Why Your 3/4-Inch Plywood Is Really 23/32: Tolerances and Three Dado Fixes, Priced
Why Your 3/4-Inch Plywood Is Really 23/32: Tolerances and Three Dado Fixes, Priced - A step into the workshop.
Stop Wasting Wood
Generate clear cut diagrams instantly with our new Cutlist Optimizer. Plan your cuts to make your next build easier.
Yo, check it. You cut the dado, you slide the shelf in, and the shelf just goes. No resistance, no squeak. It swims. You wiggle it and watch a shadow line open at the shoulder that no amount of glue is closing.
I work out of the second bedroom. My landlord thinks it's for guests. There's no room in here for a mistake I have to carry back down three flights to the curb, and a 30" x 30" panel of 3/4 Baltic birch runs $45.99 at Rockler, so a failed joint comes out of rent money. Which is why I went and read the actual standards on this one. The fraction stamped on that sheet never promised me anything.
The Fraction Is a Category Name, Not a Measurement
Start with hardwood plywood. ANSI/HPVA HP-1-2024, the Decorative Hardwoods Association standard, says in section 3.13 that the actual dimensions of a panel "shall be as agreed upon between the buyer and the seller." There's no fixed thickness anywhere in the document. And when a mill labels a panel, the standard shouts in its own capital letters: "The Thickness Class designation shall NOT be expressed in terms of actual units of measurement."
Read that twice. The 3/4 is a class name. The only hard number HP-1 attaches is a floor: a class of 1/4 or bigger may run minus 3/64 inch. So a panel legally sold as 3/4 Class can measure 0.703". That clause sets no ceiling at all.
Structural plywood plays the same game with different words. NIST's Voluntary Product Standard PS 1-19 calls the fraction a Performance Category, "a panel designation related to the panel thickness range." APA, the Engineered Wood Association, says it flat out on its trademark page: Performance Category is not an actual thickness measurement.
Now the part that fixes the whole premise. What the store calls "3/4 plywood" usually isn't 3/4 category at all. It's 23/32. PS 1-19 Table 10 puts a sanded 23/32 panel between 0.703" and 0.734", so against a 0.750" dado you're looking at a gap of 0.016" on a good day and 0.047" on a bad one. Meanwhile a genuine 3/4 PERF CAT sanded panel runs 0.734" to 0.766", which straddles 0.750". A real 3/4 sheet can be too fat for your 3/4 groove.
The spread inside one legal category is 0.031". A full thirty-second. Hang onto that, because it's the reason no fixed cutter width can ever be right. The target moves by 1/32" and the sheet still passes.
The Number You Want Is Already Printed on the Panel
Free tip. Costs you nothing but bending over in the aisle. PS 1-19 section 7.3(e) requires the panel to carry "the thickness in 1000ths of an inch" in addition to the Performance Category. Dunn Lumber's walkthrough of a real grade stamp shows what that looks like in the yard: "THICKNESS 0.703 IN" sitting right next to a "23/32 CATEGORY" mark.
Don't take that decimal as gospel about the sheet in your cart, though. PS 1-19's Appendix D says the recommended labels are "based on the minimum permissible thickness for that category," and APA says the declared decimal is generally at or near the lower tolerance. It's a floor. You can still be carrying 0.031" of category above it.
The stamp tells you which category you bought. A caliper tells you what you got. Two different jobs.
While we're down there: PS 1-19 section 5.10 bases panel measurements on 9% moisture content, and HP-1 lets a mill ship at up to 12%. The legal thickness is defined at one specific moisture. A sheet is not a fixed object.
Baltic Birch Is Metric, and the Tolerance Is Wider
18 mm is 0.7087". That's a 0.041" gap in a 3/4" dado before anybody's tolerance shows up.
Wolstenholme's published spec sheet for Baltic birch gives sanded-panel tolerances that widen with thickness, and 18 mm gets +0.7/−0.9 mm. Range: 17.1 to 18.7 mm. In our money, 0.673" to 0.736". That's 1.6 mm of legal variation on one nominal thickness, essentially a full 1/16". Worst case in a 3/4" dado, 0.077" of daylight.
Rockler sells those panels labeled in inches even though they're manufactured in millimetres, and their own listing carries the disclaimer: all dimensions are nominal and may be slightly undersized. Believe them.
MDF Flips It on You
MDF is the one sheet good actually made to the number on the label, and that's where people get bitten going the other direction.
Weyerhaeuser's specifications for standard MDF list a thickness tolerance of 0.005" and stock 0.7500" as a real thickness, not a category name. Uniboard's technical sheet for MDF to ANSI A208.2-2016 Grade 130 spells it out further: plus or minus 0.125 mm (0.005") from the specified thickness, and plus or minus 0.125 mm from the panel's own average. Held to uniformity within the sheet, not just to the target.
