Several ICs are manufactured in packages with more pins than it supports. For example, the LM317 in an SO8 package has 4 VOUT pins and 2 N/C (no connection) pins. I often want to run traces through N/C pins to ease routing, but wonder if it would make them give up the ghost. If it exists, what is the standard or rule by which manufacturers follow concerning the electrical characteristics of N/C pins? Or do I have to scour the datasheet / do my own testing every time?
4 Answers
It's a grey area. Some manufacturers tell you that's used for calibrating. Some manufacturers even will tell that you that certain pin has a function, only used by them for calibrating. Some tell you only not to connect it, or just say it's an unconnected pin.. You can't know for sure. The datasheet is information the manufacturer want to tell to you about using the device, but it might not be everything.
I recommended you do not connect them. If you get some generic IC from a different manufacturer or even batch the behaviour might be different. If you're engineering a project, you don't want to throw in unpredictability. You would have to test every single batch before you're going to use that particular batch. It depends on whatever you want to do that.
Normally "n.c."
means that the pin is not connected to the die, and you should be safe running a trace over it. In some rare occasions I've read "n.c. DO NOT CONNECT"
, which rises the question why I shouldn't, if it's not connected internally anyway. In any case I expect the datasheet to mention it expressly when I shouldn't connect to the pin.
One example is pin 4 on the LP2981. Texas Instruments says "Pin 4 (NC) must be left open. Do not connect anything to this pin"
, without further explanation. National specifies: "Post package trim. **do not** connect to this pin"
. So the pin seems to be connected as expected, otherwise it would be safe to connect. In this case "not connected"
should be read as 'no end user connection"
. Both manufacturers indicate clearly not to connect.
-
1
-
10\$\begingroup\$ Sometimes the "NC" pins have special diagnostic functions used at the factory. I think it's best to assume you cannot use the pins, rather than risk unspecified behavior. \$\endgroup\$ Jan 2, 2011 at 11:52
-
3\$\begingroup\$ Additionally the NC pins may be bonded together to one common point. \$\endgroup\$– Thomas OJan 2, 2011 at 11:56
-
1\$\begingroup\$ @Leon: yes, the infamous RFE, a.k.a.
reserved for future enhancements
. \$\endgroup\$– stevenvhJan 2, 2011 at 12:24 -
2\$\begingroup\$ @steven, I'd emphasize the answer to this question by saying "No, not unless explicitly allowed by the manufacturer." up front. \$\endgroup\$– Nick TJan 3, 2011 at 17:07
If you must do it, at least check an actual device with a meter - check resistance between the pin and +ve and the pin and ground using both polarites on the meter - if this shows infinite resistance it probably isn't bonded and is safe to connect.
-
4\$\begingroup\$ Also check continuity between individual NC pins. \$\endgroup\$– Thomas OJan 2, 2011 at 13:39
-
\$\begingroup\$ @Thomas: yes, you commented on stevenvh's answer that NCs may be connected together. Can you give an example? I wonder why they would do this, as it only costs money. \$\endgroup\$ Jun 14, 2011 at 14:14
-
2\$\begingroup\$ @Fredrico In some cases the N.C. pins can be connected to a substrate, or to the lead frame. This is to aid with heatsinking in high power devices where usually the best option to pull out heat is through the pins and into the copper traces on the PCB. I suppose it could also be used for noise supression e.g. in precision op-amps, using guard rings or similar. In either case, the N.C. pins would all be connected to each other. \$\endgroup\$– Thomas OJun 16, 2011 at 7:42
Sometimes. On the TDA89xx, the recommended layout has traces routed through the NC pins.