James Bryant Consultant Engineer - Analog Electronics
SOME THOUGHTS ON THE CONNECTION OF UNUSED PINS
One of the commoner causes of problems during circuit development is incorrect connection of integrated circuit (IC) terminals which are not actually used in that particular application - "unused pins". This paper considers the different types of unused pin, and how they should be treated.
There are three reasons why there can be unused pins on an IC. The first, very common, reason is that IC packages are standardised and the most convenient package for a particular IC may have more pins than the IC chip has terminals.
Next, some ICs are made with pins which are used for testing and/or calibrating the device during manufacture. These pins will often (but not always) be unused during normal device operation.
The final, and most obvious, reason is that ICs are often complex devices performing many functions on a single chip, some of these functions may not be required in a particular application, and in such a case the associated pins are not used. One example is an active filter using three operation amplifiers - it may save space and money to use a quad op-amp (four op-amps in a single chip/package) and there will be an unused op-amp left over; another is a Analog-Digital Converter (ADC) with a built-in reference which may be used with an external reference instead for reasons of accuracy, stability or system calibration, leaving the internal reference output unconnected.
The labelling of unused pins
In the first two cases above the pins will be labelled to indicate that they are redundant. Unfortunately there is no standard for this, even for data sheets from a single manufacturer. The commonest label is "No Connect" or "No connection" (NC) but this may have two different meanings:- [A] there is no connection from the IC chip to this package pin, or [B] there is a connection from the chip to this pin, and it must not be connected to anything.
The best data sheets distinguish between these two cases by the use of NC for an internally unconnected pin, and "Internal Connection" (IC) for a pin which is internally connected and must be left without any external connection but, sadly, this is often overlooked in the hurry to prepare a data sheet1. Where only IC is used the "Pin Function Description Table" which should follow the "Pin Configuration Diagram" should indicate what is meant, but frequently it does not. Occasionally when these resources have failed the information is to be found buried in the "Circuit Description & Applications" section at the back of the data sheet, but usually it is not, and even when it is present it is well-buried and easy to overlook.
It may be possible to measure the impedance at a pin to determine if it is unconnected or internally connected, but a high impedance input pin may have very low leakage and be very hard to distinguish from an unconnected pin. If it is important you should contact the Applications or Product Engineering department of the IC manufacturer and ask. Do not rely on information from sales, marketing or distribution sources as they are likely to play for safety and advise you to make no connection rather than take the trouble to determine if it is actually safe to do so.
Connecting pins without an internal connection
Leaving unconnected pins open circuit risks their behaving as antennas and coupling high frequency (HF) noise into the IC. This is not usually a serious problem, but may be in systems where there is possible capacitive coupling of HF noise, so the best thing to do with unconnected pins is to connect them to ground. Where this is impracticable connect them to other low-impedance nodes which are not noisy.
Of course sometimes it is convenient to allow a PC track to connect to a nearby unused pin for layout reasons. There is rarely any harm in this but if the integrated circuit is sensitive to low-level or HF signals it is wise to verify that any signal on such a PC track does not affect the IC's operation.
In general such tracks should not carry a voltage greater than 80-100V.
Connecting pins with an unknown internal connection
If the data sheet, or the Product Engineer, says that one or more pins of an IC must be left unconnected then those pins should be left open circuit. Nothing should be connected to them, and unless you are explicitly told to do so they should not be connected together either.
But this may not be sufficient - if there is a danger of capacitively-coupled noise it may be necessary to shield them from it with a grounded guard ring on the PCB, or even with a metal shield or can. To be honest, this is rarely necessary, but problems can arise from such noise coupling 2 and it is worthwhile verifying that there is no potential problem, and shielding if there is.
The problem with such pins is that unless the Product Engineer is unusually forthcoming we do not know what is connected to them. It is best to assume the worst and expect them to be sensitive to noise of any sort and vulnerable to low-level ESD, and to design your system and PCB layout so that they are not exposed to such possible dangers.
Connecting the pins of unused sections of an IC
Integrated circuits contain many electronic components interconnected to create a complex electronic system in a single chip of silicon. The complex system will contain a number of simpler sub-systems and sometimes the user of an IC will not need all of them. The question then arises - what to do with the terminals of the unused part(s) of the IC?
The most important thing is to do nothing that will interfere with the correct operation of the parts of the system you are using. After that the unused system(s) should be configured, as far as is possible, to use as little power as possible, and to disable any un-needed signal sources. This may involve hardware programming some of the digital inputs, and may also involve writing the correct configuration code to the IC during power-up - which will involve software, rather than hardware, changes. Since such configuration is different for every IC I shall not discuss it in any detail, merely to note that the principle is to configure the device so that all un-needed subsystems are, as far as possible, shut down.
Many ICs are configured by logic inputs rather than software. This means that one or more inputs of the device are hard-wired to logic 0 or logic 1 (and sometimes leaving an input open defines another logic state). These inputs should not be considered "unused" even though they never change state - they are programming pins and must be connected correctly.
What to do with unused inputs
In general unused inputs should not be left open circuit.
By our definition an unused logic input is one whose input state does not affect the operation of the parts of the system we are using, and does not affect the power consumption or oscillation of the parts we do not need. (Inputs which do affect power or oscillation were dealt with in the previous section.)
Usually the best thing to do with such logic inputs is to ground them. However, if there is an internal pull-up resistor or current source then grounding the input will cause the pull-up current to flow, which is a (very small) waste of power. If this power waste is unacceptable leave the pin open-circuit or, better if it is liable to encounter RF or electrostatic fields, connect it to logic 1.
All analog inputs have a small bias current flowing in the input device. Many types have the bias current supplied by resistors or current sources on the chip but some devices, particularly operational amplifiers (op-amps), must be provided with a bias current path lest they latch up, permanently, or temporarily while the bias is missing, and draw excess current. We shall discuss these later.
