[MossyRocks]

Now I really want a K-3iii. I love the K-3 + O-GPS1 but being able to do such things without going through the process of calibration or dealing with magnetic issues in a number of locations would be great. The only thing I can think of to make it better would be if it could use the O-GPS, compass, and level sensors to shorten the initial shot period or eliminate it for successive shots would allow capturing more shots. Now I wonder how it would do with sticking a K-3iii behind my A* 400/2.8 with the 1.4X-L or 2X-L converter opening up some nice shots of smaller DSOs that the K-3 just kind of chokes with since it doesn't have an option to enter 560mm and 800mm seems beyond what I can get a good enough calibration to do. The over1 additional stop of dynamic range that the K-3iii offers would also be a nice benefit.

[bertwert]

I thought I stumbled across an old thread when I saw and read this one. AstroTracer has been a Pentax feature for quite a while now and everyone was talking like it was something jst discovered. Those are some great shots and the third one testifies to pretty good tracking using AstroTracer. Excellent work for someone just discovering what AstroTracer can do!

Aha yes, astrotracer has been around for years - but relies on the GPS to get the camera's position and direction pointed, in order to determine how much to turn the sensor.

This new K-3iii feature replaces the GPS (it does not have on board) by taking a short ~30s photo first with star trails and then calculating from that how much to turn the sensor for the subsequent shot.

Certainly new and innovative - real impressive in my mind!

[BigMackCam]

That's amazing... How does it work without GPS to determine position??

[AgentL]

I installed the beta 1.40 firmware (which is only available for Ricoh Imaging Photo ID members at the moment), and the sky was clear. I HAD TO test GPS-less astrotracing aka Astrotracer Type 3!

I set up my tripod in the back yard and quickly shot Orion nebula using DA* 300 at F/5, 25 seconds, ISO 400. Light pollution is terrible here so I used a cheap light pollution filter. First attachment is SOOC jpeg with Astrotracer Type 3 ON. For comparison, the second one is w/o any astrotracing. You can see how nicely the tracking worked, it really felt like magic, it was just a matter of pressing the shutter button and the camera took care of the rest!

When I pressed the shutter button, there was an initial exposure of about 30 sec, and then there was another shutter actuation to take the real picture with specified duration. I tried longer exposure up to 2 minutes but the first exposure was always about 30 sec. Anyway, I assume that the camera doesn't move the sensor for the first exposure, analyzes the star trailing and figures out how much the sensor should be moved/rotated per second.

25 sec was not the limit of the astrotracer, I started with 60 seconds and it worked w/o problem, but I was overexposing and eventually settled on 25 sec. (You can still see that the center of the nebula is overexposed, 30sec or so with ISO 200 would have been better.)

I shot 7 pictures (total of 175 sec) and combined them using Sequator, and did a very quick and dirty PP job in Lightroom, cropped and resized, that's the 3rd attachment.

If I don't have to work tomorrow I'm still shooting (after adjusting the exposure) and the result would have been greater, but this was a huge win for me as the first night of GPS-less Astrotracer!

Fantastic!

[MossyRocks]

That's amazing... How does it work without GPS to determine position??

I don't think it calculates position and orientation but instead calculates the movement, x/y translation and field rotation, based off of a few shots with the camera's current orientation. I would guess a minimum of 2, but 3 or 4 shots wouldn't surprise me, so that it can determine the actual direction of movement and how much. Basically a bunch of math like with using regular astrotracer with a GPS but with the addition of some image analysis.

[Bob 256]

Now I get it!!! This is a current thread! Great idea. It most likely finds the brightest stars (compares them to an internal star map) and from that, it can calculate star movement (and positions) pretty easily (theoretically, it should be able to do a better job than the GPS based AstroTracer). A lot of auto-pointing scopes use this approach, having the user point at a couple of known stars (all that's needed) and it can then calculate positions for the rest using spherical coordinates. It appears to do a really good job from the OPs photos.

[BigMackCam]

I don't think it calculates position and orientation but instead calculates the movement, x/y translation and field rotation, based off of a few shots with the camera's current orientation. I would guess a minimum of 2, but 3 or 4 shots wouldn't surprise me, so that it can determine the actual direction of movement and how much. Basically a bunch of math like with using regular astrotracer with a GPS but with the addition of some image analysis.

