A robotic welding cell that ran 98% arc-on time for six months suddenly drops to 91%. Spatter increases. Rework hours climb. The robot path, wire feed speed, and gas flow are unchanged. The difference: the welding power source was swapped from the robot manufacturer's integrated model to a standalone third-party unit during a retrofit — and the 3-millisecond communication lag between the robot controller and the separate power source is enough to desynchronize arc start from torch motion at travel speeds above 1.2 meters per minute. This article traces how the physical and control distance between a robot and its welding power source changes what happens at the arc.
The Arc Start: Where Latency Matters Most
A gas metal arc welding (GMAW) arc start happens in three events sequenced within 50 milliseconds: wire approaches workpiece (robot motion), power source detects contact and ramps current, wire retracts to establish arc gap. If the robot is still moving toward the start point when the arc fires, the wire stubs into the workpiece. If the robot has already begun travel when the arc is still stabilizing, the first 3–5 mm of weld bead have insufficient penetration.
An integrated power source shares a real-time backplane with the robot controller — arc start and motion start are synchronized within the same control cycle, typically 250–500 microseconds. A separate power source communicates over a fieldbus (EtherNet/IP, PROFINET, DeviceNet) with a cycle time of 1–8 milliseconds. The 3 ms difference may sound trivial, but at 1.5 m/min travel speed, the torch moves 0.075 mm in 3 ms — and arc initiation is a nonlinear process where the first 2 ms of current ramp shape the weld pool nucleation. Integrated architectures win the arc start consistency battle by a margin that weld inspectors can see on radiographs of the first 5 mm of bead.
Waveform Control During Travel: The Sync Problem
Pulsed GMAW varies current between a high background level and a peak pulse 200–400 times per second. Each pulse detaches one droplet of filler metal. In an integrated system, the robot's velocity profile feeds forward into the power source's pulse timing: as the robot accelerates out of a corner, the power source increases pulse frequency to maintain constant fill rate per millimeter of travel. As the robot decelerates into the next corner, pulse frequency drops. This keeps the deposited metal volume per linear millimeter constant regardless of robot speed.
A separate power source does not receive velocity feedforward — it responds to the robot's actual position via fieldbus, always one to two position samples behind. At high travel speeds with frequent direction changes (tight weave patterns, small-diameter pipe welds), this lag produces visible variation in bead width: wider in deceleration zones, narrower in acceleration zones. The difference is typically 0.5–1.0 mm of bead width variation for a separate system versus under 0.3 mm for an integrated one — enough to fail cosmetic criteria on visible welds, though usually within structural tolerance for non-appearance-grade applications.
Spatter: The Cost of Disconnection
Welding spatter — tiny droplets of molten metal ejected from the arc — is the number-one consumable cost in a robotic welding cell after filler wire and shielding gas. Spatter builds up on the gas nozzle, eventually restricting shielding gas flow and causing porosity. It also coats fixtures, requiring periodic manual cleaning that stops production.
The spatter rate correlates directly with the precision of arc-voltage-to-wire-feed-speed matching during transient conditions — starts, stops, and path direction changes. Integrated systems maintain this match within one control cycle. Separate systems experience brief mismatches at every speed transient: the power source is still responding to the previous speed while the robot has already changed velocity. A study by a major automotive OEM reported a 35–50% spatter reduction when switching from a separate to an integrated welding architecture on a chassis line running 1.8 m/min travel speed — translating to roughly 40 fewer minutes of nozzle cleaning per shift. The robot controller architecture directly affects consumable cost.
When Separate Makes Sense
Separate welding power sources are not obsolete. They win in three scenarios:
- Multi-process cells that switch between GMAW, FCAW, and GTAW on the same robot — a standalone multi-process power source is easier to reconfigure than an integrated unit tied to one robot brand
- High-deposition-rate applications (above 10 kg/hr deposition) that need power sources with specialized waveforms (tandem-MIG, submerged arc) not available in integrated form factors
- Retrofit and brownfield cells where the robot is already installed and the welding power source is being upgraded independently — replacing only the power source costs 60–80% less than replacing the entire robot-power-source package
Travel Speed Limits for Each Architecture
- Integrated: 0.8–2.5 m/min GMAW
- Stable arc start, synchronized pulse, consistent bead width within ±0.3 mm across speed changes. The practical speed ceiling for thin-sheet automotive applications.
- Separate (fieldbus): 0.5–1.5 m/min GMAW
- Adequate for structural fabrication, heavy plate, and non-appearance welds. At speeds below 0.8 m/min, the communication latency becomes negligible relative to the weld physics — separate and integrated perform similarly.
- Separate (hardwired analog): 0.3–1.2 m/min
- Arc start is triggered by a 24V discrete signal and voltage/current setpoints are analog references. Slowest response, but simplest to troubleshoot and compatible with any robot brand. Still common in shipyard and heavy equipment applications.
Does an integrated power source lock me into one robot brand for future expansion?
Partly. The integrated power source is physically and electrically designed for a specific robot controller family. If you later add a robot from a different manufacturer to the same cell, that new robot needs its own power source — integrated or separate. However, the welding process parameters (WPS files) are generally exportable between power source brands via standard formats, so the welding recipe itself is not locked in. For cells that will expand across robot brands over time, start with a separate power source architecture from day one to avoid the future integration puzzle.
Can I improve a separate power source's arc start by upgrading the fieldbus cycle time?
Yes, within limits. PROFINET IRT (isochronous real-time) reduces cycle time to 250 µs and adds clock synchronization between the robot controller and power source. This brings a separate power source's arc-start performance close to integrated levels — within 5% on bead consistency at the start. However, PROFINET IRT requires every device on the network to support it, including the robot controller, the power source, and the network switches. Retrofitting an existing EtherNet/IP cell to PROFINET IRT for this reason alone is rarely cost-justified unless the cell also needs the bandwidth for other real-time devices.
