7

I'm interested into know what semiconductor manufacturing process use the most common PIC and AVR microcontrollers, in particular 8 bit microcontrollers.

Obviously they are not at the same level of modern CPUs and GPUs (~20 nm), what is the gap between these two "categories"? Are microcontrollers still in the order of micron?

Is the microcontrollers frequency limit (tens of MHz) only due to the worst manufacturing process or are they other factors that affect their maximum operating frequency?

Is just the extremely high cost of better manufacturing process that prevents its use in cheap microcontrollers? Or are there other reasons?

Oceanic815
  • 1,179
  • 1
  • 11
  • 19

2 Answers2

7

Finer processes require lower supply voltages, which tends to lead to lower core than I/O voltages, and multiple pins being required for power supplies. For example, 3.3V maximum on the I/O and 1.x V on the core. That's inconvenient in a small system.

Also, a lot of the die is used up by the (fixed-size) I/O bonding pads and drivers, so making the rest of the chip finer does not yield as big returns in terms of reduced chip acreage.

The mask sets are much cheaper for the coarser processes (maybe tens of thousands of dollars rather than ~$1M for a state-of-the art mask set), so it makes it more practical to have a wide range of different products.

Microchip PIC products are actually produced in at least one fab (Gresham OR) that was previously used to make (at that time) state-of-the-art memory chips (Fujitsu).

enter image description here

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • One main feature of a microcontroller is that is has GPIO. These pins need to be able to drive 5v or 3.3v logic chips connected to it. – benathon Feb 12 '14 at 09:18
5

http://ww1.microchip.com/downloads/en/DeviceDoc/DS-00169F.pdf page 4:

The baseline N-well CMOS process is released on mature manufacturing lines that support 1.2, 0.9, 0.7, 0.5, 0.4 and 0.25 micron technologies

That's across the whole PIC range, so the smaller PICs are probably on the larger process nodes.

Cost is the major factor; as is power consumption (both static and leakage), and the associated thermal design. Note that PIC offer larger and faster microcontrollers (you can do video on a PIC32), as do AVR. There's also little point in staying 8-bit if you actually want to do a lot of processing; you might as well get a larger data bus and a more convenient memory architecture than the 8-bit PIC one.

A minor issue is that at higher clock speeds you can't read from Flash in a single cycle, so you need at the very least an instruction cache to maintain execution speed.

EMC (noise emission) is more of a problem at higher speeds too.

pjc50
  • 46,540
  • 4
  • 64
  • 126
  • Both answers are good. I can't accept two answers, so I upvoted. Thanks for the data from Microchip, it's very interesting know what manufacturing process they use. So essentially Microchip is about 15 years back compared to the best manufacturing process. – Oceanic815 Feb 13 '14 at 14:52