Shop OBEX P1 Docs P2 Docs Learn Events
P2D2 - An open hardware reference design for the P2 CPU - Page 9 — Parallax Forums

P2D2 - An open hardware reference design for the P2 CPU

1679111238

Comments

  • jmgjmg Posts: 15,173
    ... The ground plane mid layer might be useful for spreading the heat a little better with standard copper but in the end it will need to be dissipated properly.
    Of course, final heat removal is always needed, but what extra copper buys you, be it 4 layers of 2oz or even 2 layers of 2oz, is 'peak removal' - the P2 package is just ~10% of the area of the P2D2 board, so if the power can be spread over 10x the area, the peak temperature of the P2 die, is lowered.
    We've used 2oz and even 4oz on boards over the years, to good effect, to disperse the heat and lower the voltage drops.
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-09-30 08:46
    I decided to mount the XCL220 1.8V switcher on my test PCB that is awaiting just the P2 chip. The pads looked awfully small but I just put down a thin layer of flux and then gingerly applied some paste from the syringe then took my chances with heating it up gently with hot air from a gas iron. When I plugged in my USB serial which supplies 5V it seemed to work so I connected up two 1R/2W resistors in series and connected to the closest 1.8V cap on the CPU side with thinner kynar wire so I wouldn't stress the cap mechanically or thermally. The resistors got hot, the USB supply was down to 4V but the 1.8V was good.

    VOLTAGE DROP
    Next I measured the voltage drop from the regulator output cap to the load and read around 16mV drop but that is at 800ma current. The XCL220 was mildly warm to the touch but not uncomfortably so. Measuring that with a gun it was around 40'C peak which sounds about right. The scope reports 112mV noise average measured from the center of the thermal pad (pin in center) to the loaded decoupler cap but most of the noise seemed to be around 40mV pp at a frequency of around 6.5MHz although the datasheet says this part switches at 3MHz but also says f OSC (MHz) = V OUT (V) / (V IN (V)×t on (ns)).
    I might put a simple ferrite bead in the 1.8V line between the isolation vias to see if that helps.

    EDIT: In fact the voltage drop from the regulator cap to the internal 1.8V "rail" was only 7mV since the load itself was connected to one of the decoupling caps. Having a copper layer for 1.8V doesn't seem worth it but I will have to check the noise when I get the P2.
    EDIT: With the ferrite bead in series with the 1.8V supply I read 40mV avg noise, so it is a good idea to incorporate a footprint for this and it is also useful for current measurements when it is removed etc.

    Once I get a P2 I will measure and post some regulation waveforms while running various tests.
  • Thats sounding really good, Peter. Great experimenting

    I've been regarding the xcl220 on my desk for a while, looking for courage to have a go at soldering it. Did you take an efficiency figure? Datasheet seems to have efficiency in low 90's at 800mA, but its a bit hard to read the graph exactly

    regards
    Lachlan
  • Tubular wrote: »
    Thats sounding really good, Peter. Great experimenting

    I've been regarding the xcl220 on my desk for a while, looking for courage to have a go at soldering it. Did you take an efficiency figure? Datasheet seems to have efficiency in low 90's at 800mA, but its a bit hard to read the graph exactly

    regards
    Lachlan

    Glad to see you were scared too :) What with the thermal ground pad and the inductor pads, those other 6 pads looked a bit too small especially since the part covers those pads.

    I did some more measurements and calculated that it seemed to be around 75% efficient which sounds a bit low but looking at the graph it does seem that once it gets up to that kind of current that the efficiency drops off. However it should be efficient at 600ma or less but I will verify that myself.

