Sky-scanning optical sensor

You'll get a large ground telescope: a 1.55 m primary mirror in a carbon-fibre cell, a carbon truss tube with a secondary mirror on a spider, a camera, and a fork mount with azimuth and elevation drives and an encoder. It is the biggest of the four examples: a first build takes about 5 minutes.

1. 1Step 1: Paste the prompt, then give the brief

   Start a new chat, paste the Speed prompt and send it. When Claude asks What do you want to build?, type:

   Type thisCopy

   SkyScan, a ground sensor that scans the sky with a 1.55 m primary mirror.

   Then switch to your Speed tab to watch.

2. 2Step 2: Read the plan

   The mirror size is stated. Claude assumes the rest, for example the mirror material, the focal length (which sets the tube length), a stiffness target for the structure (such as a first natural frequency of 10 Hz), and the load cases. Speed has a ready-made 1.5 m Zerodur mirror, so Claude starts from it and scales the primary to 1.55 m.

3. 3Step 3: Watch it build, stage by stage

   1. Start. Claude loads Speed's 1.5 m Zerodur mirror and resizes the primary to 1.55 m.
   2. Mirror cell. A carbon-fibre (CFRP) cell that supports the primary from behind.
   3. Tube. An open carbon truss, with the secondary mirror held on a spider at the top, and a camera.
   4. Mount. A fork mount with an azimuth drive, an elevation drive and an encoder. No bearing in Speed's catalogue is large enough for these axes, so Claude models placeholders and says so.
   5. Check. Claude runs the physics check, takes a picture, then runs FEA on a few named load cases, such as gravity with the tube at different elevations and a modal case.

4. 4Step 4: Read the summary

   Expect an honest answer like "strong but floppy". On the live run, all three load cases passed on strength, but the first natural frequency came out far below the 10 Hz target. Claude said so and recommended diagonal bracing. You'll also get the total mass, the moving mass, every assumption, and the note that the bearings are placeholders.

Follow-ups to try#

Add the bracing and rerun the FEA#

Add the diagonal bracing and rerun the FEA. Show me the first mode before and after.

Claude adds the bracing, runs the same load cases again and reports the new first natural frequency against the target, plus what the bracing added in mass. If the target still isn't met, it says what it would change next.

Compare a stiffer tube#

Try a stiffer tube design and compare it with this one in a table: mass, first mode and lowest safety factor.

Claude changes one thing, reruns the checks, and shows the two side by side.

Make the one-pager#

Make the one-pager.

A one-page drawing sheet with views, a parts list and notes opens in your Speed tab.

Get the FE report#

Make the FE report with every load and support you assumed listed.

Claude solves every load case and gives you a download link to a PDF report. The link lasts one hour. On a design this size it can take several minutes; see the note below.

What to check before you trust it#

• FEA here is screening grade on the parts' shapes. It shows the structure's strength, sag and first resonance. It does not predict the mirror's surface figure, which needs a dedicated optical-mechanical analysis.
• The large bearings are placeholders. Size them with a bearing maker before you trust the mount's stiffness.
• A large FE report can take longer than a chat allows for one step (280 seconds). If the chat says a step timed out, wait for the Agent panel in Speed to finish, then ask "Show me the FEA result". See Troubleshooting.
• An engineer must review the design before anything is built.

Something unclear or wrong? Tell us at support@tryspeed.ai.