So 3/4" MDF is legally 0.745" to 0.755". Against a 0.750" dado, half the population is a hair loose and half of it will not go in at all. Same fraction on the label, opposite failure.
How Tight Is Right
A dado cut to exactly the board thickness is a failed dado. Nowhere for the glue to live. WoodWeb's knowledge base has pros converging on machining dados .005" to .010" wider than the actual part thickness, and I like their field test better than any number: the joint should be just snug enough that you can drive the boards together with your hat. If it needs hammer blows, you squeezed the glue line out.
Best trick in this whole piece comes from Woodcraft's caliper guide. Zero your digital calipers inside the cut dado, then measure the mating piece. The display reads exactly how undersized the groove is. Negative means it's fat, pull a shim. Woodcraft also reports actual 3/4 plywood running anywhere from 45/64" to 3/4", and notes thickness varies within a single sheet, so measure around the edge of the panel, not once and done.
Three Fixes, Priced Out
Undersized bits. Whiteside's 470 set (7/32", 31/64", 23/32", all 1/2" shank) is $71.99 at Woodcraft. Amana's 45445, the 23/32" on a 1/2" shank by itself, is $28.84 on sale at dealer JMP Wood, $38.45 regular. Here's the trap nobody prints: a 23/32" bit cuts 0.71875". Against a maximum-thickness 23/32 sanded panel at 0.734", that groove is 0.015" too narrow. Add glue clearance and you're 0.020" short. The board doesn't go in. Against 18 mm Baltic at its 18.7 mm max, you're 0.017" small. Rockler concedes it in their own explainer: these bits get you close, but no sheet is guaranteed any exact thickness. One fixed guess that can fail tight as easily as loose.
Dado stack. Freud's SD508 Super Dado is $226.97 at Rockler, cutting 1/4" to 29/32". The SD500 series ships shims of 0.020", 0.012", 0.008" and 0.004", so your finest move is 0.004", comfortably inside WoodWeb's window. The SD608 Dial-A-Width is $313.47 and gets the same 0.004" per click of the dial with no shims at all. Genuinely adaptive. But Freud's own guide tells you to check your saw's manual for the maximum allowable dado width, and plenty of us don't have that saw. I don't. Every new sheet also means breaking the stack down and re-shimming it.
The two-pass jig. Two parallel legs, router rides between them, bearing-guided pattern bit. Per Woodworker's Journal, the bit has to be smaller in diameter than the material so you can cut in two passes, bearing riding one leg and then the other. Groove width equals leg spacing, which means the bit diameter drops out of the math entirely. You set that spacing by sandwiching an offcut of the actual shelf stock between the legs. Their spec is a 5/8" pattern bit and an adjustable leg using 1" holes over 1/4" carriage bolts. Woodcraft lists Whiteside's bearing template bit line at $30.99.
| Fix | Cost | Sizes to your actual sheet? |
|---|---|---|
| Shop-made two-pass jig + pattern bit | ~$31 plus scrap | Yes, and no measuring |
| Amana 45445 undersized bit | $28.84 sale / $38.45 reg | No, one fixed size |
| Whiteside 470 three-bit set | $71.99 | No, three fixed sizes |
| Infinity Precision Router Dado Jig | $129.90 | Partly, 1/16" index |
| Freud SD508 + shims | $226.97 | Yes, 0.004" steps |
| Freud SD608 Dial-A-Width | $313.47 | Yes, 0.004" per click |
| Woodpeckers Exact Width Dado Jig | $439.99 (reg $479.99) | Yes, no measuring |
| Caliper, prerequisite for bit and stack routes | $27.99 Pittsburgh / $35.99 iGaging | — |
The commercial jigs prove the principle costs money, not genius. Infinity's Precision Router Dado Jig is $129.90 and indexes in 1/16" increments up to 2" wide, which is coarser than the fit window, so the fine sizing comes from repositioning the slide and taking another pass. Woodpeckers' Exact Width Dado Jig is $439.99, and their pitch is the part worth stealing: you fit your stock snugly between the edge guides, so the material is the gauge and nobody measures anything. Woodpeckers will also tell you straight that sheet goods "don't come close to their nominal dimensions these days" and vary by brand, sheet type and country of origin.
Build the Jig
The shop-made version is the only sub-$50 fix on that list that adapts to the sheet in front of you, and the only one that removes measurement error instead of managing it. The undersized bit costs about what the pattern bit costs and still can't handle a fat panel. If you're cutting a hundred identical carcase parts off one lift of plywood and your saw takes a stack, buy the stack. That's a production tool and it's the fastest one here.
For the rest of us, working one sheet at a time in a room that's supposed to have a bed in it: cut two legs out of the offcut you already have, clamp a piece of the actual shelf stock between them, and let the board size its own groove. Then next time you're standing in the aisle, bend down and read the decimal on the stamp before you pay. Just remember what it is. A floor, not a promise.