Now I get it!!! This is a current thread! Great idea. It most likely finds the brightest stars (compares them to an internal star map) and from that, it can calculate star movement (and positions) pretty easily (theoretically, it should be able to do a better job than the GPS based AstroTracer). A lot of auto-pointing scopes use this approach, having the user point at a couple of known stars (all that's needed) and it can then calculate positions for the rest using spherical coordinates. It appears to do a really good job from the OPs photos.

How clever. I'm not an astro-photographer, but I love viewing others' images. This is a really great feature... and yes - it does an excellent job judging by the OP's results. Good stuff!

[Snapppy]

This is a big plus for me when thinking about the K3III. I was pretty disappointed when it was announced without built-in GPS. I love astrotracer, and use it every chance I can get. It’s actually a pretty good workflow with the built-in GPS of the K-1 - but of course this sounds like it would be easier. I’ve also found it can be subject to not working sometimes, requiring repeated calibration. This seems even simpler and solves the K3III not having GPS for me. I don’t use it for tagging, just for astrotracer

[MossyRocks]

It most likely finds the brightest stars (compares them to an internal star map) and from that, it can calculate star movement (and positions) pretty easily (theoretically,

LIkely no. Even if it had an internal star map it still wouldn't be able to get the direction of movement since it would lack the direction (clockwise or counter clockwise) of movement but would know how much. Also lenses have varying about of geometric distortion in them so to make use of that it would have to correct for the distortion each time adding additional processing and taking up more internal memory. I'm not saying it couldn't do that but it seems like that would just add complexity when all it would need would be to figure out the movement across the frame over time and not bother with trying to figure out position and orientation and then calculate the movement across the frame.

A lot of auto-pointing scopes use this approach, having the user point at a couple of known stars (all that's needed) and it can then calculate positions for the rest using spherical coordinates.

Sort of. I belive what you are referring to is the goto functionality that is on a lot of equatorial and alt-az mounts. For that to work correctly it does require proper calibration, so for an equatorial it requires doing a proper polar alignment and setting to the zero position, and for some of the alt-az amounts it requires using a GPS, usually built in, and also going through a calibration process not to dissimilar to the calibration done for astrotracer with the O-GPS1 or inbuilt one. For mounts that allow you do to a multistar alignment, mostly equatorials, it isn't to figure out where the scope is pointed but it to get a better polar alignment. In those cases you get a pretty good polar alignment where polaris or the southern celestial pole is reasonably close with the scope and mount in the zero position. From there the scope will slew to a bright visible star and you then need to get it centered in the frame and repeat with a few stars to really get it properly polar aligned.

[kwb]

The manual says "By using interval shooting together, it is possible to perform the main exposure multiple times for each test shooting."

I haven't tested interval shooting yet, that's something for next time. I was pressing the shutter button manually for each exposure (which means that there were 7 test exposures total).
Probably 4 shots per one calibration would have been OK in this case. After a few minutes the camera needs to be re-positioned anyway unless I'm willing to crop more, GPS or no-GPS. That's the limitation of Astrotracer in general.

I wonder how this compares to the standard Astrotracer. Does it actually have better performance?

At least in this example where all stars move relatively uniformly in the frame, from my experience I say it's as good as the standard Astrotracer with GPS/compass/tilt-meter as far as the end result is concerned. Better in that I didn't have to worry about calibration.
OTOH, once you do the calibration right for standard Astrotracer, you can point the camera anywhere you want in the sky without re-calibrating. That's not the case for Type 3, but that's a VERY small price because Type 3 "calibration" is relatively short (the manual says 30-50 sec) and completely automatic.

But Type 3 "calibration" could fail (or it could pass but the result could be blurry) under some conditions according to the manual, i.e.

• anything other than the stars are in the frame,
• you cannot get enough exposure for stars because of heavy light pollution etc.,
• focus is too soft,
• there's camera vibration.

They recommend GPS for starscape (landscape with stars) with wide angle lenses. Also there are some restrictions about the lenses.

• No fisheye.
• You can only use the wide and tele end of zoom lenses. The only lenses where the entire zoom range can be used are:
  • HD PENTAX-DA 55-300mmF4.5-6.3ED PLM WR RE
  • HD PENTAX-DA* 11-18mmF2.8ED DC AW
  • HD PENTAX-DA* 16-50mmF2.8ED PLM AW

I don't know what the last one about zoom range really means. The only Pentax K-mount zoom I have is 55-300 PLM, but I'll use my Sigma 16-55mm zoom to see what happens for in-between focal length.

[kypfer]

I wonder how this compares to the standard Astrotracer. Does it actually have better performance? Presumably every shot needs to be preceded by a calibration shot which halves the number of shots you can take in a given time.

