Roller Chain Selection Calculator

Which chain size, from the power and the speed — by the route that can be verified: design load against minimum breaking load over a speed-dependent design factor, with the multiple strand factors, the tooth-count and ratio limits, and every figure attributed to the maker who publishes it. No rating curve, because no rating table could be read reliably; and the breaking load is an input, because five catalogues disagree about it by up to 34% for the same ISO designation.

Last updated 24 Sep 2026

Roller chain selection

Power and speed → chain size

Power to transmit (kW)The motor’s nameplate power, or the power the machine actually needs — not both. The service factor below is what turns it into a design power.

Small sprocket speed (rev/min)Always the SMALL sprocket. It is the one that articulates most often per unit of chain and it sets the rating.

Teeth on the small sprocket17 or more for a normal drive (Renold Jeffrey). Diamond-Drives allow 12 at low speed but want 17 at medium and 25 at high.

Teeth on the large sprocket

Service factor1.0 for a uniform load on an electric motor, 1.3 for moderate shock, 1.5 for heavy shock, and up to 1.7 on an internal combustion engine with a mechanical drive. Work yours out on the service factor page linked below.

StrandsThe factor is not the number of strands. Tsubaki: “the loading is unequal across the width of the chain, therefore the transmission capability is not a direct multiple of the number of chains.”

Chain to checkThe page tells you the smallest adequate size whatever you pick here; this choice is the one whose margin is worked out in detail.

Chain pitch (mm)

Minimum breaking load, one strand (kN)Take this from YOUR chain’s catalogue. Five catalogues were compared in this batch and they disagree by up to 34% on the same ISO designation, because the standard sets a floor and each maker prints its own figure above it. The defaults here are the lowest of the five.

Where does the design factor come from?

Design factorBreaking load divided by working tension. Khurmi’s Table 21.2 gives 7 at 50 rev/min for every pitch, rising to 14.8 at 2,000 for a 12–15 mm chain and 19.5 at 1,200 for a 30–35 mm one. The whole table is below.

Not a circuit: the eight ISO 606 B-series chains drawn as bars of their minimum breaking load, all to one scale in kilonewtons, with the load YOUR duty requires drawn as a double line across them. Every bar that reaches past the line will carry the drive and every bar short of it will not, so the answer is something you read off the picture rather than off a table. The chain you selected above is outlined twice. Two things are worth noticing in the shape of the ladder. It is steeply non-linear — 16B-1 is more than three times 12B-1 for a pitch only a third larger — so going up one size buys far more margin than it looks like it should. And the bars are the LOWEST figure of five catalogues; the highest published figure for 05B-1 is 34% above the bar drawn here, which is why the breaking load on this page is an input.

11.22kNExample

5.5 kW at 1,450 rev/min on 21 teeth into 63, a 1.3 service factor, simplex, checked against ISO 10B-1

Design load against breaking load, over a design factor

P_design = P × K_s · P_strand = P_design ÷ K_m · v = p·N₁·n / 60000 · F = P_strand / v · F_break,required = n_design × F

K_s: service factor: driver type against the driven machine’s shock class. 1.0 to 1.7 in the chain family
K_m: multiple strand factor: 1, 1.7, 2.5, 3.3 for one to four strands. Dividing by it is what turns a whole drive’s design power into the power ONE strand must carry
v = p·N₁·n/60000: chain speed in m/s with p in mm. Note that at a fixed tooth count and speed, a bigger pitch runs FASTER and therefore carries less tension for the same power — which is half the reason a bigger chain helps
F = P_strand / v: working tension in one strand. Power over speed, nothing more. It ignores the centrifugal tension, which is negligible below about 10 m/s and is where this route starts to understate
n_design: breaking load over working tension. Khurmi’s Table 21.2 gives 7 at 50 rev/min for every pitch, rising to 14.8 at 2,000 rev/min for a 12–15 mm chain
F_break,required: what you take to a catalogue. Compare it against the minimum breaking load of the chain in front of you — not against a figure from a different maker’s table

