MAC Address Chips

Sep 05, 2003 17 Replies

I'm looking for "silicon serial number" type chips to give a unique MAC address for ethernet. I'm aware of the Dallas one-wire parts, but what i'd really like is an I2C/SMBus/TWI type solution. Or SPI.



I don't want to buy a block of MAC addresses from the IEEE and program EEPROMs during production. So I want guaranteed unique 48-bit numbers that don't clash with allocated MAC addresses.



Cheers TW


Dallas actually tried to make serial number chips pre-programmed with valid globally assigned MAC address a number of years back, but were blocked by IEEE.

If you are comfortable using locally assigned MAC address (see

formatting link
what I've done in one unit is use the lower 32 bits from a Dallas serial number chip (unfortunately, one wire bus) combined with a a fixed, 16 bit locally assigned prefix as the upper 16 bits of the MAC. It's a judgement call.

Did you check Rabbit Semiconductor? I think they sell uniquely serialized EEPROMS compatible with the Realtek 8019 ethernet chip.

The IEEE's position on this is unnecessarily against the "little guy", but the path you're taking isn't reliable.

The IEEE assigns the first 24 bits (OUI) of the MAC address, with the remaining 24 administered by the manufacturer. There is no way to randomly generate MAC addresses without the risk (albeit very low) of colliding with either a functional address (group addresses) or another vendor's OUI space.

The proper solution would be for IEEE to offer a low-cost option for registering small blocks within a multi-vendor OUI (e.g., $25 for 10 specific MACs). Some companies have tried to do this within their own OUI, and the IEEE actively squashes their attempts.

An alternative is to buy cheap / defunct Ethernet products, destroy them, and use their MAC, but then your device will appear to be from another registered manufacturer (who might object).

And, the MAC is usually burned into an EPROM. Why add a chip for the MAC?

Which IMHO does not mean that they could not be read out by a CPU and used with other chips too.

Markus

What's your point? He didn't specify whether or not a CPU would read the chip. In any case, it would be fairly trivial to bit-bang the MAC number out of the chip. I've done it.

Errr - yes, that was my point :)

Markus

Sorry, my parser only caught half of that double-negative.

The only way to guarantee that you won't clash is to get them from the source - ieee - sorry, that's life.

Andrew

Ted Wood wrote:

Those IEEE guys are real humps with those MAC addresses.

A couple years back a fellow was trying to do a public service for people who couldn't afford the thousands of dollars to buy a block, and bought a block he was going to divide up and sell small groups at cost.

The IEEE got on his case, yet they won't sell small blocks themselves.

Dirtbags.

The only choice you have to get unique IDs is to find a batch of discarded/obsolete network cards and use those IDs. Make sure to destroy the card to guarantee uniqueness.

Rufus

Here's a "what if":

Suppose someone set up a company that builds network cards. Customer makes a down payment on a batch, requesting the MAC id's in advance in order to finish development of the rest of the system. Company never delivers the network cards. Customer doesn't ask for down payment back, because he got what he really needed - the MAC numbers. See how it works?

There are "Individual Address Blocks" (IABs). 4096 addresses, for about $550, last time I checked.

formatting link

That's correct - I just registered one such block.

AFAIK, it's the least expensive way to get unique MAC addresses (besides cannibalizing old NIC's).

Tauno Voipio tauno voipio @ iki fi

Unless you've got free labor, I can't imagine that cannibalizing old NIC's would be cheap.

Grant Edwards grante Yow! ... If I had heart at failure right now, visi.com I couldn't be a more fortunate man!!

I meant jus stealing the addresses from them and then making sure that nobody's able to use the originals anymore.

Tauno Voipio tauno voipio @ iki fi

I know -- but even that is going to involve quite a bit of labor (espeically if the MAC aren't printed on the boards) when compared with the US$0.13 per address you paid for your 4K block.

Grant Edwards grante Yow! QUIET!! I'm being at CREATIVE!! Is it GREAT visi.com yet? It's s'posed to SMOKEY THE BEAR...

And depending on the NIC card maker, the MAC might not be unique. A number of them just cycled through their block. Some cycled through someone else's block. A few used the same MAC for all their cards in a production run.

For a hobbyist, re-using a MAC or two is no-biggy, but in a commercial environment, better safe than sorry and cheaper too.

Ron Sharp.

They do now sell blocks of 4096 for US $550:

formatting link

Not great pricing since blocks of 16777216 cost $1650.

I agree with you even though I'm an IEEE member.

If I had an IEEE-assigned OUI, and wanted to sell off Ethernet address space, I'd offer some lame product with an Ethernet port for sale for an outrageous amount of money (say $5000), expecting to sell exactly zero units. But I'd offer a "replacement" address EEPROM containing a guaranteed unique address dirt cheap.

Well here in South Africa, the trade unions have negotiated a minimum wage of approx 50 US cent per hour. In practice with the problems in Zimbabwe, labour is available at a 10th of this minimum wage. One will of course have to set up a system where unskilled labour is suitable.

Regards Anton Erasmus

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required