[Cryptech Tech] PCB prototyping machine [Was: BCP prototyping machine]

Benedikt Stockebrand bs at stepladder-it.com
Tue Feb 24 19:08:23 UTC 2015


Hi Pavel,

Павел Шатов <meisterpaul1 at yandex.ru> writes:

> Benedikt, you do know what all those capacitors are there for, don't you?

let's say I'm semi-clued enough when it comes to electronics to make an
educated guess:-)

But seriously, when it came to the breakout board you mentioned before
(http://www.proto-advantage.com/store/images/PRODUCTS/BGA0006_0.JPG) I
silently assumed that the pads to the left and right are there exactly
to place such caps in a reasonably convenient way.

And yes, like Fredrik---who also showed some serious interest in that
PCB printer thing---I've been working mostly on fairly slow speed stuff,
i.e. the avalanche noise generator; so my previous post may have been a
bit biased by that, but I guess you are too focused on the FPGA and ARM
side yourself.

> Bredboarding an FPGA is a very flawed idea.

Sure not, and perfboard isn't all that much better, either.  But: If we
can do a modular design where individual modules can be replaced, and if
*some* of these modules can actually be done with anything more
convenient than ordering PCB prototypes, then I expect that would be
rather helpful; maybe not to your part of the project, but at least to
what Fredrik and myself are working on.

I have tried not to get myself drawn into the FPGA part of the project
simply because I currently find it hard to find time for the stuff I've
already familiarized myself with, so I don't have an idea if we can
separate the FPGA (with its peripherals) from the MCU (with its
peripherals, like the RAM you mention), but I do know that Fredrik and I
can split the avalanche noise generator into at least two boards which
I've actually built using breadboard and then perfboard and then a
custom prototype PCB: The voltage supply (either using boost converter
or a charge pump) and the analog part of the actual noise generator,
plus the MCUs we've been using.

> FPGA layout is actually not our worst problem. The worst part is that
> i.MX requires DDR3-1066 memory. GHz signals require controlled PCB
> impedance, which is impossible to get outside of a factory. How are
> you going to deal with that?

Put all of that on a single module with only low speed connectors going
off that module?

>> It'll look messy like hell, and it won't be much fun to carry around,
>> but it'll make experimenting much easier.
>>
>> When I started messing around with the microcontrollers in general that
>> approach proved to be extremely useful, despite perfboard being so
>> unbelievably tedious to use.
>
> I agree, this approach works nicely with DIP and SOIC packages, but
> BGAs are completely different story.

It's not so much with the package type but the signalling speeds, and
wherever we have only low speed links we can break the design up into
modules that can be independently worked on.  And as I said, at least we
can break some of the stuff off this way.


Cheers,

    Benedikt

PS: While this isn't as important to the Cryptech project as it is to my
    own little project, going modular even beyond the prototyping stage
    also has the advantage that it makes it easier to diversify the
    design, which will also make things more expensive to untargeted
    manipulations.  But then, that's an entirely different topic.

-- 
Benedikt Stockebrand,                   Stepladder IT Training+Consulting
Dipl.-Inform.                           http://www.stepladder-it.com/

          Business Grade IPv6 --- Consulting, Training, Projects

BIVBlog---Benedikt's IT Video Blog: http://www.stepladder-it.com/bivblog/


More information about the Tech mailing list