Reply by Paulson, David.C \MA\●January 20, 20062006-01-20
Small sample quantities are available today. The evaluation board is
8-10 weeks leadtime.
________________________________
From: lpc2000@lpc2... [mailto:lpc2000@lpc2...] On Behalf
Of Joel Winarske
Sent: Wednesday, January 11, 2006 7:15 PM
To: lpc2000@lpc2...
Subject: RE: [lpc2000] LPC's w/ ENC28J60 for ethernet via SPI
> > The AT91SAM7X may not be available yet. I
don't see stock at digikey, > and
> > on the Atmel site there is no "buy from distribution" link
like there is > > for other full production chips like the
regular SAM7. Quite a few AT91 > > chips look interesting, if you can get your
hands on them ....
>
> I received two samples from All American over two months ago. No posted > errata so either silicon revision, or queued for
fab.
> > The AT91SAM7X may not be available yet. I don't see stock at
digikey, > and
> > on the Atmel site there is no "buy from distribution" link
like there is
> > for other full production chips like the regular SAM7. Quite a few
AT91
> > chips look interesting, if you can get your hands on them ....
>
> I received two samples from All American over two months ago. No posted
> errata so either silicon revision, or queued for fab.
I received word today the AT91SAM7X was recently approved for production.
They are waiting on production dates from the planning department.
Joel
Reply by Joel Winarske●January 10, 20062006-01-10
> > The Infineon 10/100 ADM7001 PHY is ~$0.86 in qty of 2500... > >
> >
> > Joel
> >
>
> I googled around for this part. I found the data sheets but not any
> prices or where I could get it.
>
> Is it available in the usa ?
Check with your favorite Infineon distributor.
AVNET has them listed online (www.partbuilder.com), min order 2500 4 week
lead. Not sure on their pricing.
Joel
Reply by donhamilton2002●January 10, 20062006-01-10
--- In lpc2000@lpc2..., "Joel Winarske" <joelw@i...>
wrote: >
> The Infineon 10/100 ADM7001 PHY is ~$0.86 in qty of 2500...
> Joel
>
I googled around for this part. I found the data sheets but not any
prices or where I could get it.
Is it available in the usa ?
thanks
hamilton
Reply by Joel Winarske●January 9, 20062006-01-09
Hi Mark,
> I have always understood that general broadcast
frames cause most
> overloading problems and it also certainly helps to not operate in
> promiscuous mode on a network with high foreign activity levels.
> Using switches to shield weaker devices can also help.
> However at the end of the day there are real limits. If frames have
> to be handled, this requires time and if the frame rate is faster
> than the throughput then something will inevitably get lost. If a
Yes I agree on all points. I think there should be a series of performance
indices for a given Ethernet solution.
Regards,
Joel
Reply by Joel Winarske●January 9, 20062006-01-09
> > Have you looked at the Atmel AT91SAM7X? >
> I took a brief look, but availability concerns me. Actually, while
> getting into an alternate CPU discussion, I came up with an
> interesting silicon cost comparison for some of my options. These
> prices below are based on single unit quantities priced via Digikey:
>
> LPC2292 $13.46
> CS8900A $8.36
> 256Kx16 SRAM $4.41 Total: $26.23
>
> EP9302 $16.14
> 64Mb (4Mx16) SDRAM $5.67
> Micrel 10/100 PHY $3.78
> 4Mb (512KB) SPI flash $2.15 Total: $27.74
>
> Hmm, $26.23 for an ARM7TDMI @ 60MHz, 256KB flash, 512KB SRAM, 10mbps
> ethernet, vs. $27.74 for an ARM920T @ 200MHz, 512KB flash, 8MB SDRAM,
> 10/100 ethernet. A whole $1.51 additional silicon cost.
In general:
Pricing solely based on single quantity is not realistic. If you buy in low
quantities, then you get end of the food chain prices. Some parts get
exponentially cheaper the higher the quantity, others reduce in a more
linear fashion. If possible the goal is to stay away from the linear
pricing model.
Yes the Cirrus is a good value if that's what you need, and you're
tooled to
handle high frequency designs.
The Infineon 10/100 ADM7001 PHY is ~$0.86 in qty of 2500...
Joel
Reply by Tom Walsh●January 8, 20062006-01-08
Tom Walsh wrote:
>Ryan Niemi wrote:
>
>>>Have you looked at the Atmel AT91SAM7X?
>>>
>>>
>>>
>>>
>>I took a brief look, but availability concerns me. Actually, while
>>getting into an alternate CPU discussion, I came up with an
>>interesting silicon cost comparison for some of my options. These
>>prices below are based on single unit quantities priced via Digikey:
>>
>>LPC2292 $13.46
>>CS8900A $8.36
>>256Kx16 SRAM $4.41 Total: $26.23
>>
>>EP9302 $16.14
>>64Mb (4Mx16) SDRAM $5.67
>>Micrel 10/100 PHY $3.78
>>4Mb (512KB) SPI flash $2.15 Total: $27.74
>>
>>Hmm, $26.23 for an ARM7TDMI @ 60MHz, 256KB flash, 512KB SRAM, 10mbps
>>ethernet, vs. $27.74 for an ARM920T @ 200MHz, 512KB flash, 8MB SDRAM,
>>10/100 ethernet. A whole $1.51 additional silicon cost.
>>
>>
>>
>>
>
>Then take the EP9302 and add:
>
>Previous cost: $27.74
>SD Socket: $3.80
>SD 128Meg card: $8.00
>64Mb (4Mx16) SDRAM -$5.67
>32MByte (16Mx16) SDRAM $12.54 Total: $46.41
>
>And you get a killer Linux powered embedded system!
>
>BTW, where did you find the cost of the EP9302? I cannot seem to find
>anyone but Cirrus that you can get those parts from (e.g. Nu-Horizons,
>etc.).
nm, I found them on Digikey.
