THE RATIONALE BEHIND COMPUTER-
Jean-Philippe
Boufflers
THE RATIONALE BEHIND COMPUTER-ON-MODULES
In a world where the race to integration, i.e
the most important functions in the smaller available space, and
competitiveness are fundamental, computer-on-modules (hereafter named COM) are
a cost-effective solution to implement. After reviewing the concept and
exposing the challenges, we explain why COM is perfectly suited to the
different embedded computer markets where size constraints and modularity is
important.
I/ Why COM?
For those who are not familiar with
what are COMs (also known as SOM for system-on-module), let us compare them to traditional boards
where you have not only the processing units but also all the interfaces,
connectors and peripherals gathered on a single board. Frequently, HW designers
would have to redesign part of the board when time comes to next PC generation. The main technical concept behind
COMs is
rather than traditional single board computer where the design is fixed
and non-flexible, COMs offers modular solution to separate main CPU module and
carrier board to make a flexible and easy-to-upgrade design. Imagine
if you would have to
upgrade your system to another CPU platform while the current
application is using single board computers? Answer is obvious, you would have to
re-design a complete new board
which is a time consuming and not cost-effective task. Alternatively, if
you were already using COMPUTER-ON-MODULE design, you could simply upgrade your
system by replacing the corresponding COMs while keeping the existing carrier
board. This is beneficial in terms of time saving, cost-effective and hassle
free. COM modules
were created to solve in an efficient manner the question of
electronics system evolution.
I-1/
Different modules form factors
The most used form-factors are identified in
fig 2 below. The reader must understand here that larger form-factors allow
users to take advantage of bigger CPUs footprints and larger power
consumption.
FORM FACTOR
|
SIZE
(mm)
|
COM EXPRESS
|
COMPACT : 95x95
BASIC : 95x125
EXTENDED : 110x155
ULTRA : 84x55
|
ETX
|
114x95
|
XTX
|
114x95
|
Qseven
|
70x70
|
COM form factors
I-2/Different
pin-out definitions
Basically, there are 7 types of pin-out
definition for modules. For
example, the type 2
connector layout defines 440 interconnect pins between the COM Express module
and the carrier board. Type 1 defines 220 pins.
Types
|
Connector
Rows
|
PCI
Express Lanes
|
PEG
SDVO
|
PCI
|
IDE
|
SATA
|
LAN
|
USB
2.0 USB 3.0
|
Display
Interfaces
|
Type 1
|
A-B
|
Up to
6
|
--
|
--
|
--
|
4
|
1
|
8 / 0
|
VGA,
LVDS
|
Type 2
|
A-B
C-D
|
Up to
22
|
1 / 2
|
32 bit
|
1
|
4
|
1
|
8 / 0
|
VGA,
LVDS,
PEG/SDVO
|
Type 6
|
A-B
C-D
|
Up to
24
|
1 / --
|
--
|
--
|
4
|
1
|
8 / 4
|
VGA,
LVDS,
PEG,
3xDDI
|
Type
10
|
A-B
|
Up to
4
|
-- / 1
|
--
|
--
|
2
|
1
|
8 / 0
|
LVDS,
1 x DDI
|
The different interfaces available
relatively to type. Connectors are different between COM express, ETX/XTX and
Q7 for-factors.
II / Why
using Computer-on-module?
“ Most of the time, our customers
develop a wide range of product, going from entry-level to very high-end solutions not only
aimed at different end-customers but sometimes also different countries, with various
economics conditions. Also those applications are frequently peripheral-rich
and typically size restrained” says Jean-Philippe Boufflers, sales manager at
AAEON.”It
becomes practical to let our customers focus on their added value, which is
clearly not designing processing units, all vendor do that very well, but
rather in developing their own proprietary SW applications or specific
programmable CPLD or FPGA circuit on the carrier board. It then becomes easy
for our customers to define which CPU=f(performance/consumption/price) is the
most suited one for every of their market segment.”
However, are COM the best friend of HW/ SW
engineers and product managers?
Typically,
we face many approaches. Let us start by the easier one. Some customers may
have competences in electro-mechanical or SW but few in electronics. In this
case, the best way is to go through some of the vendor services that consist of
buying the complete Carrier+COM (+ maybe heatsink) through him. Customer and
vendor mutually define the specification and agree on the interface and
connectors that are required. Finally customer focuses on its core competency
that can be SW, mechanical, chemical…
Second case
is when customer already has a carrier board and would like to switch from a
fully integrated system to a module-based one. Here, the vendor can help ease
the task of re-designing the carrier board by providing a design guideline
document that will explain how to route the different signals to the specific
connector that will have to be put on the carrier. Here also, the customer
could prefer to send its carrier board to the vendor R&D to get assistance
on this task.
