What the rating on the plate actually says
A duty cycle is meaningless without an amperage, and it is always stated as a pair. Sixty percent at two hundred amps means that at two hundred amps the machine can carry the arc for six minutes out of any ten and needs the remaining four with the output off and the fan running. Thirty percent at three hundred amps on the same plate means three minutes of arc and seven of cooling at that higher current. These are two points on one curve describing the same machine.
The ten minute period is the part people get wrong. It is not a ten minute block that starts when you press the trigger and resets when you take a break. It is a rolling window, and the machine is heating and cooling continuously. Six minutes of continuous arc followed by four minutes off satisfies the rating. So does welding for ninety seconds out of every two and a half minutes, which is the same ratio arrived at a different way and is much closer to how real work actually happens.
What the rating is describing is heat in the output components. Current through a transformer, an inverter stage or the output diodes produces heat proportional to the square of the current, and the cooling system removes heat at a roughly fixed rate. The rating is where those two balance at the manufacturer test conditions.
Why the derating goes as the square of current
Because the heat does. Resistive heating goes as current squared, so doubling the current produces four times the heat and the machine can therefore sustain it for a quarter as long. The relation that follows is that the allowed duty at a new current equals the rated duty times the square of the ratio of rated current to new current.
| Machine rated 60% at 200 A | Allowed duty | Arc minutes in ten |
|---|---|---|
| Running at 150 A | 100% (relation gives 107, capped) | 10 |
| Running at 175 A | 78% | 7.8 |
| Running at 200 A | 60% | 6.0 |
| Running at 250 A | 38% | 3.8 |
| Running at 300 A | 27% | 2.7 |
This is the standard relation manufacturers use and it is a good approximation of how a machine behaves, but it is an approximation. Real machines have their own published curve, thermal protection that trips at a temperature rather than at a calculation, and cooling that is not linear. If the machine came with a duty cycle chart, that chart beats this arithmetic. Where the arithmetic earns its place is in answering the question the chart does not: what does that mean for the number of parts I get out of an hour.
The two limits, and which one you are actually against
There are two independent ceilings on output and only one of them is the machine. The first is your own cycle time: arc time plus handling time is how long a part takes, and sixty divided by that is parts per hour if nothing else intervenes. The second is the machine: the allowed duty cycle multiplied by sixty gives arc minutes available per hour, and dividing by arc minutes per part gives parts per hour the machine will sustain.
The realistic rate is the lower of the two, and knowing which one is binding is the whole point of the exercise. If the handling time is binding, a bigger machine buys nothing at all and the improvement has to come from fixturing, part flow or a second pair of hands. If the machine is binding, the answer is a lower amperage, a longer cycle, a bigger machine or a second one, and no amount of shop floor efficiency will help.
A worked case. Machine rated 60 percent at 200 amps, running at 200 amps, arc time 1.5 minutes per part, handling 1 minute. The cycle is 2.5 minutes, so the natural duty is 60 percent, exactly at the limit, and the rate is 24 parts an hour using 36 arc minutes. Now improve the handling to 30 seconds. The cycle drops to 2 minutes, the natural duty rises to 75 percent, and the machine will not do it. The rate stays at 24 parts an hour and the extra half minute you saved becomes cooling time. All that work on the fixture bought nothing, because the constraint moved to the machine and nobody checked.
Consumables run on arc hours, not clock hours
Every consumable in a welding cell is consumed while the arc is on and not otherwise, which is why the burn rate calculation goes through arc hours per working hour rather than through parts. In the example above, 24 parts an hour at 1.5 arc minutes each is 36 arc minutes, so 0.6 arc hours per working hour. Wire at 4.5 pounds per arc hour is therefore 2.7 pounds an hour of clock time, and shielding gas at 35 CFH is 21 cubic feet.
That gas figure is the one that catches people out. An 80 cubic foot cylinder covers less than four hours of production at that rate, and that is before pre-flow and post-flow, before purging the line after a cylinder change, and before the gas that runs while somebody is finding out that the trigger works. Ordering gas on the assumption that a cylinder lasts a week is how a job stops on a Friday afternoon.
