P2QT - P2+regs+Flash+uSD in a 68-pin 0.91" sq Quad "Fat" Pack

Peter Jakacki

Cluso99 posted a bare bones P2 module but I wanted a really small module too with just the minimum number of pins for the 64 I/O, plus Vin, ground, Vout, and reset, so 68 pins and 0.91"sq with regs and Flash etc. However I decided to put the P2 on the reverse just by itself and still have uSD, Flash, and switching regs on the top side. This is my P2QT. Am I crazy? Yes!

I want this board to integrate into a legacy design and also check a new dual 2A switching regulator plus I'm using cap arrays as well to squeeze it up a bit more. It's a wip but one I will get done. So there is no USB or RTC or clkgen options or LEDs etc so this will be an easy board to manufacture. The uSD header faces in just like it does on the P2D2 and the card slides in over the top of the Flash.

The corner pins are reset,power,ground,3.6vout, with 16 I/O on each side. The pcb is 0.91" sq (23mm x 23mm) and might end up 4 layer. If this goes onto a pcb I can also cut out a place for the P2 chip to poke through just to get it flush.

Don't expect to plug this into a breadboard, but you can prototype with it using my microMAT boards with the 50mil pitch dual hole size pads. The P2D2 on the other hand is ready to run when you plug in the USB.

Here's a rough layout and I will edit this later to show the components. If the holes are castellated then it could be surface mounted provided there is a cutout for the P2.

Maciek

Peter, I have long suspected you must be cruel to us, poor souls waiting for your already announced boards. With this anouncement you've just made I got a proof .

JRetSapDoog

He's like Boeing or Airbus laying the groundwork for next generation planes while producing the current generation. Is he crazy? Yeah, like a fox.

It seems like a great minimalistic board; it really is a cutie (QT?). We're going to need that, something that can hopefully sell for $30 or so in quantity one, and something so compact, yet straightforward to connect up with. The features of Cluso's board overlap, but it's good to have options.

I didn't know about cap arrays. Those should be easier to pick-and-place (if doing so manually, I mean). And if wanting to surface- or flush-mount the P2QT, there's a good chance that the folks designing carrier boards for and around it will be able to include a board cutout for the P2 package.

Anyway, 16 I/O's on a side, 17 total pins per side, makes for a nice "fat pack" (but not too fat). A minimalistic board like this is kind of what I was expecting from Parallax. But their Edge board/card is nice, too. And there certainly is a place for it. But it does require a rather large card socket. And at $49 for singles (at least at present), it opens the door for other designs, which is fine as far as Parallax is concerned since it's their chip either way and "the more, the merrier" in terms of boards in the P2 ecosystem.

This board is closer to being a chip than a board, hence the "quad fat pack" in the title.

Tubular

Nice job Peter. Can I be the first customer?

I suspect the castellation is a bit ambitious for the corner pins, if these handle the power. If you shave off along all four sides what you're left with in each corner is just 1/4 of a plated hole

Getting heat out will also be interesting, but hey lets try it and see

jmg

Tubular wrote: »

I suspect the castellation is a bit ambitious for the corner pins, if these handle the power. If you shave off along all four sides what you're left with in each corner is just 1/4 of a plated hole

Good point.
I've seen many edge designs use dual plated holes for that reason, and top and bottom routing to the edge pins.

Cluso99

Nice Peter.
My concern with my original double sided design that was components on both sides was too risky. I've done plenty of P1 double sided designs where the caps were better placed underneath. However, the P2 has that big ground pad underneath which probably means you need to solder it first. Then you turn the board over and solder the jelly beans and hope the P2 doesn't move - unlikely but I decided against it. I'll be interested to see how it goes.

Peter Jakacki

The P2QT (yes it's a cutie) is really a big chip to be used as a component in another system, but it combines all the essentials in a compact "QFP" and it will be as cheap as I can make it. Of course you can mount this QFP onto your own header board or DIP40 etc or integrate it into your P2QT pcb where you don't need to worry about routing all those power and ground tracks and how to supply dual voltages at high currents etc. BTW, I still use the same technique I use on the P2D2 where the switcher produces 3.6V which is fed into a dual LDO for a clean 3.3V. The LDO is enabled by the 1.8V rail so if the 1.8V isn't working then the P2 doesn't get the 3.3V either.

