A few weeks ago we set up a camera next to a four-axis job: an aluminum structural wall section, held on a rotary table, machined in one setup. Twenty-two seconds of footage, nothing staged.
It is the least dramatic machining video we have. No shower of chips, no time-lapse of a complete part appearing. The cutter makes long, quiet climbs down a wall that looks already finished. That is the point, so this post is really about what does not happen on camera.
The finished clip is embedded at the bottom, together with the page where we keep the process notes.
The failure mode nobody films
A 0.5 mm wall in 6061 does not move while the tool is in it. It moves afterwards, when the vise opens and the stress that roughing put into the material has somewhere to go.
The pattern on parts we receive for rescue work is boringly consistent:
- finished in one setup, clamped hard
- measured flat on the machine, all good
- bowed 0.02 to 0.05 mm by the next morning
Nobody lied on the inspection sheet. The part simply had not finished moving yet. If your drawing shows a 0.4 mm wall with a 0.01 mm flatness callout, the process is the product - the geometry is decided by setup and sequence long before the finishing pass.
What we actually change for thin walls
Three things, in this order of importance.
Stock left on the wall, and time. Roughing leaves 0.15 to 0.25 mm on the wall, then the part rests. Overnight, not twenty minutes. Semi-finishing takes half of that away, another rest, and only then the finishing pass. It costs a day of calendar time and saves a scrap bin.
Workholding. This matters more than feeds and speeds on any wall under about 1.5 mm. We bore soft jaws to the part profile so contact is distributed; we switch to an expanding mandrel when the inside diameter is the datum; and on walls too thin to survive a through-slotted jaw we go to vacuum or adhesive fixturing. Finishing passes run climb-milling with light radial engagement, so the wall is pushed into its support instead of pulled away from it.
Sequence. Anything that removes support early - a window, a slot, a cross-hole - gets scheduled last, or interrupted so the wall keeps a bridge until the end. On the section in the video, the profile is roughed with the wall still tied to the parent stock, and the tie is cut in the finishing sequence.
The numbers, because they are what gets quoted
Material limits differ enough that they belong in a table rather than a paragraph:
| Material | Wall we machine routinely | Note |
|---|---|---|
| 6061-T6 / 7075 aluminum | 0.3 mm | 0.25 mm possible on short features |
| 304 / 316L stainless | 0.5 mm | work-hardens - each pass must cut, not rub |
| Ti-6Al-4V | 0.6 mm and above | springs more than aluminum at equal thickness |
| PEEK and engineering plastics | part-dependent | thermal growth dominates cutting force |
Tolerance commitment on diameters and positions is plus or minus 0.005 mm. Flatness on a 0.5 mm wall typically lands at 0.01 to 0.02 mm depending on span, and tighter figures get quoted per drawing rather than promised in a blog post.
We are a shop with 220 CNC machines in Shenzhen, so the honest version is: these are the numbers we will put in a quote, not the ones we hit once on a good day.
How you measure a wall that flexes
A CMM probe presses on the part. On the thinnest walls that pressure is a measurable fraction of what holds the wall straight, so the act of measuring changes the reading.
On those features we measure with an air gauge or an optical comparator and record three orientations per feature. The inspection sheet notes the method next to the number, so a buyer auditing the report six months later can see exactly how each figure was taken. That line - "how was this measured" - is the one most inspection reports leave out, and it is the one that decides whether two reports are comparable.
What to send with an RFQ for thin-wall work
Four things, and the fourth is the one that gets skipped:
- The drawing, with wall thickness and flatness called out explicitly - not implied by a general tolerance block
- Material and temper (6061-T6 and 6061-T651 machine differently near the limit)
- Quantity, because a 5-piece prototype and a 5,000-piece run need different process discipline
- The datum scheme you will inspect against
We quote prototypes in 3 to 7 days, production runs in 7 to 15 days, with MOQ from 1 piece and stepped pricing. Shipping is EXW Shenzhen only - we are happy to recommend reliable freight forwarders, but the export leg is yours to book.
Watch the footage
The clip is a four-axis rotary setup cutting an aluminum structural wall section in our Shenzhen facility. Same fixture discipline, same process control as production thin-wall parts - the geometry is held by setup, toolpath and sequencing from the first cut.
- Video, with the process notes: https://chxmetal.com/en/thin-wall-cnc-machining/
- Short version: https://www.youtube.com/shorts/mx2yv4Zx-2s
- Also on Vimeo: https://vimeo.com/1245683370
If you have a thin-wall part that bowed after machining, the drawing plus one of the bowed parts tells us whether rework beats remaking. We will say so either way.
Top comments (0)