Espruino Pico (next revision)

Hi,

So as I'd said on the last KickStarter update we're having to re-make a bunch of Espruino boards, and as part of that:

• Part outlines are being tweaked for less tombstoning/higher yield
  
• I'm taking all the USB and BOOT0 pins of all the Picos onto the panel that they're made on. It'll mean that all of them can be programmed and tested at once, which'll really improve matters.
  
• Jaltek are doing something about the breakout points so that the boards can be broken out more easily. It may result in the two 'rat bites' at the end protruding a mm or so more though :(
  
  

But as I guess a bunch of you have got your Picos now, do you have any thoughts about things that could be tweaked in this new revision?

Obviously I can't change things massively, and I don't want to tweak the pinout or placement, but I could for instance improve the silkscreen on the rear (I think it needs it).

Personally I'd like to add a polyfuse, but I don't see how I could fit one in.

The issue with break out is the thickness of the board. Did you ever think of doing it differently, suchas making a thinner board and snapping something on to create the thickness? It may also have an inpact on the cost of the overall bard.

As I understand from this process is, that after the Picos have been tested, the milling will happen to cut off the utility connections - is that correct?

Yes, I'd considered using a thinner PCB and something to make it thicker (maybe adding a piece of thin plastic on the bottom) but it's a lot of extra manufacturing work (as it couldn't really be automated), and actually I'm not sure it'd help the break-out that much. One of the big issues is that the break out is so near the solder joint for that last 0.1" pin.

Personally I think it could be overcome in a number of ways:

• CNC mill the break-outs away
  
• Use a laser cutter to cut break-outs (maybe not a normal one, but cutters do exist for just that)
  
• Get a different break-out tool with a very thin edge (their current one has probably a 2mm thick edge, which is causing the issue)
  
  

But sadly Jaltek aren't really very interested in what I have to say and would much rather I change the design than they change what they're doing :)

after the Picos have been tested, the milling will happen to cut off the utility connections - is that correct?

Actually I think the connections will run across the rat bite on the USB end of the board. When the board is snapped out I think they'll just snap out with it - I'm not sure if it'll be cut.

Without cutting the traces, they may ripp off the board on the 'inner'-side of the breaking. But I think the good news of the thickness of the board is that keeping two break off spots at the USB end should be strong enough to hold on to the PICO - for placement and soldering. Milling first the two long edges incl the USB sides, and then the USB tip, then drilling the holes in the breakout parts at the left and right side of USB tip... No problem w/ Jaltek the specialists, but listen to ideas never hurts... it's difficult though ;-)

I don't see any room for a polyfuse either; you've crammed every square millimeter of the board. Though it looks like you only need room for a 603 package...?

You might be able to move one of the LEDs to the side of the chip (ie put the LED right where it comes out of the chip, before the resistor)... not sure if that's useful though.

Re: silkscreen on bottom - the markings near that row of 6 pads for swd (what do they go to?) make it look like the 6 pads are just 2 pads each for vcc, gnd, 3v3, which ofc they're not.

I mentioned this during the kickstarter; why not put two little plated through holes in the spot for the 32k crystal, to make it easier to install one of those cheap + plentiful radial can type crystals? It looks like you've got the space on the layout for it, since there's just ground plane underneath there...

I think I've managed to make room for an 0603 polyfuse now - it's tight though! \o/

The markings for power are a bit difficult... I've moved them around to make them nearer the actual pins, but there's not enough space to move the SWD pads down, so the labels do have to go below them :(

With the crystal, the problem I have is that the STM32F4 is on a low power process, and the oscillator can't drive a 'normal' watch crystal reliably (at least not well enough to run it at the 32kHz it's trimmed it). It's got to have nearer a 4pF load capacitance, and I don't know if they even make the through-hole watch crystals like that.

I actually ordered 6000 crystals, found out they wouldn't work, and had to try and send them back. As expected crystals that actually work are twice the price, making them by far the most expensive part on the board (apart from the STM32).

It's not hard to buy one though, and they're not very painful in small quantities - ABS06-107. The 107 is very important (as I found out!).

They shouldn't be that much more difficult to solder than the ones with wires.

Yugh!

Do they just use shiny brass covers, or are those crystals solid gold? More than an order of magnitude more expensive than those ubiquitous watch crystals.

On that note - if I don't have a crystal there, I can use C14/C15 for something else, right (if I'm willing to solder onto the pads for the crystal)? Same goes for the A13-15 pins (which look to go to the debug connector)?

yes, if you want to. However the SWD pins may require some poking as by default they'll be in SWD mode.

There's GND + VCC on the SWD pads too - potentially there could be some tiny little shim that solders on with something on it :)

At some point, I'd love to know what I need to poke to gun SWD - but that's not an urgent priority. I'll make a github issue in the docs for it so it doesn't get forgotten.

Are there SMT tactile switches that use even less space than the current one? If so, that could provide some room...

It's quite hard to find decent SMT switches, so I think the current one is the best bet for now I'm afraid.

For SWD, it's the same pinout as the ST discovery boards. While this page is for the original Espruino board, it still basically shows you what you need (you only need 3 wires, although to get USB working you need to power the board separately as well, boot it, then plug in USB).

About the Silkscreen - in another thread the difference between the board docs and the silkscreen for Bat/VBat/Bat_in/etc came up.

Any thoughts on whether the silkscreen should change? VCC is possibly misleading for something that is just 'plus volts' - ~4.5v on USB, or whatever 'Bat' is.

So VCC is the way you can access the supply to the device? from either source.

Please see attached image there us a resistor/competent very close to top pin edge.
When soldering a header on it was very easy to short the pin to this component.

1 Attachment

So VCC is the way you can access the supply to the device? from either source.

Absolutely, yes.

... very close to top pin edge.

I noticed that too - it's already been moved up as much as I think I can get away with now :) Thanks for the heads up!

We need to put some thought into what the correct names for the pins are.

We have three:
Battery input (BAT_IN - maybe should be Vin? )
Battery/USB voltage (VBat? I like this, because it's the same as the original Espruino board - if it was called Vcc, I would expect it to be a regulated voltage, not the input voltage)
Regulated 3.3v (Now is this Vcc? 3V3? I'd personally assume that Vcc meant the processor voltage...)

Vin is a good idea... What about Vout for the other one? VBat is a bit confusing when you're running off USB - it might also imply that it could be an input.

My thought was that VCC was TTL+, and VDD was CMOS+ - so probably VDD is the right one for 3.3v? Even so it's totally confusing, and probably 3.3v is the most sensible.

Vin is okay, maybe B+ and B- for the battery inputs is an alternative.
Vout is okay.
VDD is better than VCC, but 3.3V (or clearer 3V3) is better.