The figure in most drawings is 1.56 micrograms of sodium chloride equivalent per square centimetre, and it is a legacy number. IPC now treats it as a process control value rather than a product acceptance limit: J-STD-001 revision H dropped the single ROSE pass or fail and moved acceptance to surface insulation resistance, keeping a fast ionic contamination test for daily control. So keep 1.56 on the drawing if a customer asks for it, report to it honestly, and do not believe a board is clean because it passed.
What 1.56 actually measures
The test is ROSE, resistivity of solvent extract, IPC-TM-650 method 2.3.25. The board is rinsed or submerged in 75 percent isopropanol and 25 percent deionised water, the conductivity of the extract is measured, and the result is converted to micrograms of sodium chloride equivalent per square centimetre of extracted surface. The original extraction was 10 millilitres per square inch, which is where the 10 micrograms per square inch, or 1.56 micrograms per square centimetre, comes from.
Two limits are built into that method. It sums every ionic species into one conductivity number and identifies none of them, which is why method 2.3.28 exists for ion chromatography. And it cannot see non ionic residue at all, which is most of what a modern no-clean flux leaves behind.
The number itself was set in the 1970s for 35 percent solids rosin fluxes and ozone depleting cleaning solvents. Published accounts of the standard describe the acceptance value as more or less arbitrarily chosen at the time, then retained because high reliability programmes had shipped against it for years. The chemistries it was calibrated on are gone. IPC-TR-583 also found that automated ROSE testers of that era lacked repeatability and reproducibility, and that the equivalency factors used to compare instruments were invalid. A 1.2 reading on one machine is not the same board as a 1.2 reading on another.
Why the number does not predict failure
Ionic failure is electrochemical, not a threshold. It needs mobile ions, a moisture film, a bias and a narrow enough gap for the two to meet. Take any one away and the residue is harmless. That is why two boards reading 1.2 micrograms per square centimetre can carry very different risk: one may hold chloride from a plating bath, the other a weak organic acid from the flux, and the acid is not ionic until it absorbs water.
No-clean is the sharp case. Those residues are deliberately left in place, they pass ROSE without effort because they are not yet conductive, and then the product sits in a humid enclosure at 60 degrees with 30 volts across a 0.15 mm gap. The acid takes up water, the surface resistance falls, and the failure reads as dendritic growth between two pads under a coating that was applied over the residue in good faith.
Density makes the same residue more dangerous than it was in 1975. A board with 0402 passives at 0.2 mm gaps has far more sensitive sites per square centimetre than a through hole board with 2.5 mm spacing, so the amount of residue that used to be irrelevant now decides whether the product ships. IPC's own position paper makes this point: one numeric limit cannot cover assemblies whose residue chemistry and spacing have both changed.
What to specify instead
Process control: a fast ionic contamination test on the line every shift, so a drift in the cleaner or the paste batch shows up in hours rather than at the customer. Process ionic contamination testing per IEC 61189-5-504 takes minutes and is repeatable, which ROSE never was across instruments.
Product acceptance: surface insulation resistance under bias, at the temperature and humidity the product will actually see. The acceptance most customers write is 10 to the eighth ohms with no more than one decade of fall from the first reading over the exposure. It is a slow test, and it is the only one that answers the question the drawing is really asking.
Species and level: ion chromatography per IPC-TM-650 2.3.28, with limits per ion rather than one total. Chloride and sulfate are the ones to watch after plating and after aqueous cleaning; bromide points at a flame retardant, and sodium at handling.
Three ways it goes wrong
Reading a ROSE number off the wrong area. The result is per square centimetre of extracted surface, and that area has to be the real one: both sides, plus holes, plus the fact that a panel is not a board. Enter the tank default instead of the measured area and the number is arithmetic, not cleanliness. The same board can pass or fail depending on one field in the software.
Cleaning to pass the test instead of cleaning the board. A deionised water rinse moves ions around and dilutes the extract without removing what sits under a component. Bath conductivity has to be logged and the rinse has to be renewed on conductivity, not on the calendar.
Coating over contamination. A conformal coating is a moisture barrier, not a cleaning step. Over a clean surface it raises the humidity threshold; over a flux residue it traps the acid against the board and gives condensation somewhere to sit. Coat after cleaning and after the bake, never instead of them.
What SUPERB PCBA controls
Aqueous cleaning after assembly for water soluble paste and for no-clean builds where the drawing calls for it, with bath conductivity logged per shift, a 125 degree bake to J-STD-033 before any coating, and selective coating on the Nordson line with cure verified. Where the drawing carries a number, the cleanliness sample goes to an accredited laboratory and the report states the extraction area next to the result, because a value without its area is not a measurement.
The honest boundary: we do not publish a cleanliness figure that a customer cannot audit. The process is controlled on our floor, the acceptance figure comes from a laboratory report on the agreed sample, and the SIR test that decides reliability belongs in the customer's qualification plan, on the customer's test vehicle.
What to send with the drawing
Which standard applies, whether the ionic figure is a hard acceptance or a process reference, the residue chemistry of the paste, the coating type, the spacing and voltage the product runs at, and the humidity it will see. With those five items the test plan is one page instead of an argument after the first batch.