I am working on a project and while doing the component selection I realised that many microcontrollers have MAC & PHY layer on the different chip. Except for TM4C1294 microcontroller (there might be many out there).

So my question is is there any specific reason why MAC & PHY layer job is exported to dedicated chip out of controller ?

  • \$\begingroup\$ there probably is. you can only guess at specifics. you could ask the manufacturer ... probably cheaper to use a separate device \$\endgroup\$ – jsotola Mar 13 '18 at 8:17
  • 1
    \$\begingroup\$ There was also several Cortex-M3 MCUs from the TI Stellaris range (originally from Luminary Micro) that had integrated PHYs (they may be obsolete now, I don't know). There are also a few others from Microchip, e.g. PIC18F97J60. \$\endgroup\$ – dim Mar 13 '18 at 8:37
  • 1
    \$\begingroup\$ There is also a Coldfire with built-in PHY. \$\endgroup\$ – filo Mar 13 '18 at 10:05
  • 1
    \$\begingroup\$ My compact car does not have a pickup truck bed because the car manufacturer recognized that not everyone needs or wants a pickup truck bed. \$\endgroup\$ – kkrambo Mar 13 '18 at 12:09
  • \$\begingroup\$ Got it @kkrambo. Appreciate your help. \$\endgroup\$ – Mayank Mar 13 '18 at 18:08

Powerful high speed output drivers take up a lot of surface area of the chip. And to operate them you need significant power.

When developing a chip you have power and surface area budgets that you have to work with.

There is a limited amount surface area, putting the output drivers on the chip means some other features have been removed or simplified. Or the chip needs a much bigger, and more expensive package.

There is also a limited amount of power you can put in a package, so running an Ethernet PHY might mean you have to reduce the specifications of other parts of the chip.

CAN bus tranceivers and USB are also examples of peripherals that take up significant surface area.

I can't seem to find any die photographs, maybe someone else can?

  • 3
    \$\begingroup\$ Anonther (marketing) reason is that external PHY give much more flexibility. Different products have different requirements (e.g. auto MDI/MDIX, dignostics capabilities, wake-on-lan, builtin switch, ...). So having the PHY integrated would restrict the applications for which the MCU could be used. \$\endgroup\$ – dim Mar 13 '18 at 8:31
  • 1
    \$\begingroup\$ NXP actually has a nice little Cortex M0 called LPC11C24 with both CAN transceiver and controller on-board, in a LQFP48. This is the only MCU I've seen with that feature though. \$\endgroup\$ – Lundin Mar 13 '18 at 15:09
  • \$\begingroup\$ @Lundin there is no ethernet support in LPC11C24 ! \$\endgroup\$ – Mayank Mar 13 '18 at 18:12
  • \$\begingroup\$ @Mayank It wasn't a shopping recommendation, but merely a comment stating that it is possible to put these things on-chip. \$\endgroup\$ – Lundin Mar 14 '18 at 7:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.