Resistors in Series and in Parallel: The Formulas, and the Three Things They Do Not Tell You

Two formulas, and what each tool returns

In series the same current passes through every resistor in turn, and each one drops its own share of the voltage, so the voltages add and the resistances add with them: Rs = R1 + R2 + … The series calculator does exactly that and nothing else. Give it its default list of 100, 220 and 330 and it returns 650 Ω.

In parallel the same voltage sits across every resistor and each carries its own current, so the currents add. Current is V/R, so what adds is 1/R — the conductances — and the total resistance is the reciprocal of that sum: 1/Rp = 1/R1 + 1/R2 + … The parallel calculator returns 56.90 Ω for the same three values, and it gives you rather more than a number: enter a voltage and it also reports the total current, the total power, and the current through each individual resistor. What it does not report is the power in each individual resistor, which is the figure that decides whether the circuit survives, and we will come back to that.

One consequence is worth stating as a rule, because it is the fastest sanity check there is: a parallel combination is always smaller than its smallest member. It has to be. Adding another resistor in parallel adds another positive term to the sum of conductances, which can only make the sum bigger and therefore the reciprocal smaller. Put a 1 MΩ resistor across a 1 Ω resistor and the total is 0.999999 Ω — a millionth less than the 1 Ω, and still less. If a parallel answer ever comes out above the smallest value on your list, you have made an arithmetic mistake or a typing mistake, and the second is more likely.

One member usually owns the answer

Nudge one resistor in a parallel pair and the total moves — but not by the same amount for both. The sensitivity of the total to a given member is Rp divided by that member's value. In 100 ∥ 220 the sensitivities are 68.75/100 = 0.6875 and 68.75/220 = 0.3125, and they add to exactly one, which is not a coincidence either: change everything by the same factor and the total changes by that factor, so the shares of the change must sum to one. Practically, a 1 % error in the 100 Ω moves the answer by 0.69 %, while a 1 % error in the 220 Ω moves it by 0.31 %. Widen the gap and the effect gets brutal: in 100 ∥ 1000 the small resistor owns 90.9 % of the answer, and you could be badly wrong about the 1 kΩ and barely notice.

The mirror image holds in series, and the numbers are literally the same pair swapped over. In 100 + 220 the sensitivity of the total to each member is that member divided by the total: 100/320 = 0.3125 and 220/320 = 0.6875. There it is the large resistor that owns the answer, which makes intuitive sense — in series the biggest resistor sets most of the total, in parallel the smallest one does. The tidy way to remember it: in each arrangement, the member that matters is the one doing most of the work.

Tolerance: two 5 % parts in parallel are still 5 %

It is widely believed that combining resistors averages out their tolerance. It does not, not in the worst case. Run the numbers: 100 ∥ 220 nominal is 68.75 Ω. Take both parts 5 % low and you get 95 ∥ 209 = 65.3125 Ω, which is exactly 5 % low. Take both 5 % high and you get 105 ∥ 231 = 72.1875 Ω, exactly 5 % high. This is not an approximation, it is an identity: the parallel combination is homogeneous of degree one, meaning that scaling every resistance by the same factor scales the result by that factor. If both parts are at the same end of their tolerance band, the network is too. The same holds in series, for the same reason.

What does shrink is the statistical spread, and only if the two errors are genuinely independent. Combine the sensitivities from the previous section in quadrature — 5 % times 0.6875 and 5 % times 0.3125, root sum of squares — and you get 3.78 % for 100 ∥ 220. For two equal resistors the same arithmetic gives 3.54 %, the familiar factor of one over root two. Note that this figure is identical for the series and the parallel arrangement, because the two sensitivity pairs are the same numbers swapped. It is a real reduction, but it is an expectation rather than a guarantee, and the guarantee is what a specification has to be built on.

The independence assumption is exactly where it goes wrong on a bench. Two resistors torn off the same reel came from the same batch, on the same machine, on the same day, and their errors are correlated rather than independent — which pushes you back towards the worst case, not the statistical one. They also share a temperature: warm the board and both drift the same way, and a common-mode drift is the worst possible kind for this argument. If you actually need a value tighter than the parts you have, the way to get it is a tighter part, not two loose ones. The only case where combining genuinely helps is trimming: a large resistor in parallel with a much smaller one nudges the small one down by a controlled amount, and because the sensitivity to the large one is tiny, its own error barely enters the result.

Power is where people actually get burned

Both resistors in a parallel pair see the same voltage, so each dissipates V²/R and the power goes as one over the resistance. The smaller resistor therefore takes the larger share, and the share is the same fraction we already met: Rp divided by that resistor. Put 12 V across 100 ∥ 220 and the total is 2.09 W, of which the 100 Ω takes 1.44 W — 68.8 % — and the 220 Ω takes 0.65 W. Widen the gap to 100 ∥ 1000 and the 100 Ω takes 90.9 % of a smaller total. So the member you were least worried about, the one that barely moves the resistance, is also the one doing almost none of the heating, and the small one is doing all of it.

That is why a quarter-watt part is the wrong choice more often than people expect. A 0.25 W resistor of 100 Ω can take at most the square root of 0.25 times 100, which is 5 V across it — and the moment your parallel pair sits on 12 V, it is being asked for nearly six times its rating. The parallel tool will show you the current through that resistor, 120 mA, and the total power of the network, but it will not multiply the two out per resistor, so nothing on screen says the word that matters. Do it by hand: for each member, square the voltage and divide by that member's resistance, and compare against the part's rating with margin — half of rating is a reasonable working ceiling for anything that has to last.

