Can you count three light bands on a carbon seal face and two on a silicon carbide face — and tell me which one actually needs rework? If you hesitated, you are not alone. Most technicians I train can learn to count bands in five minutes. The part that takes experience is knowing what that count means for the seal you are holding.
The reading technique is straightforward physics. The decision — accept, reject, or relap — depends on your seal material, service pressure, and face diameter.
What One Light Band Actually Measures
One helium light band equals 0.0000116 inches — that is 0.3 microns or 11.6 microinches. This number is one half-wavelength of helium light, and it is the resolution unit for optical flat measurement.
When you place an optical flat on a seal face, light reflects off both the flat’s bottom surface and the seal face. Where these two reflected beams meet, they interfere. A dark fringe appears wherever the gap between the surfaces equals exactly one half-wavelength. Each additional dark band adds one more half-wavelength of separation.
Three dark bands across a seal face means the maximum gap between the optical flat and the face is 3 x 11.6 = 34.8 microinches, or roughly 0.9 microns. That is your flatness deviation — the actual out-of-flat measurement.
If your facility uses sodium light instead of helium, one band equals 0.29 microns (0.0000114 inches). The difference is negligible for seal work. Use whichever source you have and stay consistent.
How to Count Bands Correctly
Count the dark bands, not the light spaces between them. This convention is universal in metrology, and mixing it up doubles your reading.
The Imaginary Baseline Method
Tilt the optical flat until at least one fringe line runs from one edge of the seal face to the opposite edge. That fringe is your reference. Now draw an imaginary straight line connecting where that reference fringe enters the surface on one side to where it exits on the other. Count how many other dark fringe lines your imaginary line crosses or touches. That count is your flatness in light bands.
Think of the fringe pattern like topographic contour lines on a map. Straight, parallel, evenly spaced lines mean flat terrain. Curved lines mean hills or valleys. The tighter the curves, the steeper the deviation.
Always Check Two Directions
Here is a step most technicians skip: rotate your viewing angle 90 degrees and read again. A seal face can look perfectly flat in one direction but show two or three bands of cylindrical warp from the perpendicular angle. Your flatness reading is always the worst of the two measurements, not the average.
I have seen faces pass a single-direction check and leak within days of installation. The warp was invisible from the first angle. Make the 90-degree rotation a habit — it takes ten seconds and catches problems that single-angle readings miss entirely.
How to Identify Concave and Convex Faces
Band count tells you how far the surface deviates from flat. It does not tell you which direction. A three-band concave face and a three-band convex face need completely different corrective lapping, so you must determine the geometry before deciding on next steps.
The Finger Pressure Test
With the optical flat resting on the seal face, press one edge lightly with your fingertip and watch how the fringe pattern responds.
- Fringes move toward your finger: The seal face is convex (high center, edges slope away).
- Fringes move away from your finger: The seal face is concave (low center, edges are high).
This is the only reliable field method for distinguishing surface geometry on seal rings. Memorize the mnemonic: toward means convex, away means concave.
Band Width Is Irrelevant
The most common mistake I see with technicians new to optical flat work is fixating on how wide the bands are. Band width changes the moment you tilt the optical flat — it is purely an artifact of the angle between the two surfaces. Wide bands and narrow bands can both indicate the same flatness.
What matters is band straightness and parallelism. Straight, parallel bands at any width indicate a flat surface. Curved bands indicate deviation regardless of width. Stop looking at how fat the stripes are. Start looking at whether they curve.
Material-Specific Pass/Fail Criteria
Two dark bands on a carbon graphite face is a pass. Two dark bands on a silicon carbide face may require rework. The same number, opposite decisions — because the materials have different tolerance standards.
Standard Tolerances by Material
| Seal Face Material | Acceptable Flatness |
|---|---|
| Carbon graphite / GFT | 2-3 light bands |
| Tungsten carbide | 1-2 light bands |
| Silicon carbide | 1-2 light bands |
| Ceramic (Al2O3) | 1-2 light bands |
Hard face materials — tungsten carbide, silicon carbide, ceramic — demand tighter flatness because they do not conform under load. Carbon graphite is more forgiving because it deforms slightly against its mating hard face during operation.
Service Condition Modifiers
Material alone does not set the tolerance. Two modifiers override the table above.
Pressure: For applications above 40 bar (580 psi), every material must be lapped within one light band regardless of face material. High-pressure service amplifies any flatness deviation into a leak path.
Diameter: Small seal faces under four inches should hold two light bands or fewer. Larger faces above four inches may read up to five bands and still perform, because the absolute deviation per unit of sealing width is proportionally smaller.
The Carbon Relaxation Trap
Before you start rejecting carbon graphite faces at incoming inspection, know this: carbon faces relax after lapping. A manufacturer laps the face to less than one light band, ships it, and by the time you check it in your QC room, you see three bands. That face is not defective.
Carbon graphite is a porous material that relieves internal stress after the lapping pressure is removed. The face “springs back” slightly. Those three bands will disappear once the face is installed against a flat hard mating ring under spring load. I have watched technicians reject entire shipments of perfectly good carbon faces because they did not understand this phenomenon. If you see three bands or fewer on a carbon face at incoming inspection, it is almost certainly normal relaxation — not a manufacturing defect.
The 3-Step Read: Pattern, Count, Decide
Every seal face inspection follows the same three steps. Skipping any one of them leads to bad judgments.
Step 1: Read the Pattern
Clean both surfaces thoroughly — one fingerprint invalidates the entire reading. Place the optical flat and observe the fringe pattern. Ask three questions: Are the bands straight or curved? Are they parallel or converging? Are they consistent across both perpendicular viewing angles?
Straight parallel bands in both directions means the face is flat. Proceed to counting. Curved bands mean deviation exists — count first, then run the finger pressure test to determine concave versus convex.
Irregular, broken, or non-uniform fringes across the face indicate localized damage — thermal distortion, pitting, or scratches. These faces typically need relapping, not just a recount.
Step 2: Count the Bands
Use the imaginary baseline method. Count dark fringes only. Take readings from two perpendicular angles. Record the higher number.
Step 3: Decide
Match your band count against the material and service conditions.
A two-band reading on a tungsten carbide face going into standard-pressure service passes. That same two-band reading on the same face going into 50-bar service fails — it needs to be within one band.
A three-band reading on a carbon graphite face at incoming inspection is almost certainly normal relaxation. A three-band reading on a silicon carbide face at incoming inspection needs rework.
This decision step is where most QA mistakes happen. I have seen technicians pass marginal SiC faces and reject perfectly good carbon faces in the same inspection session — because they applied one tolerance to every material. The band count was accurate both times. The judgment was wrong.
If a reading is borderline — right at the tolerance limit — re-clean both surfaces, re-seat the optical flat, and read again. Contamination and improper seating are the two most common causes of artificially high readings. A borderline face that re-reads one band lower after proper cleaning was never borderline to begin with.
Next Steps
The band count is data. The material-and-condition lookup is knowledge. The judgment call on a borderline reading at two in the morning, three seals into a turnaround — that is skill, and it only comes from repetitions.
Start building that skill by reading every seal face that crosses your bench, not just the ones that look questionable. Record the material, band count, and your pass/fail decision. Within a few dozen readings, you will stop reaching for the tolerance chart and start making decisions from memory.
The fastest way to get confident: find an experienced technician who reads light bands routinely and watch them work through five or six faces. Pay attention to what they look at first, how they position the flat, and how quickly they make the accept/reject call. That speed comes from pattern recognition, not from counting faster.
