16
\$\begingroup\$

Take a look at this example schematic:

Example Schematic

The chip in question is the PIC18F4550 and as you can see power is provided on both sides (with a 100nF cap to smooth noise I guess). Is this strictly required or could you put power into one side and just ignore the other? I know that I have just put power in one side and everything still seems to work fine but I feel uneasy with doing that. I feel that they would not waste pins that could be doing something useful by having an extra pair of power in/out pins.

For what purpose does this PIC MCU (and others like it) have multiple power pins?

\$\endgroup\$
14
\$\begingroup\$

Each Vdd and Vss pin (and AVdd and AVss, if present) must be connected, and each pair should be decoupled. Multiple supply pins with their own decoupling ensure that different parts of the chip receive a clean low-impedance supply, which wouldn't be the case if the supply was distributed inside the chip itself.

For decoupling to be effective, the connections to the pins should be short with wide tracks to minimise inductance. The supply track should go to the capacitor and then to the pin, just putting the capacitor on stubs with the supply and ground going to the pins first isn't a good idea.

\$\endgroup\$
  • \$\begingroup\$ To add: While it's not likely the case with a micro-controller, sometimes is beneficial to have separate power supplies for different parts of the chip where a specific part is sensitive to clean power. \$\endgroup\$ – user606723 Nov 23 '11 at 17:55
  • \$\begingroup\$ How can a part be sensitive to "clean power"? \$\endgroup\$ – Leon Heller Nov 23 '11 at 18:14
  • 1
    \$\begingroup\$ Perhaps I should've said, "sensitive to the cleanliness of the power"? DACs often have separate analog and digital voltage sources. \$\endgroup\$ – user606723 Nov 23 '11 at 18:16
8
\$\begingroup\$

As stated in the datasheet, all power and ground pins must be connected. This includes analog power and ground too (AVdd, AVss). Each power pin should have its own bypass capacitor as close as possible to it.

Multiple Vdd and Vss pins are probably internally connected to a point. However, due to constraints of the die, each of these feeds is separately required. The PIC may appear to run with only one pair connected, but some voltages inside the chip are then probably not as intended and various subtle things might go wrong.

\$\endgroup\$
  • \$\begingroup\$ Thankyou for the excellent response and that is exactly what I suspected; that subtle problems may be occurring without my knowledge. But what page in the datasheet did you find that they must both be connected? I swear I tried to find that information before asking this question and could not. \$\endgroup\$ – Robert Massaioli Nov 23 '11 at 13:50
  • 2
    \$\begingroup\$ @Robert: I know I've seen it, and it's a well known issue anyway. I did just take a look at the 18F4550 datasheet and didn't find it either. Perhaps not all datasheets mention it, or it's in the family reference manual or something. In any case, I know I've seen this explicitly stated by Microchip, just can't remember exactly where or by whom. \$\endgroup\$ – Olin Lathrop Nov 23 '11 at 14:36
  • 1
    \$\begingroup\$ It is mentioned in the data sheets for the 16-bit parts. \$\endgroup\$ – Leon Heller Nov 23 '11 at 18:42
4
\$\begingroup\$

While you can run a PIC off just the one pair of power/ground it isn't advised.

  • One power pin can only safely provide a certain amount of current. Having 2 doubles that current capacity.

  • The ground pins are often also heat-sinks, so not connecting them could cause too much heat to build up in part of the chip.

  • Multiple decoupling capacitors with separate feeds into the chip will improve the overall current draw and signal cleanliness of the PIC.

So yes, for simple experimentation and prototyping you can just use the one pair if you want, but for anything in production use them all.

\$\endgroup\$
  • \$\begingroup\$ More likely it is the inductance rather than resistance that is the most critical factor for a high speed micro. \$\endgroup\$ – Martin Nov 23 '11 at 13:46
  • \$\begingroup\$ I did not know that about the ground pins also being heat sinks, I guess I just never thought about it but it makes a lot of sense. And when you say separate feeds to you mean two complete separate power sources or two separate lines from the same power source (I've only taken one elec course at uni; I come from Computer Science so please forgive the probably simple question). I will take that advice to heart. \$\endgroup\$ – Robert Massaioli Nov 23 '11 at 13:54
  • \$\begingroup\$ Same lines, separate decoupling capacitor. \$\endgroup\$ – Majenko Nov 23 '11 at 13:59
  • \$\begingroup\$ @Majenko: Oh, okay. But why do I need two? Is it the physical proximity that makes it better? Because would not the two capacitors essentially be in parallel nodally speaking? ...actually is there something that I could read that explains why both capacitors are required? \$\endgroup\$ – Robert Massaioli Nov 23 '11 at 14:06
  • 2
    \$\begingroup\$ @Robert: inductance, inductance, inductance. \$\endgroup\$ – Jason S Nov 23 '11 at 14:25
4
\$\begingroup\$

You must connect them all.

Not doing so is a recipe for interesting times. Often the core will come up, and everything may seem to work, but then you start getting odd failure modes - on-chip peripherals not working, I/Os not driving to full voltage, etc

\$\endgroup\$
  • \$\begingroup\$ It sounds like you may have actually encountered those problems before? If that was true that would be pretty final. \$\endgroup\$ – Robert Massaioli Nov 23 '11 at 13:57
2
\$\begingroup\$

Processors that have multiple power supply pins do so to lower parasitic resistance and inductance from the power supply -- low parasitic inductance is very important at high frequencies for a processor that has fast supply current transients.

If you do not connect all power supply pins, with appropriate bypassing (i.e. capacitors connected with short traces), the power supply voltage as seen on the chip will experience voltage transients that may be outside the operating spec for the part.

\$\endgroup\$

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.