[Arm-netbook] IC for analog and digital buttons (EOMA-68)

Mon Aug 11 10:00:10 BST 2014

I see your point. At this precise moment, your standard is up-to-date. I
cannot complain. But just imagine that the EOMA things had gone faster and
1 year ago it was in production - before list discussion took place and
decision to add USB3.0 was made (there is a planned update to USB3.1 with
something like 10Gbit). Could we say that EOMA will be limiting for its
users (it it missed the USB3.0)? Could you claim Gbit ethernet if EOMA had
only USB2.0? Will anyone that plans on NAS-like server usage jump on EOMA
product if he knows that there is no native SATA but just USB to SATA + USB
to Gbit on one 480mbit/s port? Not to mention possibility to end up with
12Mbit/s port on some cards...
Even now it is limiting because the same A10, A20, i.MX provide SATA, PCIe
that base board designer cannot use. Not to mention MIPI and other high
speed board interfaces. Is there a good way to add a camera (or even 2) to
an EOMA based board? I think USB will come to the resque? But even now it
is over-saturated for some of the boards...
I checked EOMA page at elinux - there is a planned update of the standard
in "2-3 years". I would say, backward compatibility will be lost? New base
boards for new CPU cards? Where is the 10 years lifespan? Who will care
producing old standard cpu cards with new SoCs?
You say Gbit Ethernet - but IC1T does not have this? So I, in the position
of designer of base board, cannot rely on this because some cards with give
limited speed?
In fact you are using one single multifunctional interface as backup - I
mean the USB. If SoC does not support something, add it on USB. From the
early days of EOMA you are ruling out the OMAPs because of HS-only USB. But
the same workaround as for low end cards could have save the OMAP CPU card
- you just need a HUB (on the CPU card). Probably this was the point behind
developing SoC with HS only USB? Even with other "supported" SoC designers
are ending with USB HUBs on the base board... So there is an alternative
solutioin - base boards that need to connect internal low/full speed USB
devices need to add a hub. If my product does not export user USB out of
the "base board", and has nothing but HS capable on board peripherals, I
can safely use any card, even those with OMAPs.
Miguel speaks about <16ms reaction time (this is another topic), but
obviously he demands some performance. So, for this reason or another, his
product will never work with low end card (beside presenting "unsupported
cpu card" message). To be able to sell products he has to satisfy user
demands. Limited audio bit depth and network speed are bad for him - one
youtube review about slow game console and mass user will be lost. Yes,
geeks will know that with good cpu card everything is fine, so it comes to
the same point where we started - "someone" will define list of supported
cards, even if it is not Miguel because he is not allowed to do it by
standard.

I have another question, sorry if it is out of context here. I know PCMCIA
was designed long ago - is it certified that it will run fine with 5Gbits
signals USB3.0? Does it comply to impedance matching for all interfaces?
How about EMI? I am asking because one reason (for us) to switch to another
SoM format was newer, high speed interfaces. Our suppliers switched to
connectors that are compatible with newer standards. Maybe they could route
out HS USB or SATA on ISA connector... I even doubt if PCMCIA can cope to
USB2.0 HS routing requirements. I know it works in practice, even with
flying wires. But not for production HW.

2014-08-11 9:48 GMT+03:00 Luke Kenneth Casson Leighton <lkcl at lkcl.net>:

> On Mon, Aug 11, 2014 at 2:17 AM, krasi gichev <krasimirr at gmail.com> wrote:
> >> out of curiosity, i'd be interested to hear if the design that you
> >> are creating is intended for anything approaching a decade-long
> >> support and end-user lifespan.
> >
> > First of all, I want to make it clear that my area is closer to
> industrial
> > and sometimes automotive than consumer. Also, I am sure that I don't have
> > your experience and I don't see all the details that you have seen - and
> > large customers that expect the CPU cards.
> >
> > And, the usual disclaimer, this post will not bring any creative ideas,
> it
> > will explain my position, so feel free to stop reading here...
> >
> > But what I have learned in recent years is that there is no such thing
> like
> > "one size fits all". Even if something fits well at the current time,
> this
> > is temporal and will change in several months or an year.
> > In our devices, in past 10 years, we had passed over several form
> factors of
> > "processing boards" - PC104, ISA, ETX, and some other, our own too. And
> what
> > we got like benefits is limited to:
> > - possible second source - I hope the EOMA will be able to attract more
> that
> > one designer and producer of CPU cards
> > - possible upgrade of the CPU card - this works sometimes, but never for
> 10
> > years
>
>  ok, this intrigues me, that it is not clear why you believe that the
> interfaces selected would not last another 10 years.  do you have time
> to go through them?
>
>  * ethernet.  GbE. has been around for 2+ decades.  do you expect GbE
> to be around for another 10 years?
>
>  * USB3.  USB2 has been around for 2 + decades.  it has been upgraded
> to USB3 which is 5gb/sec at the moment..  do you expect USB3 to be
> around for another 10 years?  in fact, i understand that USB3 is to
> get at least *another* speed upgrade.
>
>  * GPIO, UART, I2C, SPI  i think it safe to say that these will be
> around for at least a decade?
>
>  * SD/MMC.  has been around for around 2 decades, and is constantly
> being upgraded.  i think it safe to say that this will be around for
> another decade?
>
>  * RGB/TTL.  has been around for 2+ decades.  it's the baseline for
> video output.  there's always going to be converter ICs, no matter
> what the latest-and-greatest standard.
>
>  i think it reason-ably safe to say that every single interface picked
> on the EOMA68 standard has a lifetime of at least one decade.  as in
> there are reasons why every single interface has already been
> long-term and will continue to be so.
>
>
> > - maybe 2-3-5 years is realistic ( I mean in industrial, in consumer 1
> > year top); with current CAD tools it is not so hard to redesign to other
> > form factor (you know it, you have made many desing recently)
>
>  yeees, but now you have set a MOQ for pricing for your clients.  the
> purpose of the EOMA68 standard is to bring the benefit of mass-volume
> pricing *even* to this smaller custom run market.
>
> > Nope, I cannot even imagine that something (commercial) will live for so
> > long. It might run fine but customers are always demanding new features,
>
>  how many of those features do not fit into USB3, SPI, SD/MMC and I2C?
>
> > better perfomance (even just better design).
>
>  what do you mean by performance?  do you mean anything other than
> better CPU, better RAM, faster RAM, more RAM or more storage?  because
> those are exactly what is on the CPU Card.
>
> > My point is (and probably I am biased by my lifestyle) that I will
> prefer to
> > put the old one in the basement, or on the e-bay, and just go for the
> > newest. Or, if you prefer, recycle it.
>
>  exactly!  so just buy a new base-unit, keep the CPU Card, you have
> just saved 30% on the cost of a monolithic unit.
>
>  ok, there's a lot here, i have to get on, more later ok?
>
> l.
>
> _______________________________________________
> arm-netbook mailing list arm-netbook at lists.phcomp.co.uk
> http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook
> Send large attachments to arm-netbook at files.phcomp.co.uk
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.phcomp.co.uk/pipermail/arm-netbook/attachments/20140811/0107daca/attachment-0001.html>