Analog input specifications love to lead with a bit number - 12-bit, 14-bit, 16-bit - and it is easy to read that as a measure of how accurate the module is. It is not. Resolution tells you how finely the module chops its input span into discrete steps, which sets the smallest change it can possibly distinguish, but a module can resolve tiny steps while still reading several of them off from the true value. Confusing the two leads to under-spec'ing a level measurement you needed to trend closely, or over-paying for bits you will never use. This guide explains how ADC resolution maps a signal into counts, why resolution and accuracy are separate, and how to calculate the smallest engineering-unit change a module can actually see.
Analog Input Resolution in one line: Analog input resolution is the number of discrete steps, expressed as bits, that a module's analog-to-digital converter divides its full input span into. An n-bit converter produces 2 to the n counts across the span, so a 12-bit module resolves the span into 4,096 steps and a 16-bit module into 65,536. Resolution sets the smallest change the module can represent, but it is not accuracy - accuracy is how close the reading is to the true value, and it is limited separately by gain, offset, drift, and noise.
An analog input module takes a continuous signal - a 4 to 20 mA current or a 0 to 10 V voltage - and its analog-to-digital converter turns that into an integer count that the PLC can read. The number of bits sets how many distinct counts are available. A 12-bit converter has 2 to the 12th, or 4,096, possible codes; 14-bit gives 16,384; 16-bit gives 65,536. Spread those counts across the input span and you get the size of one step. For a 0 to 10 V input on a 16-bit converter, the span is 10 volts divided by 65,536, so one count represents roughly 0.15 millivolts. On a 12-bit converter of the same span, one count is 10 volts divided by 4,096, or about 2.4 millivolts - sixteen times coarser.
For a 4 to 20 mA loop the arithmetic is the same but you work over the 16 mA of live span. A 16-bit module resolves 16 mA into 65,536 steps, so each count is about a quarter of a microamp; a 12-bit module resolves the same span into 4,096 steps at roughly 4 microamps per count. Note a subtlety with 4 to 20 mA: some of a converter's full range may be reserved for the below-4 and above-20 regions used for out-of-range and fault detection, so the counts actually spread across your live 4 to 20 mA band can be fewer than the raw bit count implies. Reading the module's documented count-per-span, rather than assuming the full 2-to-the-n applies to your usable range, keeps you honest about how fine the steps really are.
Resolution is the size of the smallest step the module can represent; accuracy is how close its reading sits to the true value of the signal. A module can have both a very fine resolution and a mediocre accuracy at the same time. Think of resolution as the number of tick marks on a ruler and accuracy as whether the ruler was manufactured to the correct length and read in the right light. A 16-bit module gives you a very finely divided ruler, but gain error, offset error, temperature drift, and the module's own internal noise can still push the reading off the true value by more than a single count. Vendors specify accuracy separately, often as a percentage of full scale, and it is that number - not the bit count - that tells you how much to trust an absolute reading.
This is where effective resolution matters. The last one or two bits of a high-resolution converter are frequently buried in noise, meaning the least significant counts flicker even when the input is perfectly steady. The number of bits that are actually stable and usable - the effective number of bits, or noise-free resolution - is smaller than the nameplate bit count. A part advertised as 16-bit might deliver perhaps 13 or 14 noise-free bits in a real panel. The practical lesson is to distinguish three things when you spec a module: the nominal bits, which set the theoretical step; the effective bits, which set the smallest change you can reliably see; and the accuracy spec, which sets how far the reading can be from truth. All three come from the datasheet, and treating the headline bit number as if it captured all three is the classic mistake.
To decide whether a module is fine enough, convert its resolution into your engineering units. Suppose a tank spans 0 to 30 feet mapped onto a 4 to 20 mA transmitter, read by a module that resolves the 4 to 20 mA span into, say, 32,000 usable counts. Thirty feet divided by 32,000 counts is a bit under 0.001 foot per count, so the module can in principle distinguish sub-millimeter changes in level. If instead the module resolves that span into only 4,000 counts, each count is about 0.0075 foot, and you can see roughly a hundredth-of-a-foot move. Now compare that against the smallest change you actually need to observe. If your process control requires seeing a 0.1 percent move on the level - about 0.03 foot here - both modules clear it easily; if you needed to catch a 0.005-foot creep, only the finer one qualifies. Doing this calculation before you buy prevents both under-spec and over-spec.
Fine resolution only pays off if the readings are stored and viewed at a matching fidelity, and that is where the SCADA layer comes in. A cloud platform such as Merobix historizes analog values continuously and lets you zoom a trend down to see small, slow moves - a tank level creeping up, a pressure drifting - that a coarse module or a coarsely stored value would flatten into a staircase and hide. When you are trying to catch a slow leak or a gradual fouling that shows up as a fraction-of-a-percent drift, the chain has to be fine end to end: enough effective bits at the module, enough accuracy to trust the number, and a historian that preserves the detail so an operator can actually see the move on a chart.
Not automatically. More bits means finer resolution - smaller steps and the ability to represent smaller changes - but accuracy is a separate specification governed by gain error, offset, drift, and noise. A 16-bit module can resolve tiny steps yet still read off from the true value by more than one count if its accuracy spec is loose. To compare modules properly you look at both the bit count and the stated accuracy, usually given as a percentage of full scale.
Effective resolution, sometimes called effective number of bits or noise-free resolution, is how many of a converter's bits are actually stable and usable rather than flickering in noise. The lowest one or two bits of a high-resolution converter often jitter even with a perfectly steady input, so a part sold as 16-bit might deliver only 13 or 14 truly stable bits in a real installation. It is the effective bits, not the nameplate bits, that set the smallest change you can reliably measure.
Divide your measurement's engineering-unit span by the number of counts the module spreads across its usable input range. For example, a 0 to 30 foot level on a module that resolves its live span into 32,000 counts gives about 0.001 foot per count. Then compare that step against the smallest change your process needs to see. If the step is finer than what you need to observe, the module's resolution is sufficient; if it is coarser, you need more bits or a narrower span.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.