Hasselblad X2D long exposures versus averaging

JimKasson · Nov 21, 2022

Inearlier posts I looked at the X2D shadow performance with 17-minute exposures at base ISO. I did this because a couple of people who like the long-exposure look asked me about it. But there's a better way to get that blurring effect if you want to minimize noise in the shadows: make a series of exposures with shorter exposures and average them in Photoshop or some other image editor.

This post is intended to demonstrate that.

The cameras were mounted to the Foba camera stand with an Arca Swiss C1 cube. The rest of the setup is as follows:

X2D with 38 mm f/2.5 XCD lens, ISO 64, about 10 feet from target
Single exposure at f/32 with shutter speed 17 minutes
34 exposures at f/5.6 with shutter speed 32 seconds with no delay between the images, using the built-in intervalometer
Manual focusing
Breakthrough 10-stop neutral density filter.
2 second self timer
Aputure 100 LED light, with diffuser

34 32-second exposures, averaged in Photoshop, 4.75 stops underexposed, 4.75 stop push before averaging

17 minute exposure (the DPR interpretation of the EXIF is wrong)

More details, images here:

https://blog.kasson.com/x2d/hasselblad-x2d-long-exposures-versus-averaging/

--
https://blog.kasson.com

yogi4fitness · Nov 21, 2022

Wow! Didnt realize the difference would be so stark!

SrMi · Nov 21, 2022

For averaging, I had a good experience with Octave Raw Tools. It is free, fast, and generates DNG raw image files.

bokesan · Nov 21, 2022

Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

JimKasson · Nov 22, 2022

bokesan said:
Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

Many, if not most, people do long exposures at base ISO with strong neutral density filters. I’m saying there’s a better way.

SrMi · Nov 22, 2022

bokesan said:
Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

I do not follow why you need lots of light.

When shooting on a tripod (typical for averaging), you always have more light than you need unless movement needs to be stopped.

When shooting handheld and there is not a lot of light, and your handholding capability limits your exposure (e.g., ISO 800, 1/10 sec, f/8), you can shoot a series of images, align and average them in post.

The only limitation of averaging is motion, as averaging simulates long shutter speeds.

bokesan · Nov 22, 2022

JimKasson said:
bokesan said:

Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

Click to expand...

Many, if not most, people do long exposures at base ISO with strong neutral density filters. I’m saying there’s a better way.

That's quite different from my use case, then - I rarely use an ND filter, and most of my long exposures are long because it's, well, dark.

- Chris

JimKasson · Nov 22, 2022

JimKasson said:

328725ffb8274017b6bb3ee613345f88.jpg.png

--
https://blog.kasson.com

bokesan · Nov 22, 2022

SrMi said:
bokesan said:

Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

Click to expand...

I do not follow why you need lots of light.

When shooting on a tripod (typical for averaging), you always have more light than you need unless movement needs to be stopped.

I meant to say that if you use an ND filter (aka "less light filter"), you have more light than you need.

And you need that if you want to get the same total exposure time. Yes, you can reduce shot noise by averaging exposures, but if your base exposure were 30 minutes (w/o ND filter), you would rarely have the time to take 30 such shots to average them

- Chris

colinsurprenant · Nov 22, 2022

bokesan said:
SrMi said:

bokesan said:

Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

Click to expand...

I do not follow why you need lots of light.

When shooting on a tripod (typical for averaging), you always have more light than you need unless movement needs to be stopped.

Click to expand...

I meant to say that if you use an ND filter (aka "less light filter"), you have more light than you need.

And you need that if you want to get the same total exposure time. Yes, you can reduce shot noise by averaging exposures, but if your base exposure were 30 minutes (w/o ND filter), you would rarely have the time to take 30 such shots to average them

- Chris

If your base exposure was 30 minutes at ISO 100, you could make 30 x 30s exposures at 6400 ISO (30s/ISO 6400 is about the same as 1800s/ISO 100) and AFAIU 30 exposures would result is 5.4x increase in SNR but a 30 minutes exposure would drop the DR significantly and induce other noise such as hot pixels etc. I a not sure exactly how many stops would be lost but judging from JimKasson post about that I would guess around 4-5 stops for 30 minutes?