Worked example

5.5 kW at 1,450 rev/min on 21 teeth into 63, a 1.3 service factor, simplex, checked against ISO 10B-1
Design power = 5.5 × 1.3 = 7.15 kW. Simplex, so the multiple strand factor is 1 and one strand carries all of it
Chain speed = 15.875 × 21 × 1450 / 60000 = 8.057 m/s
Working tension = 7.15 kW ÷ 8.057 m/s = 887 N
Khurmi's Table 21.2 gives 11.7 at 1,200 rev/min and 13.2 at 1,600 for a 12–15 mm pitch chain; interpolating to 1,450 gives 12.6375. A 15.875 mm pitch is outside that band and outside the next one, so it is read from the nearest — which the page says in the results rather than hiding
Required minimum breaking load = 12.637 × 887 N = 11.22 kN
ISO 10B-1's lowest published minimum is 22.2 kN, so the duty uses 50.5% of it and the chain achieves a design factor of 25.01 against the 12.64 asked for. Working down the sizes, the smallest that carries this duty is 08B-1 at 12.7 mm pitch — so 10B-1 is one size up, which on a drive that will be inspected twice a year rather than twice a month is usually money well spent

Every ISO 606 B chain against this page’s default duty, and what five catalogues disagree about

The default duty is 5.5 kW at 1,450 rev/min on 21 teeth with a 1.3 service factor, simplex. The required column changes with the chain, because a bigger pitch runs faster at the same tooth count and rpm and so carries less tension — and because Khurmi’s design factor is read from a different published band for a 12 mm chain and a 32 mm one. The three middle columns are the reason the breaking load is an INPUT on this page and not a lookup: iwis/JWIS, Cross+Morse, Wippermann, RS Components, Farnell print figures for the same designation that differ by up to 34.1%. ISO 606 sets a minimum; a maker’s own chain beats it by whatever margin that maker chooses, and only the maker’s own catalogue tells you which. The defaults here are the lowest of the five, which is the conservative choice and close to the figures usually quoted as the standard’s own minima. This page sizes a part; it does not certify one. Where the answer carries a consequence — a load path, a lifting duty, a pressure boundary, a fastener holding something that can fall — confirm it against the design code that governs the application, and against the manufacturer’s own rating, before relying on it.
Designation	Pitch (mm)	Roller Ø (mm)	Lowest minimum breaking load published (kN)	Highest (kN)	Spread (%)	Required at the default duty (kN)	Verdict
04B-1	6	4.00	3.0	3.2	6.7	29.67	too small
05B-1	8	5.00	4.4	5.9	34.1	22.26	too small
06B-1	9.525	6.35	8.9	10.4	16.9	18.69	too small
08B-1	12.7	8.51	17.8	19.4	9.0	14.02	carries it
10B-1	15.875	10.16	22.2	27.5	23.9	11.22	carries it
12B-1	19.05	12.07	28.9	32.2	11.4	11.42	carries it
16B-1	25.4	15.88	60.0	72.8	21.3	8.56	carries it
20B-1	31.75	19.05	95.0	105.0	10.5	8.65	carries it

Khurmi Table 21.2 — design factor for bush roller chain

Columns are the smaller sprocket’s speed in rev/min. Three things are worth reading off it. Every band starts at 7 at 50 rev/min, so the low-speed factor does not depend on the pitch at all. Every band rises with speed, which is the textbook’s way of acknowledging a failure mode this route cannot see directly — a breaking-load calculation has nothing in it about roller and bushing galling, which is what actually limits a chain at speed. And the dashes are not missing data: they are speeds the published table does not rate that pitch for, which makes them a speed limit stated in the only form that is defensible. The bands cover 12–15, 20–25 and 30–35 mm and have gaps either side; a pitch in a gap is read from the nearest band here and the results say which. Nothing is interpolated ACROSS bands, because the three rows are three separately published rows and not three samples of one surface. These dimensions come from a published standard’s table, not from a formula. The standard itself is cited below and the printed values are attributed to the catalogue they were taken from; a different publisher may round differently in the last digit.
Chain pitch	50	200	400	600	800	1,000	1,200	1,600	2,000
12 to 15 mm pitch	7.00	7.80	8.55	9.35	10.20	11.00	11.70	13.20	14.80
20 to 25 mm pitch	7.00	8.20	9.35	10.30	11.70	12.90	14.00	16.30	—
30 to 35 mm pitch	7.00	8.55	10.20	13.20	14.80	16.30	19.50	—	—

