for as well. Admittedly the SD card is a little anachronistic but I do have that 7805. :smile:
    1280 x 960 - 205K
  • YanomaniYanomani Posts: 1,524
    edited 2018-07-07 05:00
    jmg wrote: »
    cgracey wrote: »
    I'm pretty sure ON Semi does not have anything. This was a special accommodation they made to use the exposed-pad Amkor package.

    Let's hope Amkor do not drop that package !
    Peter's approach to load the areas thermally with a number of resistors sounds good enough for now.

    As a follow-up; where there is any commercial interest involved, there will be a solution provider, eager for profit.

    Althought Amkor isn't by any means alone in the market, up to this moment I couldn't find another offering of a such large exposed pad as P2 will use.

    But, for the sake of completeness and due to the lack of similar (read as 'available') information at Amkor's site, I'm linking some info I could find at three of its competitors web sites.

    Statschippac's Exposed Pad LQFP web page -> cross section shows a possible embedded ground ring, surrounding the chip die. Despite the fact that such a solution tends to shorten the (costly) wires, that connect the GND pads (if they where positioned at the borders of the chip) to the bottom exposed pad, I'm unsure if a large die as P2 has could be accomodated within such leadframe/molded body limits.

    statschippac.com/~/media/Files/Package%20Datasheets/LQFP.ashx

    Shinko's Enhanced thermal performance Leadframe web page -> two images at the bottom of the page (worth a two thousand words) showing leadframe options; left-one shows the leads internal protrusions, the embedded ground ring and the exposed pad, in a three-step setup; right-one shows the leads and the exposed pad, in a two-step setup only.

    shinko.co.jp/english/product/leadframe/etp-lf.html#link3

    Due to the size of P2 die, I supose that Amkor's current offering would closely ressemble Shinko's bottom right image, but it's only a guess of my part.

    Spil is another option, but as Amkor's site, it's a little jealous when it comes to their products details.

    https://spil.com.tw/products/?u=8*0*4

    Sure, this is not a compreensive listing of players sharing that market, but it shows some plausible alternatives.

    Henrique
  • Just to keep OT

    If some (perhaps many) of you are a bit curious about why Amkor does offer a such large exposed pad in its 100-pin, 14mm x 14mm LQFP package, perhaps the following links could explain that fact, but, as usual, it's just a guess of my part....

    https://c44f5d406df450f4a66b-1b94a87d576253d9446df0a9ca62e142.ssl.cf2.rackcdn.com/2018/02/ePad_LQFP-TQFP_DS231.pdf


    https://c44f5d406df450f4a66b-1b94a87d576253d9446df0a9ca62e142.ssl.cf2.rackcdn.com/2018/02/FusionQuad_DS587.pdf

    https://patents.google.com/patent/US7211471

    Why I've come to the above conclusion? I'll explain...

    During my search for leadframes, exposed pads and alike, I've summed 1 + 1 + 1 and got a result greater than 3 that I couldn't reject, by principle.

    That leads me to the conclusion that someone (Amkor) is leveraging someway from the fact that it needs to produce a bottom exposed metal area that could be sawed (according to its own patent) to generate its FusionQuad® Technology, 176-lead, 14mm x 14mm body-sized package offering.

    Thus (fiat lux) the LQFP, 100-lead, 10.3mm x 10.3mm exposed pad was born, just by chance.

    As I said above, just a guess... :smile:

    Henrique
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-07-09 01:45
    rogloh wrote: »
    Peter, just for interest, here's where your P2D2 board with a P2 would get first used by me - to replace a dual P1 system I have built for video/kbd/mouse control of an old Z80 machine (a Microbee). The P2 itself would be ideal with its generous internal hub RAM, and sufficient IO pins for the requirements. Your board is a perfect fit here really and I love the green PCB for the full retro look from the 80's I've been aiming for as well. Admittedly the SD card is a little anachronistic but I do have that 7805. :smile:

    Hey, that reminds me of my first PC style board, just before the I started using the Propeller I did a bit with the new ARM chips from Philips (NXP). Actually, since it was like a PC without a PC I called it "noPC".
    noPC-1At.jpg

    But when I discovered the Propeller I did a board like this:

    CE0074%20-4068.jpg

    So you can imagine what I would do with the P2! :)

    BTW, stick with USB connectors in place of bulky PS/2 since USB/PS2 device mode is determined by the signaling, not the connector. I probably won't use VGA style connectors as I prefer to have an adapter cable these days.
    1009 x 771 - 561K
    859 x 651 - 210K
Sign In or Register to comment.