So it seems that you would gain from doing multiple exposures in this scenario too.

JimKasson · Nov 22, 2022

colinsurprenant said:
bokesan said:

SrMi said:

bokesan said:

Hardly surprising, given that the averaged image got 34 x more photons.

It seems fair to point out that this technique works only when you have more light than you need.

Click to expand...

I do not follow why you need lots of light.

When shooting on a tripod (typical for averaging), you always have more light than you need unless movement needs to be stopped.

Click to expand...

I meant to say that if you use an ND filter (aka "less light filter"), you have more light than you need.

And you need that if you want to get the same total exposure time. Yes, you can reduce shot noise by averaging exposures, but if your base exposure were 30 minutes (w/o ND filter), you would rarely have the time to take 30 such shots to average them

- Chris

Click to expand...

If your base exposure was 30 minutes at ISO 100, you could make 30 x 30s exposures at 6400 ISO (30s/ISO 6400 is about the same as 1800s/ISO 100) and AFAIU 30 exposures would result is 5.4x increase in SNR but a 30 minutes exposure would drop the DR significantly and induce other noise such as hot pixels etc. I a not sure exactly how many stops would be lost but judging from JimKasson post about that I would guess around 4-5 stops for 30 minutes?

So it seems that you would gain from doing multiple exposures in this scenario too.

With dual conversion gain sensors, raising the ISO above base ISO, mean stacking, and keeping the f stop and total exposure time the same as a single long exposure helps a little by lowering the read noise, but the big win is by getting more photons on the sensor in the mean-stacked series.

WOOLF66 · Nov 23, 2022

Sorry if asking a stupid question:

If understanding You right with averaging you take several pictures with lower time and average them in Photoshop; so I don´t see any reason why the second underexposed picture should look any different than the first one. So if this is true why not take the first exposure 34 times and safe a lot of time?

JimKasson · Nov 23, 2022

WOOLF66 said:
Sorry if asking a stupid question:

If understanding You right with averaging you take several pictures with lower time and average them in Photoshop; so I don´t see any reason why the second underexposed picture should look any different than the first one. So if this is true why not take the first exposure 34 times and safe a lot of time?

Go back to the original post and look at the f-stops.

WOOLF66 · Nov 23, 2022

JimKasson said:
WOOLF66 said:

Sorry if asking a stupid question:

If understanding You right with averaging you take several pictures with lower time and average them in Photoshop; so I don´t see any reason why the second underexposed picture should look any different than the first one. So if this is true why not take the first exposure 34 times and safe a lot of time?

Click to expand...

Go back to the original post and look at the f-stops.

Yes, I think I understood that:

Instead of taking one shot at f/32 for 17min, you average in Photoshop 34 exposures at f/5.6 and 32sec.

My question:
Do You really need 34 exposures, You could also average 34 copies of the one f/5.6 and 32sec ? No?

JimKasson · Nov 23, 2022

WOOLF66 said:
JimKasson said:

WOOLF66 said:

Sorry if asking a stupid question:

If understanding You right with averaging you take several pictures with lower time and average them in Photoshop; so I don´t see any reason why the second underexposed picture should look any different than the first one. So if this is true why not take the first exposure 34 times and safe a lot of time?

Click to expand...

Go back to the original post and look at the f-stops.

Click to expand...

Yes, I think I understood that:

Instead of taking one shot at f/32 for 17min, you average in Photoshop 34 exposures at f/5.6 and 32sec.

My question:
Do You really need 34 exposures, You could also average 34 copies of the one f/5.6 and 32sec ? No?

That would produce an image that looked like the image you used to get the 34 copies. The average of {x,x,x,x,x,x,...} is x.

Photon noise is stochastic (and, incidentally Poisson), which is why this trick works.