Do you think you can take one calibration shot and then take many shots without recalibration?

Quote from an "official" Pentax page Firmware for PENTAX K-3 Mark III released on December 7 (Young Generation) | PENTAX official :-

"IT'S A DREAM-LIKE FEATURE, BUT IT'S ALSO LIMITED. BASICALLY, IT IS RECOMMENDED TO TAKE A PICTURE OF A STARSCAPE USING AN O-GPS1 WHEN TAKING A PICTURE OF A STAR USING A WIDE-ANGLE LENS"

So there you have it … not better performance, but as good as, in some circumstances.
OK for deep-space objects like the Andromeda galaxy, not so good for the whole Milky Way!
The calibration seems to be done by "making" a calibration shot and analysing the result, then applying these results to the eventual exposure.
This would imply that, depending on where in the sky you point your camera, and/or possibly what focal length lens is fitted, that a separate calibration sequence would be necessary for each different framing, to compensate for the potentially different rates of star movement.

[Bob 256]

LIkely no. Even if it had an internal star map it still wouldn't be able to get the direction of movement since it would lack the direction (clockwise or counter clockwise) of movement but would know how much. Also lenses have varying about of geometric distortion in them so to make use of that it would have to correct for the distortion each time adding additional processing and taking up more internal memory. I'm not saying it couldn't do that but it seems like that would just add complexity when all it would need would be to figure out the movement across the frame over time and not bother with trying to figure out position and orientation and then calculate the movement across the frame.

Sort of. I believe what you are referring to is the goto functionality that is on a lot of equatorial and alt-az mounts. For that to work correctly it does require proper calibration, so for an equatorial it requires doing a proper polar alignment and setting to the zero position, and for some of the alt-az amounts it requires using a GPS, usually built in, and also going through a calibration process not to dissimilar to the calibration done for astrotracer with the O-GPS1 or inbuilt one. For mounts that allow you do to a multistar alignment, mostly equatorials, it isn't to figure out where the scope is pointed but it to get a better polar alignment. In those cases you get a pretty good polar alignment where polaris or the southern celestial pole is reasonably close with the scope and mount in the zero position. From there the scope will slew to a bright visible star and you then need to get it centered in the frame and repeat with a few stars to really get it properly polar aligned.

Once two or more known stars are located (only one needed if it's identified and a star map is available), they can be mapped to a sphere and if the axis of that sphere is known (which it is for the celestial sphere) movements of all other objects on that sphere can be calculated (your location on earth is also needed). Lens distortions would make little difference since the lens with those distortions is being used to identify the anchor points (in fact this is a bonus since the exact magnification for the lens can be determined which is needed in the calculations).The final result is a slew angle and speed for the sensor. Another approach that might have been used would be to use feedback to anchor some established star images in place by moving the sensor in between image samples (this again would require some bright star references for best performance).

You're right in saying equatorial mounts need some calibration up front because they can only move in one way (around the polar axis) when tracking. Newer scopes ("computerized") are available, using micro-stepper (or servo) motors and encoders, and only require pointing at a couple of known stars to calibrate. They can be guided in much the same way as sensor-slewing (an x-y motion using two steppers or servos). It's a more complicated system than the old one ($$), but a whole lot easier to set up. No polar alignment is required because the mount isn't a polar mount (it's basically an x-y positioner, some alt-azimuth type). All of these can accomplish go-to type operations.

Anyway, it's nice to know that this technology has migrated to just plain old cameras

[Kevin B123]

Once two or more known stars are located (only one needed if it's identified and a star map is available), they can be mapped to a sphere and if the axis of that sphere is known (which it is for the celestial sphere) movements of all other objects on that sphere can be calculated (your location on earth is also needed)

I think you are over analyzing it here. The sensor is a flat plane with projected bright spots of light on it. All it needs to do is note which pixels register the bright spots at the start and after 30 seconds and the relative movement is then found. Now the sensor copies that movement during an exposure. Brilliantly simple and effective. No need to know anything about azimuth or elevation or compass heading or anything else as I see it.

[The Squirrel Mafia]

This is awesome!

That's amazing... How does it work without GPS to determine position??

I believe that it does not need position or anything at all, only the curve the stars make during the 20-30 seconds. From that it creates an equation that is the basis for the movement of the sensor. It is genious!

Why it does not work well with wide angle is because of the to short trails that is created during 20-30 seconds. The more tele the more accurate I guess. Why did´nt I think of this?