Probably as many variables when screw cutting – Worn leadscrew, grind of the cutting tool & tip, full or partial insert, work deflection, depth of cut, operator. exact pitch or near enough with small %age error, etc

I suspect that threads produced with a Coventry or similar die-head are likely to be more accurate than anything screw-cut on a lathe of unknown history.

It is said that the quality of long die-cut thread suffers due to simultaneous variation in cutting forces and the force needed to pul the die along the three. With a spring loaded die head, like the Coventry system, thread cutting and traverse forces are essentially constant.

Feeding via the feed screw rather than by hand is likely to be best. Which is what I usually do if its a thread on the QC box. It helps that my S&B 1024 is metric and my P&W B is imperial so most things can be easily covered. Except 19 tpi.

Nailing down specifications is pretty difficult but Wikipedia AI says :-

The American Standards for Accuracy of Engine and Toolroom Lathes (ASA) says typical lead errors fall into the following ranges: [1]

1. Cumulative Lead Error (Linear Accuracy)
This measures the cumulative error over a specific length of the screw, usually measured per foot (300 mm): [1, 2]
Toolroom Lathes (High Precision): Maximum allowable error of ±0.0010 inches per foot (≈ 0.025 mm per 300 mm).
Engine Lathes (Standard Industrial): Maximum allowable error of ±0.0015 inches per foot (≈ 0.038 mm per 300 mm).
Hobbyist / Inexpensive Lathes: Typically range from ±0.002 to ±0.005 inches per foot (≈ 0.05 to 0.12 mm per 300 mm), as these leadscrews are typically thread-rolled rather than precision-ground.[1, 2]

2. Short-Distance / Incremental Variance
This measures local pitch errors over a short span to catch quick fluctuations or “drunken” threads: [1, 2]
Toolroom Lathes: Max variance of ±0.0004 inches in any 4-inch span (≈ 0.01 mm per 100 mm).
Engine Lathes: Max variance of ±0.0005 inches in any 4-inch span (≈ 0.012 mm per 100 mm). [1]

3. Backlash and Wear Realities
While pitch accuracy dictates the precision of the thread form while cutting in a single direction, overall operational accuracy is highly dependent on backlash: [1, 2]
New Acme/Trapezoidal Setup: Typically exhibits 0.05 mm to 0.15 mm of backlash between the leadscrew and the half-nut.
Worn/Used Setup: Easily degrades to 0.20 mm to 0.40 mm of backlash over time. [1]

Working out exactly what effects such errors have on thread properties is a nice exercise for the reader.

So lead accuracy is probably not a great worry on a lathe in good condition.

Form errors due to imprecise grinding or tool setting up seem the most likely reason for noticeable error in a screw-cut thread.

Clive

I often make threads on the lathe. Most of them are short (3 times the diameter of less). I use the CNC lathe (G33) for this because it is “accurate” and easy. By accurate I mean they fit my test nut or the counter part it is used for.

For a very tight fit (double nut to preload bearings) I use the CNC lathe where the spindle is driven by a stepper (C-axis). That assures the Z-axis and spindle (C-axis) are always in sync. Then the counter part is also used to get a tight fit.

I can’t make “high accurate threads” because I do not have the tools to check the thread made. The 3 wire method I use sometimes can only measure outer threads.

I do have non adjustable threading dies but these are used very rarely. When used, the nut always fits so “accurate” in my book.

On the CNC router I mill threads and check these with the counter part.

Don’t forget, I do this for a hobby.