Multiple strand factors, and what each extra strand actually buys

Renold Jeffrey and U.S. Tsubaki publish the same five figures, which is why they are treated here as settled. The interesting column is the third: a duplex chain gives 85% of two simplex chains, a triplex 83.3% of three, a sextuplex only 76.7% of six. Tsubaki explain why in one sentence — “the loading is unequal across the width of the chain, therefore the transmission capability is not a direct multiple of the number of chains” — and the last column is the practical consequence: the first extra strand is worth 70%, the second 47%, the fifth only 18%. Past about four strands you are usually better off with a bigger pitch. Note that these are POWER factors. A duplex chain’s catalogue breaking load is typically about 1.75 times the simplex figure, slightly more than the 1.7 power factor, so working per strand against the simplex breaking load — which is what this page does — is the conservative reading by a few per cent. A drive is rated for the duty it sees, not for the power it nominally transmits. The service factor used here is stated; the manufacturer’s own factor for your machine class and daily running hours takes precedence over any general table.
Strands	Multiple strand factor	Capacity per strand	Shortfall against a direct multiple (%)	Gain over one fewer strand (%)
1	1.0	1.0000	0.0	—
2	1.7	0.8500	15.0	70.0
3	2.5	0.8333	16.7	47.1
4	3.3	0.8250	17.5	32.0
5	3.9	0.7800	22.0	18.2
6	4.6	0.7667	23.3	17.9

The drive limits this page checks, and who publishes each one

None of these is this page’s opinion. Each is a sentence in a named document, and where two publishers differ the page prints both rather than averaging them. The one figure that is genuinely contested is the wear limit, which is why it has a page of its own.
Limit	Figure	Source
Teeth on the small sprocket	17 or more	Renold Jeffrey’s drive checklist. Diamond-Drives give 12 for slow drives, 17 for medium speed and 25 for high speed.
Reduction in one step	7:1 optimum, 10:1 practical maximum	Renold Jeffrey: “speed ratio should be 7:1 or less (optimum) – 10:1 minimum”.
Centre distance	30 to 50 pitches	Renold Jeffrey and Tsubaki, independently. See the chain length calculator.
Wrap on the small sprocket	120° minimum	Tsubaki’s installation notes, as a requirement. 90° for a hanging drive only.
Design factor	7 at 50 rev/min, rising with speed	Khurmi Table 21.2, reproduced above.
Multiple strand factor	1.7, 2.5, 3.3, 3.9, 4.6	Renold Jeffrey and U.S. Tsubaki, agreeing.
Service factor	1.0 to 1.7	Renold Jeffrey and U.S. Tsubaki, agreeing cell for cell. Work yours out on the drive service factor calculator.
Wear replacement limit	1.5% to 3% elongation	Four publishers, four different figures. The chain wear calculator prints them all.

Why there is no rating curve here, and what is here instead

There is no horsepower rating curve on this page, and that is a decision rather than an omission. A chain’s real rating is a curve with two failure modes: link-plate fatigue sets the limit at low speed, roller and bushing galling sets it at high speed, and the useful rating is the lower of the two with a peak between them. Reproducing that needs a published rating table. Three separate attempts to read one in this batch — Martin Sprocket’s, a distributor’s and a maker’s — returned figures for the SAME chain, tooth count and speed that differed by up to 54% at 1,800 rev/min, and one of them put chain 40’s minimum tensile strength at 2,100 lbf, which is chain 35’s. That is the signature of a shifted column, and a rating read out of a shifted row is worse than no rating at all, because it looks authoritative. So this page takes the other published route and says which one it is taking.

