SESSION 5 · 09/26 · DIDN'T WORK
A new tube, two nights, and not one frame of sky
Got the 76/700 on the mount and pointed it at nothing successfully. Cloud, a finder I never aligned, and a field of view the size of a keyhole. The morning's daylight testing is where the actual answers came from — including that the Barlow I've been recommending to myself is doing nothing on this tube.
The tube
A 76/700 replaced the 76/350. Same aperture, so the same hard ceiling — Dawes 1.53", about 2.8 km on the Moon. Nothing about that changes.
What changes is the focal ratio, and it matters more than I expected. A spherical
mirror only stays diffraction-limited above about 4.5 x D_inches^(1/3), which for a 3"
mirror is f/6.5. The old tube was f/4.6 — under the line, so it probably never
reached its own theoretical limit. The new one is f/9.2, comfortably over. Same
ceiling, but this one should actually get there.
Two dividends I didn't know to expect: collimation tolerance is roughly 4x looser at f/9.2, and at 0.41"/px the 1.53" limit spans about 3.6 px, which is correct sampling rather than the under-sampling I had before.
The night: nothing
The Moon wasn't available at all — it rose at 02:56 and never got above 17 degrees before dawn. Saturn was the target, and it was the best-placed thing I've ever had: transit at 02:09 at 51.7 degrees, airmass 1.22. Compare that with the 30 August session at 7 degrees and airmass 8.1.
Then:
- 70% cloud at 20:00, low cloud in a 16 km/h breeze. It thinned to 19% by 02:00, which I never got to use.
- Saturn's rings are edge-on. Tilt 0.15 degrees — a ring-plane crossing. So even with clear sky it was going to be a 45-pixel featureless oval, not the postcard. Worth knowing before blaming the scope.
- Venus was a 36% crescent at 32.5" and set at 21:30, under the worst of the cloud.
- I tried Alpha Centauri at 38.5 degrees in the southwest and could not find it. At 700 mm the field is 23 arcmin. Hunting in that without an aligned finder is not a technique problem, it's a geometry problem.
The one measurement worth keeping: a snapshot of a streetlight showed a 212 px edge rise, which works out to 2.83 mm of defocus. Small, and now I know which number to chase.
Also binned an 8.1 GB SER — 30 seconds of an out-of-focus pole. Full-frame RGB at 21 MB a frame adds up terrifyingly fast. Saturn is 45 px across; there is no reason to record 3264 of them.
The morning: where the answers were
Daylight, a road sign at 80 m, and finally some numbers.
Shot with the 76/700 plus the 2x Barlow — f/18.4, 1400 mm. Edge rise 25 px.
My first instinct was "still soft". That was wrong, and the correction is the useful part: at f/18.4 diffraction alone cannot put an edge in less than about 9.7 px. So 25 px is only 2.6x the floor. In angle that's 5.2 arcsec of blur, against a 1.53" limit — and across 80 m of daylight air, that's heat shimmer, not the telescope.
| config | f/ | arcsec/px | diffraction floor |
|---|---|---|---|
| 76/350 bare | f/4.6 | 0.825 | 2.4 px |
| 76/700 bare | f/9.2 | 0.413 | 4.9 px |
| 76/350 + 2x | f/9.2 | 0.413 | 4.9 px |
| 76/700 + 2x | f/18.4 | 0.206 | 9.7 px |
Look at the last row. The Barlow on this tube spreads the same detail over 2.5x more pixels and adds nothing. My pixels are about 1.4 um, which wants roughly f/5–f/7 for critical sampling. f/9.2 is already at the top of that range. f/18.4 is past the point where extra magnification can carry information.
That is a direct reversal of what I wrote on 30 August, when the Barlow genuinely was the biggest single improvement available. It was — on the f/4.6 tube, which was under-sampled at 1.8 px per resolution element. The advice was right and the hardware moved out from under it.
Three things that were wasting my time
I lost nine runs to a changed output path. The captures I thought had silently failed
were on disk the whole time — 28 GB of them, in ~/Pictures/2026-09-06/ under the
object name testing, because the path had been changed in the GUI and I was searching
AstroDMx_DATA. See the output path is app state.
I had a tidy theory ready — the camera came back on /dev/video5 after the tube swap, so I
blamed a stale device handle — and it was wrong. The app never missed a frame.
The floor. Testing indoors on a suspended wooden floor: my own weight deflects it and tilts the tripod. At 23 arcmin of field, a deflection I can't feel walks the target out of frame. Where I stand changes the alignment.
Near targets eat focuser travel. This one is genuinely counter-intuitive and it nearly had me condemn the tube:
| target distance | extra focuser travel vs a star |
|---|---|
| 30 m | 16.7 mm |
| 50 m | 9.9 mm |
| 100 m | 4.9 mm |
| infinity | 0 |
I was trying to focus on a streetlight at maybe 50 m, which needs 10 mm more outward travel than a star does. If I'd run out of travel I would have concluded the camera can't reach focus on this tube — when Saturn, needing 10 mm less, would have been fine.
Smaller things found
- Auto white balance is being re-asserted by the app. The capture log says
WB = AUTOeven though it was set to manual underneath. AstroDMx's own buttons are the ones that count. - The observing site isn't being logged. The capture log has
Error: bad lexical castwhere Goolwa should be. The location was written straight into the app's database on 1 September and the format is wrong. Needs entering through the Geographical Location dialog instead. - Filenames are stamped in Sydney time, half an hour ahead of Goolwa. The frame named
21-12-41was taken at 20:42 local. That matters the moment I try to line a capture up against where something actually was in the sky. - Daylight frames are badly veiled — a dark sign never gets below 129 out of 255. Stray light down an open tube. A cardboard hood extending past the front would fix it, and would do the same favour near the Moon's bright limb.
What the morning actually produced
Nine SER runs, 10 seconds each, ~151 frames at 15 fps, plus two snapshots. Road sign, a rower out on the water, and the far shoreline. 28 GB.
Runs 1–4 were shot with Exposure = AUTO and WB = AUTO, which makes them close to useless for stacking — auto anything means every frame has a different transfer curve. From run 5 I went manual: 8 ms, then 16 ms for the rest. That progression is the right one.
What these runs cannot settle is the Barlow question, and it took me a while to admit it. Edge rise across three nominally identical no-Barlow runs came out 50, 28 and 41 px. That spread is larger than the effect I was trying to measure. I also moved between runs, so the scale changes for reasons that have nothing to do with optics.
Run-to-run focus variation dominates everything else right now. That is the finding. Not the Barlow — the fact that I cannot yet hold focus from one run to the next, on a springy floor, with a focuser I am still learning.
The Barlow verdict stands on the physics rather than on these frames: 1.4 um pixels want about f/7, the tube is f/9.2, and f/18.4 cannot carry more information no matter how it looks.
Where this leaves it
Two nights, no sky. But the tube is understood, the Barlow question is settled, the defocus is a measured 2.83 mm rather than a mystery, and there are 28 GB of daytime frames to practise processing on.
The verdict on the optics still hasn't been taken — daylight at 80 m is shimmer-limited at around 5 arcsec, three times worse than the telescope. That test needs a star. Alpha Centauri, once the finder is aligned, is a double a few arcseconds apart. If it splits cleanly into two, the 76/700 is doing its job.