Re: [ox] FPGAs (was: Re: Die Anwendbarkeit ...)
- From: graham belegost.mit.edu
- Date: Sun, 17 Dec 2000 13:10:39 -0500 (EST)
Hi Graham,
some more questions about FPGAs:
graham belegost.mit.edu schrieb:
such as you describe for some years. And in fact there are groups
working on designs for FPGAs (eg OpenCores). BUT commercial users
of such chips are very worried that their designs may be too easily
copied.
This means that FPGAs once programed can be read and reasembled?
Or can be copied from those chips to another?
There are 2 potential 'problems' for commercial users:
i. The bits used to program an FPGA can simply be copied and used
to program another FPGA (say by a competitor making a cheaper copy
of a board including the FPGA). This is very difficult to prevent,
since the bits are usually stored in a ROM next to the FPGA (one
way to prevent it, which I believe is used in flight control systems
for missiles where the army does not want their opponent to be able
to copy the design, is to use only static RAM with a power source
which can be switched off when required so that the design 'vanishes').
But this kind of copying is illegal and not most FPGA users main worry.
ii. If the internal structure of an FPGA is known, it is in principle
possible to 'disassemble' the programming bits to recreate the original
high-level design. If some small changes are made, and then the design
is reassembled, it may no longer be straightforward legally to say whether
the new design is the same as the original.
One techniques being proposed to stop this is a kind of 'watermarking'
of designs, where redundancy is introduced to be able to identify even
a changed design as being descended from the original.
The latest FPGAs also include encryption, so that the bitstream
cannot be disassembled without first breaking the code.. Logically,
this should mean that it is no longer necessary to keep the bitstream
format a secret, but the manufacturers will not listen to this
argument.
As a result, they put pressure on the FPGA manufacturers
to make it difficult to get at the internals of the chips. The
manufacturers are happy to co-operate for their own reasons.
Generally FPGA manufactures are interessed in increasing their
production - whether for "free", "illegal" or "normal" products.
Right?
The argument that hobbyist use will increase sales does not impress
them. They are interested in volume sales, whether commercial or
military (the military market is a big part of fpga sales).
However, some people have suggested in the past that it may be possible
to use this approach to persuade some of the smallest FPGA manufacturers
to co-operate, since they may be desperate to find anything which
gives them a competitive edge against the largest. Personally I'm
sceptical, but who knows?
The result
is that the internal data format for FPGAs is a secret. No-one can write
free software to program the chips (the manufacturers do release
some 'free beer' programs
What means 'free beer' program? A proprietary program which some
limited features (limited time)?
I mean zero financial cost but no source code. The limit is not usually
time but size of device handled, and sometimes size of design allowed.
These limits are not that severe: it is perfectly possible to use this
software to do real designs.
to do this). There was one FPGA design
which was inherently open - the programming data from the chip could
be read and changed at any time from a normal processor bus. The
company which made it was bought by Xilinx (the Microsoft of the
FPGA world) and production closed after a couple of years. There
are no other FPGAs with this property, and Xilinx are very aggressive
about their patents (they just won a patent case against their
nearest competitor, Altera).
Why is Xilinx so aggressive to keep their design closed - because of
the pressure you mentioned above? And what about the competitor?
Aren't they interessed in generating a market and therefore in
making their design free?
See above. Also, the competition between these companies is quite fierce.
They all reverse engineer one another's products with electron microscopes
etc, so in practice the internal structure of all the FPGAs is known
to each FPGA manufacturer. But I suspect the need to do this reverse
engineering gives any innovating company a few months lead over its
competitors. But of course I have no idea what really goes on at
decision making levels in these companies.
This makes it difficult to think about
producing a 'free' FPGA (quite apart from the fact that no-one
has yet produced a free IC of any type, though some groups are
trying. Free electronic design groups are split over whether it
is possible to produce free designs using non-free software (I
guess the majority opinion is 'yes').
Does any useful non-free software for programming the FPGAs exist?
Or is it only this type of 'free beer' program?
If you accept the use of non-free software, it is perfectly possible
to do FPGA design work - maybe not on the very latest, largest FPGAs,
but those are horribly expensive in any case.
Maybe silly questions, but I am not very familiar with such stuff.
Not silly at all but hard to answer.
I tried to summarize a newsgroup discussion of these topics once: see
http://www.opencollector.org/news/Bitstream/
don't expect any clear conclusions though!
Are you on CCC 27th to 29th of dec in Berlin?
I was hoping to talk as part of the F-CPU group. But I'm going
to Portugal for Christmas (my wife's mother is ill) and it seems
very unlikely now I'll be able to come for cost and time reasons.
If you are going, you should try to talk to Whygee or others from
f-cpu: some of them have a lot more fpga experience than me.
Thanx,
Stefan
Best wishes
Graham
PS I will be away from the 19th to Jan 3rd and may not have access
to a computer over that period
_________________________________
Web-Site: http://www.oekonux.de/
Organisation: projekt oekonux.de