Analog inputs with on-chip bias should not be grounded at DC as their circuitry may saturate and waste current. They may simply be left open circuit but, as we have already seen, such open circuit pins are sensitive to pick up of noise, RF or electrostatic fields, so it is better to ground them to AC with a 10-100 nF ceramic capacitor.
The input of a direct coupled amplifier without internal bias can often be grounded, turning off the amplifier without causing excess current draw. But sometimes taking the input too close to Vdd or Vss increases current draw (amplifiers which behave in this way may mention it on their data sheet, or you may need to experiment to determine their Vin/Ipower variation). Amplifiers where this happens should have their input connected to a potential somewhere between the supply rails - see the end of the next section.
If an op-amp is overdriven, the output stage will saturate at one of the supply rails, and the op-amp will consume some excess power, often this is not very much, but sometimes it may be substantial. Many common configurations of an unused up-amp will overdrive it. (An overdriven op-amp may also take a long time [relative to its bandwidth] to recover from overdrive - this is not important in the present circumstances but should not be overlooked in cases where it might matter.)
If the terminals are all left unconnected, there is a real risk that stray electrostatic fields will cause an input to go outside the supply rails. This can sometimes cause latch-up and destroy the whole chip. Even if latch-up does not happen, a DC field may cause saturation and power waste, or the amplifier may amplify an alternating electrostatic field and, if it saturates, modulate its own supply current sufficiently to cause crosstalk to other devices on the chip.
Some users connect one input to the positive supply and the other input to the negative supply. This again saturates the output and wastes power; and it may also exceed the differential input voltage rating and damage the device. Even if damage does not occur, some input stages draw several tens of milliamps under these conditions, wasting even more power.
Grounding both inputs, or shorting them together at some other potential, also causes the output stage to saturate, since the offset voltage of an op-amp is never exactly zero; shorting them together and not biasing them has the same latch-up risks that we have already mentioned.
What we should do is connect the device as a follower (output to inverting input4) and connect the non-inverting input to a potential somewhere between the supply rails. With a dual-supply system, ground is ideal, but connecting to the positive or negative supply of a single supply system will cause saturation and the resulting power waste if the offset voltage has the wrong polarity. The "potential somewhere between the supply rails" may be any point in the circuit with a suitable potential, since the loading caused by the op-amp input is minimal.
Or you might use it as a buffer amplifier in a part of your system that does not need one but might perform slightly better if it had one.
Inputs and outputs of analog switches/multiplexers
Analog switches are symmetrical devices where a logic level determines whether or not there is an analog signal path between two terminals. Multiplexers are arrays of analog switches which connect one terminal to just one of an array of others under logic control, allowing us to select one analog signal of many. Analog switches are symmetrical so we may consider both analog terminals as "inputs".
These analog inputs of analog switches and multiplexers must never be left floating if they are unused but must always be connected to ground or some other convenient potential5 between Vdd and Vss of the analog switch.
Unused digital outputs should always be left unconnected.
Analog outputs include the outputs of amplifiers, voltage references, and oscillators of various sorts. Most of these should also be left open circuit but some may oscillate without capacitive, or, very occasionally, resistive, damping. Such oscillation may disrupt the operation of other parts of the system and must be prevented by the use of an appropriate capacitor or resistor - the data sheet should make it clear when this is necessary.
Current outputs are not very common and in general it will be possible to power down unused current output terminals. Active unused current outputs will often require a pull-up (or pull-down) connection to a supply or to ground to prevent incorrect operation of other parts of the circuit. (Kirchoff's Law says, in effect, that a current always has to go somewhere - it cannot just disappear - and if a current output is not terminated the current may go and disrupt something on the chip.) This pull-up/down will generally be a short circuit but may sometimes need to be a resistor so that power dissipation takes place in the resistor and not in the IC chip.
Power and ground pins
All ground pins should be grounded. Where a system has analog and digital grounds the rules concerning the separation (or not) of analog and digital grounds should followed with great care6. Most ICs that have power dissipation tabs require that they should be grounded - even if the dissipation is low enough that the tabs need not be connected to a heat sink of any sort they must still be grounded.
In general all power pins should be connected to the correct power line - and correctly decoupled. Occasionally it is possible to save power, and safe, to disconnect (and sometime ground - see the device data) power connections to unused parts of the chip but be very cautious when doing this and don't unless you are quite sure that that it is safe.
To summarise:- even if a particular terminal of an IC is unused in some applications it is still necessary to understand its function, what potentials are (or should be) present on it, what currents may (or must) flow to/from it, how sensitive it is to electrostatic or RF influences, and if it requires capacitive or resistive loading of any sort. In other words, as we have so often said before, RTFDS1 - and do what it tells you.
1 Read the Friendly Data Sheet. See the link from RAQ 4 "Caveat emptor".
2 The first radio I designed for manufacture was badly desensitised by coupling between an ungrounded quartz crystal can in the frequency synthesiser and an unused pin in the IF amplifier IC some 11 mm away. The production engineer did not want to have to solder a ground lead to a metal crystal can during production but until this was done the receiver sensitivity was degraded by over 30 dB.
3 See RAQ 46 "What shall we do with the unused op-amp?"
4 If the op-amp is a current-feedback type, or has a minimum stable closed-loop gain (Acl) >1 the connection should be made with a 100 kΩ resistor rather than a short-circuit to ensure stability and prevent possible oscillation.
5 See RAQ 3 "Isolating the Key Detail (or Lunching With a Mermaid and Pickled Herring)."
6 See RAQ 9 "Grounding Converters (or Philosophy to the Rescue!"