Sorry for the short notice, I forgot until today.  Hallmark's Will Philips and
a Rep. from Standard Microsystems Corp. will be here today to talk about SMC's
new COM20020 chip.

The COM20020 is a single chip LAN contoller with built in microsequencer and
2K bytes of SRAM.  It will transmit and recieve data at a rate of 2.5 MBPS
and will cost about $16 in quanities of 5K/YR.

From what I've read about the chip it seems very promising for a low cost,
low board space replacement for the current network board.  In fact if it is
as good as it appears, we won't even need to put it on a stack card, but can
put it right on the mother board.

I will be bringing the reps to the Eng. conf. room.  Please attend if your
interested.  Bring questions, we need to make sure this thing is for real.

Thanks,
Brian

Well, here goes again.  I wrote this file once in mail, but mail lost it and
I have to do it again...

Yesterday a few of us had a meeting with the sales rep from Standard
Microsystems Corp., SMC for short.  He confirmed some of the information I had
learned about a new chip they are offering, called the COM20020.  But we had
other questions which he could not answer, so he told me that an Engineer would
call today to answer the list we had formed yesterday.  For those of you who
don't know, I'll give a short list of the features of this ARCNET controller
chip as well as report on what I learned this morning.

Part number:	COM200X0
Vendor:		SMC
Price:		Approx. $16/5K/Yr (COM20020, 2.5MBPS)
Transmit speed:	2.5 MBPS
		5.0 MBPS binned.
Onboard SRAM:	2K x 8 in COM20020
		1K x 8 in COM20010
Protocol:	ARCNET

The COM20020 is a full featured network interface chip which is capable of
handling up to 512 byte packets.  It handles all network protocol including:
data encoding, transmision, receiving, decoding, error detection and correction,
and sends interupts to a host processor or microcontroller when packets are sent
or recieved.

If the host wishes to send a packet, then all he need do is transfer the data
into the COM20020's dual port SRAM, and tell the onchip microsequencer to send
the data.  When data is received the COM20020 interupts the host and if another
packet is recieved while the host is reading the first, the COM20020 will store
it in free memory.

The host can send broadcast packets or individual packets.

The best case time needed to broadcast a packet can be calculated as follows:

Tb = ((Ps + Is) * Bp * Tr) + Tt

Where:
	Tb is the broadcast time per node in microseconds.
	Ps is the packet size in bytes.
	Is is the information header size in bytes.
		2 byte CRC
		1 byte count
		3 byte address
		1 byte type
		1 byte preamble
		---------------
	  Total:8 bytes
	Bp is the bits per packet: 11.
	Tr is the transmission rate in us: 0.4 us.
	Tt is the time to pass the token: 25.8 us.

The equation can be expressed more simply:

Tb = ((Ps + 8) * 4.4) + 25.8

To give an example of use in our application lets run some test cases: assuming
we had a game the had to know what every other game was doing every 16 ms, which
is common in our applications.  The question is: depending on packet size, how
many nodes could be located on a single network, each node broadcasting one
packet to every other node within 16 ms.

Packet size	Tb	Max number of nodes
bytes		us
64		342.6	46
128		624.2	25
256		1187.4	13
512		2313.8	6

These are the best the part can do, if network errors are rampent, then
everything will slow down.  This is one thing we need to verify, that errors
are rare. 

With 5 MBPS parts, the number of nodes will double, so if in the furture we
need more data or more bytes per packet, then the option will exist to upgrade
to the faster part.

The current plan is to cludge some samples onto an existing board and see of
they perform as they should.

If you have any questions you can mail me (MCKEE), call me (X1731), or drop by
my office (cube outside pats office)

Thanks for reading,
Brian