  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-10-01 04:11
    I've updated P2D2 rev H with a few extras including option to use internal layers or just double-sided. There's also an extra edge with a reset button (either side) and larger LEDs plus external power and Prop plug although this section can be safely trimmed off too. Other changes are:
    RESET pullup
    Large 3mm center hole
    Ferrite bead for 1.8V supply

    There's no room for extra ground posts as I usually place a 2-pin header down for probes but there are so many ground pads that will take pins anyway. (although I could probably extend out the other edge just for this purpose. I also figured that there is such little need for a special transistor reset circuit just for cheap ebay USB serial that it would be easy enough to have a tiny 4-pin adapter pcb that I will do along with my version of the Prop Plug that includes high-speed serial and powers the board too. Use a Prop Plug, use a cheap ebay USB serial with the adapter that plugs straight into it or my enhanced high-speed Prop Plug with the latter also powering the P2D2.

    Bear in mind that this is a module, a bit like a Flip chip, it is designed to be plugged into your matrix board or pcb just like a big chip but with all the important stuff already on it, just add water.
    I will post up my dev board that this plugs into next. BTW, I plug my module in upside down.

    P2D2H-2.png
    P2D2H-3.png

    There is also the internal power layer just under the ground plane.
    1017 x 774 - 688K
    1324 x 892 - 74K
  • Hi Peter Jakacki

    Just awesome!

    During the past months I know I've said something in the same line, but I feel the need to do it again...

    You are a real artist, who happens to express yourself thru whichever you intend to design, including, but not limited to your hardware and software creations.

    Henrique
  • Looks nice Peter.

    Very glad you made that new PCB edge trimmable as I'm slightly space constrained in that location.

    I was just wondering if it also makes sense to bring out the 1.8V to one of the bottom pins currently allocated to 3.3V or 5V instead for anyone who chooses to feed in the 1.8V from some external supply solution or if they want access to the voltage for powering other low voltage peripherals and/or testing the board...?

    May be nice to show the pin numbers on the bottom of the board for the reverse mounting installations for probing, though it may affect the clean look you have. Needs a fun logo or something too, LOL.

  • Henrique - Thanks heaps for the encouragement but I'm just another "cog" in the forum doing their bit :)


    BTW - I changed the color of the board in the previous post!

    @rogloh - yes, good idea to add the pin numbers on the reverse as well and I will look at bringing out the 1.8V too!

    1017 x 774 - 632K
  • Looks good Peter

    There's a copper stub just above the silkscreen '06' on the top layer. I'm pretty sure its a copper island but just in case its a disconnected track...
  • Now it's Sparkfun reddy. :lol:

    What does the middle pin of the new 3 pin power header do?
  • A work of art Peter!
    Seems a shame to populate it with parts :lol:
  • jmgjmg Posts: 15,173
    .... I also figured that there is such little need for a special transistor reset circuit just for cheap ebay USB serial that it would be easy enough to have a tiny 4-pin adapter pcb ..
    I dropped a SOT23 footprint on the back, keeping the PCB the same area, but with your extra PCB area, it can fit in there easily ?
    Another tiny board, is just something else to lose..

    ...... that I will do along with my version of the Prop Plug that includes high-speed serial and powers the board too. Use a Prop Plug, use a cheap ebay USB serial with the adapter that plugs straight into it or my enhanced high-speed Prop Plug with the latter also powering the P2D2.

    Do you mean HS-USB ? Which HS-USB part does that use ?

    A while back you mentioned an option of decoupling caps on the back, is that still planned ?
    They seem to fit neatly next to the 1v8 vias. The design keeps the top-side decoupling.

    Given the P2 Icc numbers Chip has reported, the original Microchip SOT89 regulators are not really going to cut it, but some Linear regulator option will allow lower noise analog designs.

    I think a couple of S15GLWHRVG diodes on the rear (or even off board, so their heat is elsewhere), can allow a LDO linear regulator to be used, by dropping the voltage overhead.

    That would just mean a separate feed to the 1.8V regulator is needed.