Also, the point of this exercise is to simulate a long exposure. 34 identical shorter-exposure images wouldn't do that.

WOOLF66 · Nov 23, 2022

Thanks Jim for the info, got it!

Wolfram

JimKasson · Nov 23, 2022

SrMi said:
For averaging, I had a good experience with Octave Raw Tools. It is free, fast, and generates DNG raw image files.

I can't get that to work with X2D files. It's looking for an EXIF field in the converted DNGs called 'createdate'.

SrMi said:
SrMi said:

Error: A file was processed that is missing the CreateDate EXIF tag. Perhaps your file mask included non-DNG files?

Click to expand...

There is no such field in DNGs that I make with DNGConverter from X2D raw files. There is a field called 'Date Created'.

Lens : XCD 38V
Date Created : 2022:11:15 10:07:35
Document ID : xmp.did:c602e28c-4198-3f40-acc6-9e36c34f4ad2

To add to my difficulties, the program is written in such a way that the run-time environment seems to ignore most of the breakpoints I set. I'm using Matlab R2022a.

Here's how I'm invoking it:

createStackedDngs('e:\Calibration\X2D\Long Exposure bookcase\DNGs','maxtimedelta', 0);

Any ideas?

SrMi · Nov 23, 2022

JimKasson said:
SrMi said:

For averaging, I had a good experience with Octave Raw Tools. It is free, fast, and generates DNG raw image files.

Click to expand...

I can't get that to work with X2D files. It's looking for an EXIF field in the converted DNGs called 'createdate'.

JimKasson said:

JimKasson said:

Error: A file was processed that is missing the CreateDate EXIF tag. Perhaps your file mask included non-DNG files?

Click to expand...

Click to expand...

There is no such field in DNGs that I make with DNGConverter from X2D raw files. There is a field called 'Date Created'.

Lens : XCD 38V
Date Created : 2022:11:15 10:07:35
Document ID : xmp.did:c602e28c-4198-3f40-acc6-9e36c34f4ad2

To add to my difficulties, the program is written in such a way that the run-time environment seems to ignore most of the breakpoints I set. I'm using Matlab R2022a.

Here's how I'm invoking it:

createStackedDngs('e:\Calibration\X2D\Long Exposure bookcase\DNGs','maxtimedelta', 0);

Any ideas?

Hmpf.

I would contact the author Horshack. He is active on DPR.

JimKasson · Nov 23, 2022

SrMi said:
JimKasson said:

SrMi said:

For averaging, I had a good experience with Octave Raw Tools. It is free, fast, and generates DNG raw image files.

Click to expand...

I can't get that to work with X2D files. It's looking for an EXIF field in the converted DNGs called 'createdate'.

SrMi said:

SrMi said:

Error: A file was processed that is missing the CreateDate EXIF tag. Perhaps your file mask included non-DNG files?

Click to expand...

Click to expand...

There is no such field in DNGs that I make with DNGConverter from X2D raw files. There is a field called 'Date Created'.

Lens : XCD 38V
Date Created : 2022:11:15 10:07:35
Document ID : xmp.did:c602e28c-4198-3f40-acc6-9e36c34f4ad2

To add to my difficulties, the program is written in such a way that the run-time environment seems to ignore most of the breakpoints I set. I'm using Matlab R2022a.

Here's how I'm invoking it:

createStackedDngs('e:\Calibration\X2D\Long Exposure bookcase\DNGs','maxtimedelta', 0);

Any ideas?

Click to expand...

Hmpf.

I would contact the author Horshack. He is active on DPR.

Thanks. I’ll take it up with Adam.

Hasselblad X2D long exposures versus averaging

Community Leader

Attachments

Veteran Member

Veteran Member

Senior Member

Community Leader

Veteran Member

Senior Member

Community Leader

Senior Member

Member

Community Leader

New member

Community Leader

New member

Community Leader

New member

Community Leader

Veteran Member

Community Leader

Similar threads

Keyboard shortcuts