[biz-engineer]

3. Put the camera in high speed burst mode

Ah that's a good idea, because the dead time between frame is shorter in burst mode than it is in interval mode, so burst mode is better for star trials for instance.

I leave shake reduction on

I would do it on a tripod, sensor SR on tripod can be worse than no SR.

60x30sec averaging

Excellent images. In that second image we see some "banding" upper right corner above the lights. I've also encountered that issue when stacking exposure as there is not enough noise in the image to blur tone transitions of image areas with two successive tone values. I start to see that problem easily occur when shooting at 100 ISO and stacking more than 8 exposures, there is not enough noise to dither tone bands in JPEG outputs.

[MossyRocks]

I start to see that problem easily occur when shooting at 100 ISO and stacking more than 8 exposures, there is not enough noise to dither tone bands in JPEG outputs.

If you are doing any heavy processing you will quickly hit the limits of using jpegs as you have compression artifacts and are stuck with 8 bits per channel. Stick with raw.

[biz-engineer]

If you are doing any heavy processing you will quickly hit the limits of using jpegs as you have compression artifacts and are stuck with 8 bits per channel. Stick with raw.

Yes, I agree, makes sense. How to avoid banding at the last stage of the process, i.e export to JPEG? Upsample before JPEG export?

[MJKoski]

The banding is real and happens when processing 16bit TIFF files as well. Image becomes too clean.

[MossyRocks]

How to avoid banding at the last stage of the process, i.e export to JPEG? Upsample before JPEG export?

I just stick to raw imports, and 16 bit tiffs. Only once I am done do I do the export to JPEG and have never had a problem with banding. Even when working with astro images where I've stacked over 1000 images I still haven't had a problem with banding. I also don't turn on any noise reduction, not even the defaults, when doing the raw to tiff conversion so that may be where the banding comes from.

The banding is real and happens when processing 16bit TIFF files as well. Image becomes too clean.

I don't see how. For example this image is one where I did the simulated ND filter and stacked around 40 or so images to get the blur of the clouds moving:


Same with this image because I was going for a super res shot of the end of the second sunset at my lake property so I could do a big print to hang on the wall:


Even this image where I used a stack of 193 images and then did a massive amount of processing lacks banding. This one also had dark frame subtraction and bias frame subtraction to reduce the systematic noise in each shot before they got stacked. Granted it deep sky stacker spits out a 32bit tiff of the final stack to use as input for further processing. However photoshop doesn't let you do much processing other than some initial levels adjustments on a 32bit tiff. So after that it got converted to a 16 bit tiff where the vast majority of that massive amount of processing took place:


If you are seeing banding I would be looking for places in your process where a lot of data got tossed out. Averaging/median filter, sharpening, software noise reduction, down sampling, levels clipping, etc. would all be places where I would start to look.

[MarcusCalidus]

Generally my process for doing this is:

Great workflow. Thank you for sharing. I will definitely try this out .

[MJKoski]

My examples were all in-camera multiexposure averages saved as one .PEF RAW. 14bit RAW is not enough to provide smooth tonal separation with high enough frame count. When this happens, banding is visible with flat tone curve already in RAW conversion stage.

Photoshop stack could be better but requires insane amount of memory to stack 256 frames.

[MossyRocks]

Photoshop stack could be better but requires insane amount of memory to stack 256 frames.

Stack in batches. If I stack things in photoshop I don't stack more than 25 or so shots at once for the reason you mention, usually going for 16 at a time. So stacking 256 frames in photoshop I would do it as as 16 batches of 16 and then stack the results. This also helps with getting photoshop to properly align the images as with big stacks it gets confused at times.

[MJKoski]

I will give it a shot. However, I prefer in-camera averages which gives usually non-problematic results and does not waste disk space. One nice .PEF is just perfect

[photoptimist]

The biggest advantages of interval composite are:
1) it is built into the camera and works on every lens (no costly filters to buy, carry, and mount);
2) it can handle really long durations (e.g., stacking shots collected over hours);
3) it's easier to compose and meter without a dark ND filter;
4) the "save process" option lets you pick the amount of stacking and get a usable shot if something disturbed the scene late in the sequnece;
5) using "averaging" mode to stack short-shutter-time shots will have less dark current noise than the equivalent high-ND-long-shutter-time shot.

The biggest disadvantages of interval composites are:
1) repetitious artifacts if the scene has slow-moving bright objects (fixable with either a shorter or longer interval between shots);
2) shutter and mirror vibration can add some blur (fixable with a solid-enough tripod)


It's worth noting that interval composite has three stacking modes -- average, additive, and bright. "Average" is probably the best one for replicating the ND-filter effect. "Additive" can be useful for low-lighting conditions and light painting but it does accumulate dark-current noise. "Bright" is great for star trails and meteor showers, has strange creative effects with peopled scenes, does not accumulate dark-current noise but is noisier than the "average"mode. (I wish Pentax included a "dark" mode just for fun!)

[BrianR]

Here is a rare example were 512x average worked (technically):


In the image, there are chunks of ice going around in circles due to water movements.

Terrific photo! It looks more like the result of blend mode set to 'Bright' though?

[biz-engineer]

When this happens, banding is visible with flat tone curve already in RAW conversion stage.

Banding comes from total absence of quantization noise at the level of displaying a image over 8bits per channel. The reason why you experience banding and MossyRocks didn't , is because he is shooting night skies at high ISO and you are shooting very long exposures at base ISO. MossyRocks's stacks start with low SNR (High burst mode of 30 sec. exposures at ISO1600, ISO3200, ISO6400 exposures, containing a lot of noise to begin with) , taking a lot of stacking for the noise to be reduced up to bits (~48db) of SNR. Stacking K1 image at ISO 100 already reach 48dB of SNR with 8 images stacked. It doesn't matter TIFF, RAW of whatever coding depth is use, as soon as the SNR exceed the max SNR of 8bits per channel coding, there isn't enough variance of values of pixel of slow tone gradation to avoid banding when displayed by 8bits/channel display peripherals (graphic card + display). Adding a little bit of noise to the 8bits/channel output and the banding in skies immediately disappear.

[MossyRocks]

averages saved as one .PEF RAW. 14bit RAW

That is actually rather disappointing that when doing in-camera averaging you still just get a 14 bit raw instead of 16 bit. I understand the technical reason for 14 bits with a one off shot or other interval combination methods as providing more bits of output just falls into the category of making crap up but with averaging you are able to get to that real 16 bits and would even be able to drive the noise down below the lowest order bit with enough shots.

[biz-engineer]

That is actually rather disappointing that when doing in-camera averaging you still just get a 14 bit raw instead of 16 bit.

for me it's intriging that camera software doesn't use multiples of 8bits (16bits = 2 x 8) for processing image data, because everything is partitionned by 8bits multiples (memory hardware and momory allocation for the software). It's actually more annoying to have to program software to manipulate 14bits instead of 16bits. Nut wouldn't DNG or PEF be larger over 16bits? Does DNG allow 16bits?

[MJKoski]

Terrific photo! It looks more like the result of blend mode set to 'Bright' though?

You are right. It may be bright-mode which I used. 512x in-camera averages usually fail with magenta tint as I demonstrated.