The route it takes is the one MechanixCalc state for ISO 606 in one sentence: compare the design load against the chain’s minimum breaking load divided by a speed-dependent safety factor. Design power is the power you are transmitting times a service factor. Divide that by the multiple strand factor and you have what one strand must carry. Divide by the chain speed and you have a tension in newtons. Multiply by a design factor and you have a required breaking load in kilonewtons — which is a number you can take straight to a catalogue and look up, and which this page also turns into a chain size for you. Khurmi’s Table 21.2 supplies the design factor: 7 at 50 rev/min for every pitch, rising to 14.8 at 2,000 for a 12–15 mm chain. The whole table is printed below, gaps and all, and the gaps are the most honest part of it — they are speeds the table does not rate that pitch for.

What this route cannot see, stated plainly. It has no high-speed failure mode in it. A breaking-load calculation knows about tension and nothing about galling, so it will happily tell you a chain is fine at 3,000 rev/min when its maker’s rating table would not. Khurmi’s rising factor is a partial acknowledgement of that and not a substitute for it, which is why the chart above plots the required breaking load twice — once at a flat factor of 7 and once with Khurmi’s speed dependence — so you can see how much of the answer is the textbook’s correction rather than the physics. It also ignores the centrifugal tension, negligible under about 10 m/s and not above. Where a maker publishes a rating for your chain at your speed, that number wins, and it is not close.

The breaking load is an input, and this is the single most important thing on the page. Five catalogues were compared here — iwis/JWIS, Cross+Morse, Wippermann, RS Components, Farnell — and their minimum breaking loads for the same ISO 606 designation differ by up to 34.1%. ISO 05B-1 is published at anything from 4.4 to 5.9 kN; ISO 16B-1 from 60.0 to 72.8. That is not five people reading one table badly. ISO 606 sets a FLOOR, and each maker prints its own product’s figure above that floor by whatever margin its own steel and its own heat treatment allow. So the defaults here are the lowest of the five, and the field is editable, and the number that belongs in it is the one on the catalogue page for the chain you are actually going to buy. Use a competitor’s figure and you have silently borrowed up to a third of someone else’s safety margin.

Two other things the multiple strand factor tells you. It is 1.7 for duplex and 2.5 for triplex, not 2 and 3, and Tsubaki give the reason in one line: the load does not share evenly across the width of the chain. Per strand that is 85% for duplex and 83% for triplex, falling to 77% at six strands, so the first extra strand buys 70% more capacity and the fifth buys 18% — which is why past about four strands a bigger pitch is usually the better answer. And a subtlety worth knowing: those are POWER factors. A duplex chain’s catalogue breaking load is about 1.75 times the simplex figure, a little more than the 1.7, so working per strand against the simplex breaking load as this page does is conservative by a few per cent rather than optimistic. Once you have a size, the sprocket geometry calculator gives you every diameter the drawing needs, and the chain wear and elongation calculator tells you when to throw it away. For belt drives the same design-power idea leads somewhere quite different — the belt tension and shaft load calculator shows what tensioning costs the bearings.

Frequently asked questions

Why does this page not use a horsepower rating table?

Because it could not obtain one it could verify. A chain’s rating is a curve with link-plate fatigue at one end and roller and bushing galling at the other, and it has to come from a published table. Three attempts to read such a table in this batch returned values for chain 40 at 17 teeth that ranged from 0.69 to 0.80 hp at 100 rev/min and from 5.82 to 8.96 hp at 1,800 — a 54% disagreement at the top end — and one of the same extractions put chain 40’s minimum tensile strength at 2,100 lbf, which is chain 35’s figure. Rows were being shifted. Publishing a rating curve built on that would have been the worst thing this page could do, so it uses the breaking-load route instead and says so.

Is the breaking-load route conservative or optimistic?

