How do I determine the shutter speed to use for a roll?

I’ve just digitized my first roll of film and I realize that I don’t know what shutter speed to actually use.

What I did was to take the first exposed frame, set the camera to aperture priority mode and have the camera measure it for me. The aperture priority mode gave me a 1/40 shutter speed when I was using ISO 100 and F8. I read that some prefer to set the shutter speed 1-stop higher, so I decided on 1/30 when I switched back to manual mode. I then went through the entire roll and shot all the frames with ISO 100, F8 and 1/30.

I did check with aperture priority mode for some of the other frames when I shot them. For some the camera would give me 1/125 instead of 1/40, if they had been the first frame I would’ve used a completely different shutter speed for the entire roll. This made everything feel very arbitrary. Would it be better to use the brightest or the darkest frame of the roll when I measure the shutter speed instead of using the first one?

I also read somewhere that I can use an unexposed frame and measure the shutter speed on that one. I did try that but then I got a shutter speed of 1/200, and all the sample shots I took on exposed frames were extremely dark.

So, is using the first frame as the reference for the entire roll the way to go or should I do something else?

1 Like

If you meter through blank frame, same rules as metering in nature apply: The exposure displayed will lead to a “middle gray”. – In a given exposure time, the maximum amount of light will pass through a blank frame. The more you increase the amount of light in a given time, the faster will it “fill” the sensor, up to a certain point where that amount exceeds the sensor limits. So, when digitizing, you’ll want to avoid this amount.

Metering on an “actual” image, perhaps adding another stop of light, may lead to an suitable scan exposure – which may suit other frames or not.

I write an artice about that, hope this helps: Find the best exposure for your DSLR scanning


"I also read somewhere that I can use an unexposed frame and measure the shutter speed on that one. "

I think you could do a version of this but I think what was meant is that you base your exposure on this, so you lengthen the exposure that your camera gives you for a blank frame and add 3 stops (say), which would be 1/25th sec in this case. However you might vary this according to experience, how were your conversions at 1/30th and f8?

Nate does seem to favour keeping the exposure the same throughout the roll:

Tom Kyle’s page looks good to explain the principles and ‘ETTR’

I did see the article, Find the best exposure got your DSLR scanning, but I felt it went a bit over my head. I needed a more beginner friendly approach. I’ve re-read it a couple times now and I think I understand the idea, but I’m not sure if I’ll have the knowledge to actually do it in practice, and I won’t have unexposed frames for all the rolls.

I’ve done a second attempt with the first roll, and instead I used spot metering on the unexposed space between two frames, it gave me 1/250 and with 2 1/3 added stops it resulted in 1/50.

The 1/30 looks better after the initial conversion, a larger part of the photos just need some fine-tuning, but the result I start with is better I think.

With 1/50 I need to do a lot more adjustments, the photos are much darker, the white balance and tone is more off and I still haven’t been able to brighten them without getting them over saturated and with too much contrast. Perhaps I’m just bad at post-processing, but for me 1/30 is easier to work with on this roll.

Spot metering on the unexposed space and 3 full stops would result in 1/30 on the first roll. Perhaps that’s something I can try on the second roll and see where it gets me.

1 Like

I think what you’re finding is that you need an exposure of at least 1/30 sec at f8 for that film since 1/50th isn’t giving good conversions. Since there is a general recommendation to give more exposure rather than less you might also try slightly longer, 1/20th or even 1/15th just to see if the conversions are better overall. With all these slow shutter speeds you have to absolutely meticulous about avoiding any vibration in your setp so remote or timed release is vital. I do know that some on here might use a fixed shutter speed for a roll as Nate suggests but that for ‘difficult’ negatives, say under or over exposed ones, they might take a range of exposures. I think that pretty quickly you’ll get your eye in.

Just try a series of the same negative and bracket from proposed exposure upwards, until the histogram gets close to the right edge of the histogram area… doing this might help.

In thinking about exposure for roll analysis, which I have yet to test but will be soon, I think it important to bear in mind that we are dealing with negatives so outcome will be the reverse of input. In this case, the negatives start with the ultimate shadows occupying the right of the histogram and the ultimate highlights the left. This has the fortunate effect that if we use the ETTR approach, we will be using 50%+ of the lightness levels on what will end-up as the shadow areas in the reversed image, which is probably a good thing for both shadow detail and noise suppression; and in line with that, what ends-up as highlights will be receiving the least amount of lightness levels. Remember how it works - for a 12-bit camera sensor, there are 4096 levels and one stop-equivalent down we are at 2048, etc.

So, if an important objective is to avoid clipping of both ultimate highlight and ultimate shadow detail, we need to be just as concerned about the left side of the histogram as the right (as we should in normal positive photography anyhow). ETTR will help with this, but not necessarily enough. It depends on the photo. So how this relates to roll analysis is this: if the full dynamic range of each negative in the roll fits within one exposure per frame without clipping right or left, that is an ideal situation for roll analysis. However if there were individual photos in the roll where it is not possible to pack all the DR into one exposure, perhaps there may be an issue for those frames - I shall be looking out for this. It may not invalidate the basic premise of roll analysis as Nate explains it, but it could mean that some custom re-exposure for some frames may be necessary. For example, I am finding in my work that there are indeed negatives showing clipping at one end or another, and to accommodate that, I need to do HDR at the capture stage by varying the exposure of the same frame ensuring I have one exposure with all the right side in gamut and another with all the left side in gamut, then doing HDR blends in Lightroom. This is usually highly successful for producing satisfactory results amenable to fairly easy post-capture editing.

I’m not sure there is a need to complicate things here.

Shoot at like F8 and then use a live histogram in manual mode to expose to the right without clipping highlights? Whatever the shutter speed is, use that?

The live view histogram does not show where the highlights (in RAW!) start clipping. So its right edge shows the wrong limit. One can do that as you’re proposing though, Quinton. It indeed is simpler, possibly better than „just metering“, but not „as far as possible.“

I’m probably missing something, but I can see blown out highlights on the Canon platform.

Anyway, I don’t think digitising film is an exact science anyway. There is no solution.
Scanner software will have their own algorithms for finding black and white points, then Nate has his for NLP, then a million other factors involved during before and after as well as white balance.

I just think shoot RAW, expose to the right so you capture as much as possible digitally and everything else is done by eye. Don’t worry too much about everything else :slight_smile:

Edit: oh, just saw your article. Interesting tom

It will, if you set your camera for UniWB. The following sites can be useful

Short version: UniWB mimics a RAW histogram by setting the white balance in such a way that the histogram (which is mostly used from preview data) closely corresponds to what the sensor captures.