In series everything inverts, which is the useful half of the symmetry. The same current runs through both, so each dissipates I²R and the power goes as the resistance: the largest resistor runs hottest. With 100 + 220 carrying 50 mA, the 100 Ω dissipates 0.25 W and the 220 Ω dissipates 0.55 W — the same 68.8 % share, on the other resistor. The rule to keep is that in each arrangement the member taking the most power is the one where the calculator is least likely to warn you, because it is the member you are paying least attention to.

What the two calculators will and will not accept

The series tool takes one text field of values separated by commas, semicolons or spaces, and it is strict about what a number looks like. Type 100, 220, 330 and it works. Type 1k, 2k2 and it returns zero, because neither token parses as a number and both are silently dropped. Type 10 kΩ and it returns 10 Ω, because the space splits the entry into 10 and kΩ, the second is discarded, and the first survives as a bare ten — a silent factor of a thousand, with no warning of any kind. The same trap catches a space used as a thousands separator: 10 000 becomes 10 and 0, and the total is 10.

The parallel tool is more forgiving in a way that is arguably worse, because it uses a lenient parse per box: type 1k in a field and it reads 1, type 4k7 and it reads 4. You get a number rather than a blank, and the number is wrong by three orders of magnitude with nothing to hint at it. In both tools the fix is the same and dull: type plain ohms, digits only, with a decimal point if you need one. Write 4700, not 4k7.

One behaviour is worth flagging as physically wrong rather than merely fussy. The parallel calculation drops any value that is not strictly positive, so entering 0 alongside 100 and 220 gives 68.75 Ω — the zero is ignored. Electrically a zero-ohm path in parallel is a dead short and the correct total is zero ohms, which is exactly the answer you would want to see if you had just typed a shorted link into your netlist. The guard is there to stop a division by zero, and the cost is that the one input that should scream comes back silent. The series tool has the opposite gap: it does not filter at all, so a negative value is summed as written and 100 and −50 give 150 Ω in the parallel tool but 50 Ω in the series one.

Frequently asked questions

Why is the parallel total always below the smallest resistor?

Because you are adding paths, not obstacles. Whatever current the smallest resistor was already passing, it still passes when you wire another resistor beside it — and the new resistor adds current of its own. More total current for the same voltage means less resistance by definition. The algebra says the same thing: the reciprocals add, so the sum of conductances is strictly larger than the largest single conductance, and the reciprocal of a bigger number is a smaller number. A useful corollary for estimating in your head is that a very large resistor in parallel barely changes anything: 1 MΩ across 1 Ω gives 0.999999 Ω, so if one member is a hundred times the other you can usually ignore the big one to within one per cent.

Can I put two resistors in parallel to get a value I do not have?

Yes, and it is a legitimate technique with three costs to weigh. The first is tolerance: as this article showed, combining does not tighten the worst case, so if you needed a precise value you still need a precise part. The second is power, because the two members will not share it evenly and the smaller one takes the larger share — size it for what it actually gets, not for half the total. The third is that you have doubled the number of joints and components that can fail, on a board where the whole point of the exercise was to avoid ordering one part. Where it genuinely earns its place is trimming a standard value slightly downwards: put a large resistor in parallel with your nearest preferred value and you move it by a controlled, small amount, and because the sensitivity to the large one is low its own tolerance barely matters.

Do two 0.25 W resistors in parallel make a 0.5 W resistor?

Only if they are equal. Two identical resistors in parallel do split the power evenly, so a pair of 0.25 W parts of the same value handles 0.5 W between them — that is a real and common trick, and it also halves the resistance, which is often the actual reason for doing it. Two different values do not share evenly at all. In 100 ∥ 220 the smaller part takes 68.8 % of the total, so a pair of quarter-watt components fails when the network passes about 0.36 W, not 0.5 W, because the 100 Ω hits its own limit first. Always check the worst-loaded member rather than the total: square the voltage across the pair, divide by the smallest resistance, and that is the number the smallest part has to survive.

Why does the parallel tool show currents but not powers?

It shows the total resistance, the total current and the total power, plus a breakdown of the current in each individual resistor — but the per-resistor power is not on screen, and that is the one number that decides whether a component survives. You can recover it in one step from what is displayed: multiply each resistor's current by the voltage you entered. In the three-resistor default at 12 V the currents are 120 mA, 55 mA and 36 mA, so the powers are 1.44 W, 0.65 W and 0.44 W. Notice that the first is already six times a quarter-watt rating. If you take one habit away from this article, make it that multiplication — it takes a second, it uses only numbers the tool already gave you, and it is the check that stops a resistor going brown.

How should I type values like 4k7 or 10 kΩ?

Do not. Type the plain number of ohms: 4700 for 4k7, 10000 for 10 kΩ, 0.22 for 220 mΩ. Neither tool understands engineering notation, and neither one tells you so. The series calculator drops any token it cannot read as a number, so 1k, 2k2 comes back as a total of zero, and 10 kΩ comes back as ten ohms because the space splits it and the letters are discarded. The parallel calculator's per-box parse is more permissive and therefore more dangerous: it reads the leading digits and stops, so 1k becomes 1 and 4k7 becomes 4. In both cases the failure is silent and the error is a factor of a thousand, which is exactly the sort of mistake that survives a design review because the number on screen looks like a number.