I’m designing sundials and building models of the designs in Sketchup. A given design will work fine in some locations (i.e., Midwest US) but will be off by several minutes in others (i.e., Western US). As a check, I imported a location into the free web version of Sketchup in the western US to set the geographic location. I chose a location directly on the 120 W. meridian, the baseline for Pacific time. I erected a vertical pole, turned on shadows, and set the time to 12:00 and the date to September 1; this should result in the sun casting shadows directly north, but instead it casts a shadow slightly to the east; backing off the time to 11:55 am renders the shadow correctly. September 1 was chosen because the Equation of Time on that date is very near zero.
I noted, too, that the imported map imagery is rotated a few degrees clockwise - it looks like this is the same angular error I’m seeing in shadow direction.
There are similar but smaller errors at other locations; in Michigan, the error is about one minute.
All documentation states that the green axis in Sketchup is aligned with true north, but this is not the case. I suspect Sketchup is using a poorly georegistered map for its web version, but I’m not certain.
Note: I can easily rotate the model itself if that’s the only solution, but I’d need to know exactly what the problem is so I can calculate the correct counter-rotation to apply.
Note: the ~5 minute difference between accurate and displayed positions is equivalent to an angular solar position error of about a degree.
This could be wrong, but I recall reading that Sketchup north is aligned to northing of the UTM grid for the location. Would that explain what you see?
Any map grid is aligned to true north only at its datum. And it is worse - solar north aligns to true north only four times a year. And as the solar time in Sketchup uses time zones instead of local solar time, this applies only to the meridian of the time zone.
SketchUp uses grid north, I imagine so that it then lines up with north shown on imported maps. How far off from true north that is depends on how far away from a line of longitude you are.
You can use this extension to adjust where true north is:
As noted, I selected September 1 to ensure alignment (this is one of those four dates); and 120 W longitude is plunk on the timezone baseline, so there’s no time zone offset.
Ugh. This appears to be the problem; if I move to the middle of a UTM zone (~112 W longitude) the problem is minimized, and gets worse out nearer the edges of the zone (which is where I noticed it being particularly bad). As noted, the error is about a degree, close to what I’d expect near UTM zone edges. This means Sketchup’s solar renderings lack precision, which is critical for this particular application.
Sadly, I’m using the free web-based version and extensions don’t seem to be an option. I may have to switch to Blender for this project, which seems like massive overkill. Maybe I can calculated the precise UTM northing error and just counter-rotate the model.
Yes. As noted, I eliminated the sundial as a test and just used the shadow of a vertical pole, which should point due north at noon per my description. Instead it’s about 4-5 minutes (about a degree) east of north.
The problem seems to be that Sketchup uses UTM grid north instead of true north - an odd choice given that it touts solar modeling as a feature.
As a hobbyist, this is an unacceptable error. The misalignment itself is near the five minute accuracy being designed for. It’s also annoying that not only isn’t it documented well, what documentation there is is misleading, claiming that the green axis is aligned with true north but failing to mention that map locations are not.
Will the Shop version do this properly with the extension mentioned above, or only Pro?
Blender seems like a heavy lift, but will at least work correctly.
If you are thinking of a desktop application as a solution to the missing feature in the web application, you could download SketchUp Make 2017 from here:
That is free for non-commercial work, and it can use the solar north extension.
Confirmed that this is a true north versus grid north problem. The grid convergence - the angular difference between these two - is given by:
γ = arctan [tan (λ - λ0) × sin φ]
where
γ is grid convergence,
λ0 is longitude of UTM zone’s central meridian,
φ, λ are latitude, longitude of point in question
After looking up the UTM zone for my problematic location and its central meridian, the above gave a convergence of ~3.2 degrees; measuring the discrepancy of my vertical pole, as described earlier, likewise gives a measurement of 3.2 degrees. So a counter rotation eliminates the problem in a quantitative way, or at least confirms that this is the issue.