    I'm studying Linear regulators to find ones that have Power Good (On Chip's list), good thermal specs, small packages, and ideally wide supply tolerance. ( and low prices... )
  • roglohrogloh Posts: 5,833
    edited 2018-10-06 03:10
    Cheesy PCB logo.
    EDIT [image deleted]
  • The Microchip part numbers are really only placeholders for better parts in compact SOT-89 packages. I ended up using AZ1117CRs for the 3.3V supplies. Anyone who is going to fully load the 1.8V regulator, and then fully load the I/O, and then run it at 280MHz is welcome to extreme stress test all they like. I just use it for "practical" applications :)

    @rogloh - LOL - hey, that's great!
  • jmgjmg Posts: 15,173
    edited 2018-10-01 05:29
    cgracey wrote: »
    Late to this thread, but I have a few suggestions, based on our experience yesterday. Maybe this is just a checklist for what Parallax needs:
    <snip>
    3) Make it twice as big, or 4x, even. Get some distance between parts and use way stronger voltage regulators.
    4) Go four layers and have a dedicated internal or bottom ground plane for heat dissipation.
    5) Put a power connector and power switch on for proper power cycling and startup.
    6) Add a 1.6V brownout detector that pulls RESn low.
    7) Put some jumpers in for opening up VDD and VIO(s).
    8 ) Centralize 3.3V VIO generation and have C-L-C filters on sets of 4 or 8 pins with a jumper on each supply side.
    9) A nice GND post would be good.

    Chip,
    On Regulators, do you prefer fixed 1v8 / 3v3 regulators, with a smaller BOM, or do you prefer Adjustable Regulators, that add 2 resistors to each footprint.
    The latter is more flexible, but costs a little more in BOM and PCB area..

    Power Good output seems to be available, do you want that on both 3v3 and 1v8 supplies ?

    What happens in a P2, if the rails come up separately, or if 1V8 applies with no 3v3, or 3v3 is applied with no 1v8 ?
  • Cluso99Cluso99 Posts: 18,069
    edited 2018-10-01 05:45
    rogloh wrote: »
    Cheesy PCB logo.

    Love it :)
  • LOL Roger!
    Ties in perfectly with "Use the FORTH"
  • ozpropdev wrote: »
    LOL Roger!
    Ties in perfectly with "Use the FORTH"

    Hah, you just outdid yourself with that one Brian.
    :D
  • P2D2H-4_02.jpg
    925 x 849 - 274K
  • Cool. Somehow it kind of works for me, in a fun way. If you do end up wanting to use it and need the original image to play around with, I took it from here (free license it says) and just modified it using very basic tools in Mac Preview. If required I'm sure you can get a cleaner result with decent image editors, though it doesn't look too bad on your render.

    https://icons8.com/icon/42656/r2-d2

  • @rogloh - yeah, I just grabbed the image and scaled, rotated, and shifted it into perspective in gimp. Obviously the original if it is in vector graphics would be even better so I downloaded the svg from your link, thanks!

    How does everyone feel about this logo, it's so cute but appropriate?
  • The logo is perfect. If license permits, why wouldn't you.
  • Cluso99Cluso99 Posts: 18,069
    Yeah. Love it :)
  • Do it Peter! It is just great!
  • jmgjmg Posts: 15,173
    rogloh wrote: »

    I was just wondering if it also makes sense to bring out the 1.8V to one of the bottom pins currently allocated to 3.3V or 5V instead for anyone who chooses to feed in the 1.8V from some external supply solution or if they want access to the voltage for powering other low voltage peripherals and/or testing the board...?
    It would also be useful to have separate pin access to the 1.8V regulator Vin point, as that allows users to control their power envelopes with external pre-regulation, but in a safer way than a 'Connect bench supply to core 1.8V bus...' That's too easy to fritz the P2.

  • jmgjmg Posts: 15,173
    There is also the internal power layer just under the ground plane.

    Looking more closely at that Power plane, that seems to make these items from Chip impossible (by bonding the VIO pins hard to an inner plane) ?
    chip wrote:
    7) Put some jumpers in for opening up VDD and VIO(s).
    8 ) Centralize 3.3V VIO generation and have C-L-C filters on sets of 4 or 8 pins with a jumper on each supply side.

