CTI is a laminate property measured in volts, and the number that decides your spacing is not the laminate figure itself but the material group it falls into. Standard FR-4 at CTI 175 to 399 is material group IIIa, and at 400 V working voltage with pollution degree 2 that asks for 4.0 mm of creepage. A CTI 600 laminate in group I asks for 2.0 mm at the same voltage.
So the useful question is not what CTI the laminate carries. It is what group the finished surface behaves like, and how much of that 4.0 mm you can afford to spend. Most 400 V industrial boards pass on group IIIa and give up the extra 2 mm of keep-out. The boards that cannot spare the space are the ones that need a higher CTI material, and they are also the ones where a dirty surface costs the most.
What the CTI test actually measures
IEC 60112 places two platinum electrodes on a 20 x 20 x 3 mm specimen, drops 20 mm3 of 0.1 percent ammonium chloride solution between them 50 times at 30 second intervals, and steps the voltage from 100 V to 600 V in 25 V increments. The test fails when a current of 0.5 A or more holds for 2 seconds, or when the specimen burns. CTI is the highest voltage that survives all 50 drops without tracking.
What fails is the surface, not the bulk. The solution bridges the electrodes, the current heats a narrow path, the path dries, a spark passes, and the resin under that spark chars. Carbon conducts far better than the resin it replaced, so the next drop finds an easier path and the track widens. A laminate that passes 600 V does so because its filler system resists that carbonisation, not because it insulates better through its thickness. Dielectric strength and CTI measure two different things, and a board can hold 4,000 V through the dielectric and still track at 250 V across the surface.
Why a clean board holds far more than its CTI suggests
Tracking needs a conducting film. On a dry, clean board the leakage between two 400 V tracks is in the nanoamp range and nothing happens at any realistic spacing. Add flux residue, dust, salt fog or condensation and the same gap carries enough current to heat and char. That is why a laminate passes 600 V in a laboratory and the same material tracks at 250 V in a humid cabinet: the material did not change, the surface did.
Residue left under a coating is worse than no coating at all. The coating now holds the contamination and the moisture in one place, and the tracking path forms under a film that hides it from inspection. This is the part of the problem that shows up as a puzzle in the field, because the board looks sealed. Measurement matters more than appearance here, which is why we work to a measured ionic cleanliness limit rather than a visual pass, and why surface tracking is kept separate from conductive anodic filament, which grows through the laminate along the glass bundles instead of across the surface.
How the CTI number turns into millimetres
Creepage in IEC 60664-1 comes from three inputs: working voltage, pollution degree, and material group. At pollution degree 2, which covers most indoor and enclosed industrial equipment, the table gives 2.0 mm for group I, 2.8 mm for group II and 4.0 mm for group III at 400 V RMS. At 800 V the same columns read 4.0, 5.6 and 8.0 mm. At 250 V they read 1.25, 1.8 and 2.5 mm. Reinforced insulation doubles the creepage figure.
Two consequences follow. First, one material group can halve the keep-out, which matters on a dense control board far more than the price difference between two laminates. Second, coating the assembly to a recognised standard lets the surface be assessed at a lower pollution degree, and that is often cheaper than changing material. Neither route survives a board that was dirty before it was coated.
Where high voltage boards actually go wrong
The first mistake is buying a CTI 600 laminate for the certificate and then sealing flux residue under the mask. The certificate describes a bare specimen. The board describes the process that built it.
The second is reading CTI 175 as a board rating. Solder mask, marking ink, residue and coating are the surface that tracks. If the product standard wants group I, the whole surface system has to behave like group I, and that means residue control, cure control on the coating, and a cleanliness measurement instead of a look under the lamp.
The third is sizing creepage on the wrong voltage. Creepage follows the RMS working voltage; clearance follows the impulse voltage. A 400 V DC bus with 1.2 kV switching transients is comfortable on creepage and can still fail on clearance, and a board that passes both at sea level can fail at altitude, where clearance is corrected and creepage is not.
What we hold on a high voltage build
Laminate at CTI 175 and above as standard, in Tg 130, 150 or 170 °C, with breakdown at 4,000 V and above, UL-94 V0, and the CTI class printed on the per-lot laminate certificate. Inner layers at 2 oz and outer layers up to 10 oz where the current needs it, plating at 20 µm and above, and 0.05 mm line and space where the spacing has to be squeezed. Full parameter detail sits on FR-4 process parameters.
On the assembly side: ionic cleanliness measured on the finished board, and selective coating to IPC-CC-830 with the cure verified rather than assumed. What we do not do is run a CTI test on a finished board. CTI is a laminate test that belongs to an accredited laboratory and to the material supplier. Our part is to keep the surface clean enough, and the coating intact enough, that the laminate number is not wasted, and to pass the supplier certificate through with the lot data.
If your product standard asks for a group I surface and the board cannot give up 4 mm, the honest answer is that the stack changes, not the cleanliness alone. We would rather say that at quotation than after the first batch fails a hipot and tracking test.
Next step
Send the stack, the RMS working voltage, the pollution degree your enclosure implies, and the material group your product standard asks for. We will tell you which group the stack meets, whether the spacing fits, and where the constraint actually sits before a panel is ordered. Assembly limits are on turnkey PCBA assembly.