Conservative at low speed and optimistic at high speed, and the page says which side you are on. At low speed a design factor of 7 against the minimum breaking load is a genuinely cautious number. At high speed the route has no failure mechanism in it for galling, which is what actually kills a fast chain, so it will pass a drive a maker’s table would refuse. That is why the page prints the fastest speed Khurmi’s table covers for your pitch, and warns you when you are past it. Past the table, get the maker’s rating.

Why is the minimum breaking load something I have to type in?

Because five catalogues disagree about it by up to 34% for the same designation. ISO 05B-1 is published at 4.4 kN by iwis and 5.9 kN by Farnell; ISO 16B-1 at 60.0 and 72.8. Both are honest: ISO 606 specifies a minimum and a maker whose chain exceeds it prints what its chain actually does. The defaults here are the lowest of the five, which is also closest to the figures usually quoted as the standard’s own minima. Replace them with the figure from the catalogue page for the chain you are buying.

Why is a duplex chain not worth twice a simplex one?

Because the load does not share evenly across the width. Renold Jeffrey and U.S. Tsubaki independently publish 1.7 for two strands, 2.5 for three, 3.3 for four, 3.9 for five and 4.6 for six, and Tsubaki explain it in a sentence: “the loading is unequal across the width of the chain, therefore the transmission capability is not a direct multiple of the number of chains.” Per strand that is 85%, 83%, 83%, 78% and 77%. The practical reading is in the last column of the strand table: the first extra strand buys 70% more capacity, the fifth buys 18%. It also assumes both sprockets are cut as one piece; two separate sprockets side by side do not share load the way the factor expects.

How many teeth should the small sprocket have?

Renold Jeffrey’s drive checklist says 17 or more, and Diamond-Drives give it against speed: 12 for slow drives, 17 for medium and 25 for high. The mechanism is chordal action. The chain leaves a sprocket as a polygon, not a circle, so its speed varies by 1 ÷ cos(180°/N) over every tooth — 1.73% at 17 teeth, 3.53% at 12, 6.42% at nine. That variation is a cyclic load on everything downstream, and it is invisible to a breaking-load calculation, which is why the figure is printed in the results rather than folded into the answer.

What is the maximum ratio for one chain reduction?

Renold Jeffrey: “speed ratio should be 7:1 or less (optimum) – 10:1 minimum”. Two independent things go wrong as the ratio climbs. The wrap angle on the small sprocket collapses, and Tsubaki want at least 120° of it. And the large sprocket becomes the dominant cost and the dominant inertia in the drive. There is a third, more surprising one: the chain-length formula every catalogue prints is a series expansion whose accuracy depends on (R₂ − R₁)/C, and at 10:1 on 30-pitch centres that parameter reaches 0.6, where the formula is out by more than a whole link.

Does the service factor include the daily running hours?

In the V-belt family yes, in the chain family no — and that is a real difference between the two, not an oversight in one of them. The chain service factor tables Renold Jeffrey and U.S. Tsubaki publish have exactly two dimensions, the driver type and the driven machine’s shock class, and no hours axis at all: hours enter a chain drive through the rating table’s own life basis instead. The V-belt tables have three bands of daily hours. Khurmi’s textbook chain method has a third shape again, with an explicit hours factor AND a lubrication factor. All three are on the drive service factor calculator, printed side by side rather than reconciled.