“It happened in
the past that our technical resources had to examine the customer’s carrier
board to check for adaptation. AAEON design their module with optional resistor
to adapt the Pull up and Pull down values for signal compatibility with the
customer's carrier. Then, the BOM (Bill Of Material) is fixed and customer
can take advantage of going forward with computer-on-modules” estimates
Pierre-Francois CASTA sales manager of MSI department at EQUIPEMENTS
SCIENTIFIQUES FRANCE.
It also
comes to the case where some customers, for any reasons, wish to switch from COM vendor A
to COM vendor B. In an ideal world, it should be very easy to do that but in a real world, electronics sciences
like to complicate life of
product managers and there is likely chances that when you turn everything ON,
the light will still be OFF… So what is happening?
Although
the pin definition is standardized by some leading PICMG consortium, the
interoperability between COM vendor A and COM vendor B is not a so obvious
topic.
It is now
and then that the notion of service takes its full sense. Many customers
perceived the benefits of using COMs but also feel lost in the middle of ocean
if left alone. “It is our daily job to
propose our assistance to our customers. Ideally, we can examine their carrier
board, define BIOS settings in order to get a first boot up of the system. Of
course, this task goes faster if end customer accepts to send a carrier board to the vendor site to
allow in-depth examination. In some particular cases, highly sensitive domains such as military or medical applications,
it is not possible to get a carrier board so we use our local engineer resources to help customers or even provide dedicate engineer for on-site support, say Kevin
Chiu, Head of COM and SW department of AAEON HQ.
It is clear
that a careful and precise list of the requirements for the system has to be
set. Latest COM CPU platforms, despite of their promising characteristics,
could have some limitations due to drivers not yet released for some particular
OS or simply cannot support legacy ISA interface. And what if your new system
generation needs USB 3.0?
It might
happen too that customers wish to migrate from MiPS architecture to x86 ones. As long as the
pin-out definitions of the computer-on-modules raw(s) are standardized, there
should not be any problem to switch from one architecture to another. Only will
be needed some adjustment in the Bootloader, if running under Linux.
One should
also pay attention to the SW application. It can be common to customers to take
advantage of the latest CPU platforms offering multi-cores/multi-threads
running under 64 bits OS but it may happen that they are deceived by final
performance when they run their application. In this case, care must be taken
and SW be reviewed to check for the structure/commands used in their code (and
also sometimes, BIOS settings to be checked too).
When SW
developers want to access some particular HW resources like buses or manage
brightness, fan speed etc.., another useful service can consist of delivering
part of the job. “To ease the task of SW
developers, one can turn toward a set of predefined APIs delivered as a service
that help shorten time to market and avoid having to spend considerable time
and effort digging into HW specification” states Kevin Chiu.
III / The Art
of Combinationhereas we used to say ‘one size fits all’, in COM
world we would say ‘multiple sizes fit multiple applications’. Taking into consideration the system
size constraints will let you decide for
the COM form factor. Considering thermal constraints will let you select the
CPU platform and possibly a solution with active fan or passive thermal
solution. Performance
requirements will also guide your choice for example choosing Intel latest 3rd
generation Core
i®called
Ivy Bridge®. For sure, one must not forget the needed I/O’s that will guide
you to choose the right
carrier board (with appropriate type regarding type for module).
IV/Benefits
for customers
“Compared to the traditional re-design of
complete board, using modules allow companies to deploy more quickly a HW
upgrade (ROI, time to market etc.
As already
stated, using COM allow one company to focus on their core competencies and
added-value rather than on re-inventing the wheels ever and ever.
The
scalability of COM is also a real interest as it can fit from the most portable
device to much bigger systems where heat can be more easily dissipated.
V/Conclusion
It’s been a
while that the notion of plug’n’play is surrounding us every day. Even if you
cannot think of simply removing old module, inserting the new one and claims
the job is done, it is clear that COM is an efficient and cost-effective way of thinking about your product
evolution. Modular design is not synonym with short-term lifecycle and of
course in IPC world, guaranteeing access to modules even after 7 years is
really an important matter too. Scalability along with time-to-market
consideration makes COM the perfect companion for your embedded devices as long
as it comes with a full set of services (from API or BIOS services to full
on-site consulting) to ease the product development especially when HW team is
willing to move straight from a traditional single board computer to a
modularized approach. Apart from the highly flexible aspect of COM to reflect
the ever-changing needs of today’s electronics systems, service is definitely
the big differentiator to turn your project using COM into a success. Vendors
provide 80% of the design while customer focus on the 20% left and so finally
it becomes a win-win solution.
For more
information,
Tunnel
Creek (TC), Luna Pier (LN), Cedar View(CV), Sandy Bridge(SB) and Ivy Bridge(IB)
are proprietary names of Intel.
