2
\$\begingroup\$

I created a circuit in which all the address inputs of an AT28C256 are connected to outputs of shift registers (74HC164).

In the future I will do many of these circuits, however I haven't yet decided if I need the full 32KB of memory.

Let's just assume that I later decide I need only 8K of memory for my application (for example, AT28C64).

After looking at the datasheets, I notice the difference in pin-out between the two chips is that on the AT28C64, A13 and A14 are replaced with NC (no connects).

Is it OK anyway to connect such pins to outputs of shift registers (and be able to make a AT28C64 like a drop-in replacement for AT28C256) or are the NC pins internally connected to something?

All chips in my question are DIP ICs.

\$\endgroup\$
2
  • 4
    \$\begingroup\$ Well no connect could mean do not connect or does not connect. Read the fabulous data sheet \$\endgroup\$ – Andy aka Mar 9 '18 at 0:42
  • 1
    \$\begingroup\$ Yeah. What Andy said. And if the datasheet is not absolutely clear, contact the vendor. They can usually answer these types of questions. Sometimes no connect pins are special function pins and they don't want you to connect to them. Other times there may not even be a wire bond between the silicon and the pin, so you are free to do whatever you want with it. \$\endgroup\$ – mkeith Mar 9 '18 at 3:18
4
\$\begingroup\$

The datasheet of the AT28C64 makes it pretty clear: There are "No Connect" (NC) pins, those are not connected in the chip so you can hook them up to a voltage. There are also "Don't connect" (DC) pins which you shouldn't connect but the DIP package doesn't have any of those. That's the obvious part.

The not so obvious part is that there are multiple versions of the AT28C64 and some of them have a READY output instead of a NC on pin 1. To avoid shorting things out, you should put a 330 Ohm (or so) resistor in series with pin 1 of the EEPROM so that if you plug in an AT28C64 with a READY output, the current gets limited.

With that resistor in series with pin 1, swapping in an 8K ROM instead of the 32K one should work just fine.

\$\endgroup\$
2
\$\begingroup\$

What JEDEC says...:

no (internal) connection (pin) (NC)

A terminal that has no internal connection and that can be used as a support for external wiring without disturbing the function of the device, provided that the voltage applied to this terminal (by means of the wiring) does not exceed the highest supply voltage rating of the circuit.

NOTE 1 If higher voltages are acceptable, this should be stated.

NOTE 2 The IEC equivalent term is "blank terminal"; nevertheless, the IEC abbreviation is "NC".

References: JESD21-C#, 1/97 JESD77-B, 2/00 JESD78A#, 2/06 JESD99B, 5/07

Source: https://www.jedec.org/standards-documents/dictionary/terms/no-internal-connection-pin-nc

More JEDEC:

NC, NO CONNECTION: A pin to which no internal electrical connection is present or allowed.

Source: https://www.jedec.org/sites/default/files/2_00R20.pdf

Real Life

Have a look at "Recommendation for NC Pins (Devices: 33912G5, 33911G5 & 33910G5)" from Freescale Semicoductors: https://www.nxp.com/docs/en/application-note/AN3829.pdf

You will quickly notice that (a) this document does not deal with EEPROMs and that (b) not all NC pins are equal. Even on one single device, different NC pins can have different permitted voltages! There is also at least one Freescale device where a group of NC pins is connected internally to each other (but not to an other group of NC pins in the same IC).

Texas Instruments seems to define "NC" as "No connect (DO NOT CONNECT) externally. Failure to leave NC pins open can cause faulty operation.". On Semiconductor seems to think the same.

ST says (for an EEPROM, so this is, in the strict sense, somehow relevant to the original question): "No connect (NC) pins should be tied to VSS.". Also, "No function (NF) pins should be tied to VSS."

Atmel seems to consider NC an "input" pins (even though without a function) and allows the same voltage range as any other (digital) input pin. Pins which should be left floating (unconnected) are marked "DU" ("Don't Use") by Atmel.

At least one manufacturer mentions that a close-by RF source can inject noise into an IC when an NC pin is left floating. In that (quite possibly rare) case, they recommend tying it to DC (GND/Vcc).

Is it safe to assume that IC manufacturers will do the "right thing"?

Nope. While in a few select cases, manufacturers have indeed taken into account that another of their ICs could be used as a drop-on replacement (like a newer/larger *PROM, which is deliberately made pin-compatible with the older part), they usually don't. Like the 28C256, which is not a possible drop-in replacement for the 27C256. Fun!

TL;DR

You are safe with the scenario you describe with the Atmel AT28C256/AT28C64. I am not aware of any mainstream EEPROMs where an NC pin might have an unpleasant effect when connected to a "regular signal" (just don't connect it to an RF noise source). However, it cannot be completely ruled out that some exotic devices have NC pins with internal connections.

\$\endgroup\$
1
\$\begingroup\$

For memory chips in particular, pins that get used as address pins in larger parts and are NC pins in smaller parts can usually be connected. This does exactly what you say, i.e. it allows you to upgrade the size of the memory by just changing the part.

In general, except in cases where the manufacture is specifically supporting compatible footprints between different parts, you should not connect to the NC pins.

\$\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.