Related calculators

References

Renold Jeffrey. Roller Drive Chain Selection. The source for the chain service factor table used here (uniform 1.0 / 1.0 / 1.2; moderate 1.3 / 1.2 / 1.4; heavy 1.5 / 1.4 / 1.7, by electric motor or turbine, internal combustion with hydraulic drive, and internal combustion with mechanical drive), the multiple strand factors (1, 1.7, 2.5, 3.3, 3.9, 4.6), and the drive checklist: “small sprocket should have 17 or more teeth”, centre distance “within the optimum range of 30–50 pitches”, and speed ratio “7:1 or less (optimum)”.
U.S. Tsubaki. RS Chain Drive Selection (chains.ustsubaki.com). Confirms the same service factor table and the same multiple strand factors cell for cell, from a different maker — which is why those two tables are treated here as settled rather than as one publisher’s opinion. Also prints the chain speed relation S = P·N·n/12 (inches, rpm) and the chain length and centre distance forms.
Tsubakimoto Chain. Coefficients Used in Selection, chain-guide.com 4.1.2. States the reason the multiple strand factor is not the number of strands: “the loading is unequal across the width of the chain, therefore, the transmission capability is not a direct multiple of the number of chains.”
MechanixCalc. ISO 606 Calculator — Roller Chain Selection & Sprocket Geometry. Prints the ISO 606 sprocket forms d = p/sin(π/z), da = p(0.6 + cot(π/z)) and df = d − d_roller, so the tip-diameter form is the same in the ISO family as in the ANSI one; and states the selection route used here — “the chain is selected by comparing the design load (nominal power × service factor) against the chain’s minimum breaking load divided by a speed-dependent safety factor”.
R.S. Khurmi and J.K. Gupta, A Textbook of Machine Design, chapter 21 Chain Drives and chapter 20 V-Belt and Rope Drives (the chapters as distributed by Al-Mustansiriyah University and Al-Mustaqbal University). Source for Table 21.2, factor of safety for bush roller chain against the smaller sprocket’s speed; for the service factor as K1 × K2 × K3 (load, lubrication, hours); for the recommended teeth on the smaller sprocket against velocity ratio; for the V-belt tension ratio 2.3 log(T1/T2) = μθ cosec β with β the groove HALF angle and a groove angle of 32° to 38°; for the centrifugal tension Tc = mv²; and for the maximum-power condition Tc = T/3.
iwis / JWIS. Catalogue: precision chains, British Standard roller chain table. The lowest of the five minimum-breaking-load columns compared in this batch, and the one that matches the figures usually quoted as ISO 606’s own minima (16B-1 at 60.0 kN, 20B-1 at 95.0 kN). Used as the default on the selection page.
Cross+Morse. British Standard Precision Roller Chain Conforming to ISO 606. Second of the five ISO 606 transcriptions cross-checked here.
Wippermann, via TransDev. Simplex to DIN 8187 (ISO 606-1982 & SMS 1613). Third of the five, and the one that also notes that electrogalvanised or nickel-plated chains may reach only 80% of the stated tensile strength.
RS Components. Simplex roller chains according to DIN 8187-1 (European standard), datasheet 0900766b80b613a7. Fourth of the five.
Farnell. Chain Drives — roller chain, datasheet 17402, conforming to BS 228, ISO R606 and DIN 8187. Fifth of the five, and the highest column in every row.
ISO 606:1994, Short-pitch transmission precision roller and bush chains, attachments and associated chain sprockets (second edition, 1994-02-15). Cited by number; only the standard’s preview was reachable, which confirms clause 3.6 (a measuring force applied when a chain’s length is measured) and clause 5 (chain wheels, described in its own foreword as ‘the unification of all the relevant national Standards’). The dimensional figures here are taken from five named catalogues instead — see the breaking-load note.
PFlow Industries. Roller Chain Maintenance and Lubrication. “In most roller chain drives, the chain is considered worn out when it has reached 3% wear elongation”; “allowable wear is limited to 200/N (N = number of teeth on largest sprocket)”; and the point this batch used to drop a max-speed column: “the recommended type of lubrication [is] shown in the horsepower tables in the respective standards”, types I/II/III being manual or drip, bath or slinger disc, and oil stream or pressure spray.
Diamond-Drives by Timken. Frequently Asked Questions. Maximum allowable wear elongation “approximately 3% for most industrial applications, based upon sprocket design” and “approximately 1.5%” where centres are fixed or the drive must run smoothly; the 200/N relationship, N being the teeth in the large sprocket; and minimum tooth counts against speed — 12 slow, 17 medium, 25 high. Its stated range of application for 200/N contradicts Reliable Plant’s; see the note on the wear page.