    The existing bottom layer track design of fingers to each pin, allows for some user changes, so I'd suggest making the 2nd inner plane solely 1.8V (which is by far the highest current draw)
    ( or maybe a mix of 1.8V and GND, if that helps cool regulator areas)
  • roglohrogloh Posts: 5,833
    edited 2018-10-02 00:08
    Actually Peter reading more I think that original icon license has some terms in order to use it commercially in a product and/or if modified. So you will have to check that yourself. But it wouldn't be hard to start from scratch and make something somewhat similar if you think it doesn't permit the use. Here's the license:

    https://icons8.com/license/
    https://creativecommons.org/licenses/by-nd/3.0/

  • cgraceycgracey Posts: 14,204
    edited 2018-10-02 10:24
    jmg wrote: »
    cgracey wrote: »
    Late to this thread, but I have a few suggestions, based on our experience yesterday. Maybe this is just a checklist for what Parallax needs:
    <snip>
    3) Make it twice as big, or 4x, even. Get some distance between parts and use way stronger voltage regulators.
    4) Go four layers and have a dedicated internal or bottom ground plane for heat dissipation.
    5) Put a power connector and power switch on for proper power cycling and startup.
    6) Add a 1.6V brownout detector that pulls RESn low.
    7) Put some jumpers in for opening up VDD and VIO(s).
    8 ) Centralize 3.3V VIO generation and have C-L-C filters on sets of 4 or 8 pins with a jumper on each supply side.
    9) A nice GND post would be good.

    Chip,
    On Regulators, do you prefer fixed 1v8 / 3v3 regulators, with a smaller BOM, or do you prefer Adjustable Regulators, that add 2 resistors to each footprint.
    The latter is more flexible, but costs a little more in BOM and PCB area..

    Power Good output seems to be available, do you want that on both 3v3 and 1v8 supplies ?

    What happens in a P2, if the rails come up separately, or if 1V8 applies with no 3v3, or 3v3 is applied with no 1v8 ?

    Jmg, having open-collector brownout detectors on both VIO and VDD supplies to pull down on RESn is a really good idea. RESn also needs a pull-up resistor to VIO and a pushbutton to GND.

    As switching regulators go, I think the same part for both VDD (1.8V) and VIO (3.3V), but with different resistors on each to set the output voltage would be good. Each capable of 2A would be critical..

    On the other hand, parts which internally sense regulation voltage are safer, because their voltage outputs can't get dangerously off due to conductive contamination or probing mishaps (temporary short causes VDD to go to 10V!).

    This approach is maybe on the big side for P2D2, but would be nice for a larger 4-layer carrier board.
  • Actually that might be an option on this revision to allow for either a 3.3V LDO or the XCL220 3.3V version rather than two LDOs. This version also has the extra mid layers to help spread the heat and lower the noise that might be a problem with extremes but I'd like to think that most of us will not require such a complicated or costly bom just to use the P2. That's why the P2D2 uses double-sided but allows for 4 layers and I will have both versions available.

    Is power-good a real necessity? Is there a problem with not having it?
  • cgraceycgracey Posts: 14,204
    edited 2018-10-02 12:53
    Actually that might be an option on this revision to allow for either a 3.3V LDO or the XCL220 3.3V version rather than two LDOs. This version also has the extra mid layers to help spread the heat and lower the noise that might be a problem with extremes but I'd like to think that most of us will not require such a complicated or costly bom just to use the P2. That's why the P2D2 uses double-sided but allows for 4 layers and I will have both versions available.

    Is power-good a real necessity? Is there a problem with not having it?

    I think it would be important to have brownout detectors on the two power rails to ensure reliable startup and recovery from brownout. Our first BASIC Stamp did not have a brownout detector and, Man, was that flaky. I've put brownout detectors on things ever since.
Sign In or Register to comment.