Contact tips and nozzles are deliberately left as inputs with no default, because there is no honest default. Tip life varies with wire type and size, current, whether the wire is clean, whether the liner is worn, and how much spatter the settings produce, and the spread between a good setup and a bad one is more than an order of magnitude. Track your own over a few weeks and put the real number in.
The things that make the rating optimistic
A duty cycle is measured at a stated ambient temperature with clean airflow. Everything that departs from that reduces it. A shop at 95 degrees in August, a machine pushed into a corner where the fan recirculates its own hot air, a filter clogged with grinding dust, an intake blocked by the leads coiled on top of it: each of those is a real reduction in the cooling side of the balance and therefore in the arc time you get. None of it appears on the plate.
The thermal protection tripping is the visible symptom and it is not the whole cost. Running consistently at the edge of the rating ages the output components even on the occasions when the protection never operates, and the failure that eventually results is the expensive kind. If the calculation on this page says you are at 95 percent of the allowance every hour of every day, the machine is undersized for the work regardless of whether it has tripped yet.
To get the arc time per part in the first place, the travel speed calculator turns a joint length and a bead size into arc minutes and a wire consumption figure, which are the two inputs this page most wants. For the cost per joint rather than the cost per shift, the weld cost calculator works from the geometry directly. And if the constraint turns out to be electrical supply rather than the machine, the shop tool circuit grouping calculator covers what else can be on the same supply.
The hazards of the welding itself are named in full on the travel speed calculator. A cell running near its duty cycle all day is also a cell producing fume all day, which makes the ventilation question more pressing rather than less.
Questions people ask
Does the ten minutes restart when I stop for a moment?
No, and this is the most common misunderstanding about duty cycle. The window is rolling, not resetting. The machine is heating whenever the arc is on and cooling whenever it is not, continuously, and the rating describes the ratio at which those balance. Welding six minutes and resting four satisfies it. So does thirty-six seconds of arc in every minute, which is the same ratio and is much closer to real production. What does not satisfy it is nine minutes of arc followed by a minute off, even though that is ten minutes containing some rest, because the machine will already have overheated at around six.
What happens when I exceed it?
The thermal protection operates. The output cuts, an indicator comes on, and the fan continues to run until the machine has cooled enough to reset, which typically takes several minutes and cannot be hurried. On a long weld this means the arc stops mid-pass, which leaves a stop that has to be cleaned and restarted properly. The protection is doing its job and no damage is done by an occasional trip. What does cause damage is habitually running at the edge, because the components age with thermal cycling whether or not the protection ever operates.
My machine says 20% at maximum. Is that useless?
Not useless, just honest about a corner of its range. Twenty percent at the maximum output means two minutes of arc in ten at that current, and small machines are rated that way because their maximum output is a short burst capability rather than a working setting. What matters is the duty cycle at the current you actually weld at, which is usually well below maximum and therefore has a much higher allowance. Put your real working amperage into this page rather than the maximum, and the picture is normally far less alarming than the headline figure suggests.
Is a water cooled torch a duty cycle upgrade?
It raises the duty cycle of the torch, which is a separate rating from the machine. Torches, guns and leads all have their own thermal limits, and on higher current work the torch is often the binding constraint rather than the power source. Water cooling addresses that end of the problem and does nothing at all for the power source, so if the machine is what is tripping, a water cooled torch will not change it. Check which component is actually getting hot before spending money on either.
How do I measure my real arc time per part?
Time it directly, with a stopwatch, on several parts, and take the average rather than the best one. Most people substantially overestimate their arc time because the whole cycle feels like welding, and the arc is on for a smaller fraction of it than it seems. If you can calculate the travel speed for your settings and bead size, arc time also falls out of the joint length arithmetic, which is what the travel speed calculator does and is a useful cross-check against the stopwatch. When the two disagree, the stopwatch is right and the difference is usually restarts, repositioning and pauses that nobody counted.