Stories
Slash Boxes
Comments

SoylentNews is people

posted by janrinok on Wednesday April 09 2014, @11:39PM   Printer-friendly
from the what-will-they-think-of-next? dept.

Ras Pi foundation announces a new product: the compute module.

The compute module contains the guts of a Raspberry Pi (the BCM2835 processor and 512Mbyte of RAM) as well as a 4Gbyte eMMC Flash device (which is the equivalent of the SD card in the Pi). This is all integrated on to a small 67.6x30mm board which fits into a standard DDR2 SODIMM connector (the same type of connector as used for laptop memory). The Flash memory is connected directly to the processor on the board, but the remaining processor interfaces are available to the user via the connector pins.

While not yet what I imagined by only the name i.e a unit to build a shoebox-sized Beowolf cluster of 1K-RasPi-cores the new form factor and pin-out should make this endeavour easier (ahem... for someone skilled in PCB design, EE practician, and a soldering-fu master that has achieved enlightenment... not quite my profile).

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 1) by NoMaster on Friday April 11 2014, @04:29AM

    by NoMaster (3543) on Friday April 11 2014, @04:29AM (#29856)

    Late back, but:

    It's perfectly possible with a cheap setup; you don't need $1000's worth of gear. I normally use an Ersa analogue station (don't like the balance of the Metcals) & an old modified Olympus stereo dissection microscope, but just for giggles I did one of the boards with my old gear - a cheap Atten station with T18-05C & 08D tips, working under a maggylamp from the hardware store. Total cost AU$100.

    Working distance was as crap as you'd imagine (I ended up coming in from the side & using the corner of the 0.8mm D tip, as I couldn't get the face of the C tip square onto the joint), and you need plenty of flux, a steady hand, a good eye, and the experience to know what the solder is doing while the joint is hidden by the tip - but, soldering the 0.5mm & 0.65mm pitch TSSOP & VSSOPs, the results are as good as any hand-soldered SOT-223 or SOIC.

    And that was hand-soldering every pin individually - I'm old-school enough that I can't bring myself to do the typical "flood 'em with solder & wick up the excess" method everyone seems to favour, and it doesn't work well on home-made boards without soldermask anyway...

    --
    Live free or fuck off and take your naïve Libertarian fantasies with you...
  • (Score: 1) by tftp on Friday April 11 2014, @05:36AM

    by tftp (806) on Friday April 11 2014, @05:36AM (#29879) Homepage

    Yes, that was a difficult job. My microscope is currently one of those Omano things [microscope.com]. It works fine, and the lens is protected from smoke with a clear filter. I never work under a magnifying lamp - just not my thing. I don't even have such a lamp.

    One of big concerns with "flood them and wick them" is overheating of the part. I try to avoid such a method.

    With regard to homemade boards without a soldermask... is it worth it if you can order three boards made, with SM and silkscreen, for $33 each [4pcb.com]? Or such a service is not available in .au? A 2-layer board would be pretty horrible in terms of EMC, but a hobbyist isn't likely to try for CISPR 22 Class B. Given the minimum trace/space that they accept, one can build a copy of IBM/360 on one board without making even a single jumper (as long as you put power rails on top, as they used to do :-)

    • (Score: 1) by NoMaster on Friday April 11 2014, @09:30AM

      by NoMaster (3543) on Friday April 11 2014, @09:30AM (#29933)

      My microscope is currently one of those Omano things.

      Very nice at that price! I'm off to investigate it further. The (lack of) DoF on my old Olympus is an issue...

      One of big concerns with "flood them and wick them" is overheating of the part. I try to avoid such a method.

      Agreed, and that's another reason I don't do it (the first two being that basically it makes me cringe, and by the time you wick up the excess it doesn't seem any faster than pin-by-pin...)

      With regard to homemade boards without a soldermask... is it worth it if you can order three boards made, with SM and silkscreen, for $33 each?

      True - but if you send them out you can't have prototypes in your hand by lunchtime ;). The ones I did yesterday are built and functionally tested already; I'm about to rejig the design a bit and will send them out to be made while I finish the software.

      Since getting out of the trade I mostly do small low-speed data & sub-VHF radio stuff (e.g. those boards are I2C + interface logic, 4MHz signal amp/buffer, programmable silicon oscillator, & associated passives), so can usually get away with predominantly single-sided (the other side is left mostly as groundplane, except for the odd jumper track & where it has to be pulled back e.g. for amp stability).

      A 2-layer board would be pretty horrible in terms of EMC, but a hobbyist isn't likely to try for CISPR 22 Class B.

      As you say, EMC compliance rarely raises its head for home-built stuff (although I sometimes think it should, particulaly with the rise of switchmode supplies designed by hobbyists!), and all my commercial experience is with telco stuff (serial 56 under the Aus regs). I also sometimes think things like the RPi are a bit of a sneaky end-run around the regs - the RPi itself has passed but, given that they're starting to appear as a component of at least semi-commercial gear, I wonder how much of that is being tested (& would pass)?

      Is it still CISPR 22 Part B in most of the world? Aus seems to have already started transitioning to CISPR 32; I'll have to look up the differences sometime...

      --
      Live free or fuck off and take your naïve Libertarian fantasies with you...