@Cluso99 - I am relying on the P2 and its big ground pad to hold the P2 in place when it gets mounted and reflowed first, and after cooling I will stencil the top side, place components, and then the panel will rest on an insulating and support layer (during the second stenciling too probably). This goes back into the reflow oven and all should be good. If need be I could make up a dummy pcb with cutouts just to support the P2QT panel when it's flipped and reflowed.

@Tubular - I will either have to order these with through holes or get some made as a castellated version. It's true that the corner pads need special attention in that case but I will initially just make them as TH pads.

The P2D2 of course is a complete system especially because of the USB serial among other things.

bob_g4bby

Just thinking aloud - as a former emc engineer, my preference for a cpu 'plugin' for process control would be 4 layer every time + keep the ratio of 0v pins to signal pins as high as you can, else cross-talk between signals and general ground-bounce will be a liability, especially in noisier industrial environments. Keeping 0v return with signal, especially needed at high frequencies. The I/O can really motor , so potentially plenty of really sharp edges. Try and get as much 0v fanning out from the centre heat pad to the edges of the module as possible and avoid long slots in the 0v plane due to lines of holes. Don't shift layer with the 0v plane. Decouple supplies with a mixture of 10n and 1n.

Peter Jakacki

@bob_g4bby - I've been down the track where I tried to do everything with the P2D2 and now I want to do as much as I can in as little space as possible. The 3.6V output is nice but that may end up being a jumper to connect to ground instead. However the P2 chip itself only has a single ground, and that ground goes back to a single corner pin. Internally on the pcb the corner ground is routed separately to the regs from the P2 ground. I will play with the layout to see what I can do, but I don't want to overly complicate it this time.

Tubular

One thing thats possible if you want to do as you say bob_g4bby is to use 1.27x2.54 pin headers pointing radially (else use right angles). One row of the headers connects to all the signals on top layer adjacent to P2 physical pins. Add some high temp insulator below ~ 0.3 or 0.4 mm thick, then solder a big earth ring to the other row of pin headers, and connect to the P2 earth. In the end, you have 68 signals and 68 paired grounds.

You can get screened 0.025" pitch ribbons ok

Spectrastrip make a twist and flat ribbon, but I think i've only seen 0.05" pitch
https://au.element14.com/amphenol-spectra-strip/132-2801-016/ribbon-cable-16way-per-m/dp/1207478?

Yanomani

Hi Tubular

Unfortunately, it isn't at next corner store...


https://iec.net/product/30-gauge-68-conductor-025-inch-pitch-twisted-pair-ribbon-cable-for-scsi-lvd-320/

Peter Jakacki

Maybe I can do more with grounds but the P2 chip itself only has a single ground and somehow we manage. Regardless, I don't want to complicate this initial design. I may end up adding an extra pad just inside two of the corner pads as extra grounds.

The proof of the pcb will be when I use it, and I will take it from there if need be.

bob_g4bby

Measurements on these first (low cost) boards will be what counts and then change if necessary. A plug-in cpu board is bound to come with some compromise over a p2 mounted directly on the main pcb in any case. Just prompting you to look for track inductance effects, if your I/O is going to be fast. The unwanted signal amplitude on any pin depends heavily on mode of use. ( e.g. No-one will switch all the other 63 pins from high to low simultaneously )

jmg

bob_g4bby wrote: »

Measurements on these first (low cost) boards will be what counts and then change if necessary. A plug-in cpu board is bound to come with some compromise over a p2 mounted directly on the main pcb in any case. Just prompting you to look for track inductance effects, if your I/O is going to be fast. The unwanted signal amplitude on any pin depends heavily on mode of use. ( e.g. No-one will switch all the other 63 pins from high to low simultaneously )

Yes, I'd agree that a single corner gnd path on a 100+MHz digital part is 'optimistic'.
A 2.54mm nudge of PCB size would allow a centre gnd pin on each side, & would lower crosstalk effects, and that could free a corner pin for maybe EXTCLK in option ?

ErNa

I think it's a good idea to focus on the P2D2 et al and not to open a new box. In the meantime anybody can make proposals but don't distract Peter. Peter has to serve the orders already in for a product that hopefully will not be out when ready. We can not stand the time to the next election! without going gracey outside of GOC!