>>Have you looked at the Atmel AT91SAM7X?
>>
>>
>
>I took a brief look, but availability concerns me. Actually, while
>getting into an alternate CPU discussion, I came up with an
>interesting silicon cost comparison for some of my options. These
>prices below are based on single unit quantities priced via Digikey:
>
>LPC2292 $13.46
>CS8900A $8.36
>256Kx16 SRAM $4.41 Total: $26.23
>
>EP9302 $16.14
>64Mb (4Mx16) SDRAM $5.67
>Micrel 10/100 PHY $3.78
>4Mb (512KB) SPI flash $2.15 Total: $27.74
>
>Hmm, $26.23 for an ARM7TDMI @ 60MHz, 256KB flash, 512KB SRAM, 10mbps
>ethernet, vs. $27.74 for an ARM920T @ 200MHz, 512KB flash, 8MB SDRAM,
>10/100 ethernet. A whole $1.51 additional silicon cost.
Then take the EP9302 and add:
I took a brief look, but availability concerns me. Actually, while
getting into an alternate CPU discussion, I came up with an
interesting silicon cost comparison for some of my options. These
prices below are based on single unit quantities priced via Digikey:
Hmm, $26.23 for an ARM7TDMI @ 60MHz, 256KB flash, 512KB SRAM, 10mbps
ethernet, vs. $27.74 for an ARM920T @ 200MHz, 512KB flash, 8MB SDRAM,
10/100 ethernet. A whole $1.51 additional silicon cost.
Granted however that the EP9302 solution takes up more board space,
requires somewhat more complicated code to bring up the chip w/ cache
and MMU config, and there's a lot more core power decoupling
requirements (possibly necesitating going to 4 layers vs. 2 w/ the
LPC). I've been down the EP9301/2 route before, and I'll admit that
designing and bringing up an ARM920T microprocessor design *is* quite
a bit more complex than the LPC microcontrollers.. On the otherhand,
it can run eCos or Linux (if some more flash is added) with the Cirrus
board support package nearly out-of-the-box..
That bugs me that a full ARM920T ethernet solution comes that close to
cost to the ARM7TDMI LPC's. Its things like that which make our lives
difficult in selecting chips! Now if the ethernet LPC's w/ onboard
PHY's were on the scene in some reasonable timeframe, it would be a
no-brainer..
For anyone confused on my SPI flash choice above in the comparison,
the reason is that the Cirrus ARM920T's (and Atmel ARM920T if I recall
correctly) will happily boot off a SPI flash. Minimal pin and netlist
count for the flash interfacing (board routing mayhem, especially when
you have SDRAM on the bus too!), then I whip up a simple bootloader
that loads the firmware from serial flash to SDRAM. I usually
compress the serial flash image with an LZO variant and decompress in
the small bootloader program while loading to SDRAM.
-Ryan
--- In lpc2000@lpc2..., "Joel Winarske" <joelw@i...> wrote: >
> > Alternatives were Cirrus EP9301/2, but the package is large and I
> > still need an external PHY anyway. Atmel AT91RM9200 borders on cost
> > sensitivity and also requires external PHY. Not to mention external
> > flash and RAM. So I'm stuck with two chips minimum anyway, so
might
> > as well try to go with the LPC2292 as the most suitable CPU choice
and
> > select an ethernet solution around it.
>
> Have you looked at the Atmel AT91SAM7X?
> http://www.at91.com/Pages/products/microcontroller/AT91SAM/AT91SAM7X/at91sam > 7x.html
>
> Joel
>
Reply by Mark Butcher●January 7, 20062006-01-07
Hi Joel
I havent experienced this problem but if a powerful Coldfire gets
overloaded then the NE64 will certainly also have problems.
It is possible to specifically block ARP frames in the hardware
filter so that they dont disturb but this would mean that other
devices on the network wouldnt be able to find the NE64 since it
would not be able to react to ARP requests. If this can be
tolerated, the filer could be deactivated for the time the NE64 is
resolving local addresses, but it makes for a very special solution
rather than a general one.
Arp frames themselves are not very demanding since most can be
ignored if the IP address is is not equal to the local one (although
a local ARP table is generally maintained which includes all network
activity seen).
I have always understood that general broadcast frames cause most
overloading problems and it also certainly helps to not operate in
promiscuous mode on a network with high foreign activity levels.
Using switches to shield weaker devices can also help.
However at the end of the day there are real limits. If frames have
to be handled, this requires time and if the frame rate is faster
than the throughput then something will inevitably get lost. If a
100M LAN is really operating at the limit, each byte should be
processed in about 80ns and that requires a certain level of power.
It is difficult to say where the limit for a particular application
in a particular network lies but it doesnt make much sense to
connect a NE64 in promiscuous mode next to a Cisco router - probably
neither a Coldfire. For many other jobs they are nevertheless fine.
Cheers
Mark
www.mjbc.ch
--- In lpc2000@lpc2..., "Joel Winarske" <joelw@i...> wrote: >
> Hello Mark,
>
> > - For low-range embedded Internet applications the MC9S12NE64 is king. > > PCB design is extremely simple and a double
layer board works fine, > > even when using all peripherals - see some
photos at
> > http://www.mjbc.ch/pics/uTasker/uTasker1_eng.html (if only SPI/LAN is > > required, it can almost fit under a standard
RJ45 connector)
>
> I've experienced networks that brought high powered Coldfire based solutions > to their knees due to level of ARP traffic. How
does the MC9S12NE64 respond > with abnormally high ARP traffic?
>
> On the upcoming LPC Ethernet solution I really hope they spend some extra > effort on Ethernet hardware filtering and IP
hardware checksum via DMA. >
> Joel
>
