Over the last two months, I’ve sold a lot of my film photography gear, and bought a whole bunch of other film photography gear. In an effort to save a few bucks, I’ve also dipped my toes into vintage camera repair.

On the whole, my repair efforts have been a fascinating exploration of electromechanical design. In the 1960s and 1970s, through-the-lens light metering and automatic exposure were considered core features of consumer cameras, but semiconductor technology hadn’t yet reached a point where all of the camera’s functionality could be shoved into a single chip. The cameras that resulted from this situation are ingenious blends of mechanical and electronic design. Just take a look at what you’ll find inside of a Minolta SRT-102, a camera whose design comes mostly from the mid-1960s:

Underneath the SRT’s elegant mid-century shell, there’s a rats nest of pulleys, springs, strings, and gears. It’s fundamentally a mechanical design, yet the signs of electrification are clearly present - note little bundles of wires snaking around the mechanical bits.

The SRT and other cameras of this time period are elegant marriages of electronics and mechanics. It’s a joy to poke around inside this thing, and discover all the clever tricks they played to make the whole thing work. These designs have some of the intelligence of the electronic designs of the 1980s, but their mechanical roots make them far easier to maintain and repair decades after they were manufactured.

There’s a catch, though. Although it’s easier to service these older mechanical cameras, they also need to be serviced much more often than their electronic counterparts. These mechanical cameras need regular tune-ups in order to work properly. They’ll keep working for decades longer than their electronic counterparts, but they’ll work a lot worse over time… unless you give them a little TLC.

Mechanical shutters are the worst offenders when it comes to mechanical camera maintenance. These precise, fast-moving devices rely on springs and ratchets to work properly. As the years wear on, those springs get stretched out, and the lubricants that keep everything running smoothly get gummy and gross.

In order to give a mechanical shutter the necessary tune-up, you need to be able to check the shutter speed somehow. These shutters can move really, really quickly, though. A stopwatch isn’t going to help you here, and neither is your phone’s camera - neither is fast enough to check a shutter that’s only open for a couple milliseconds. Off-the-shelf shutter test devices are available, but they’re usually targeted at professionals and they’re wildly expensive. So, let’s make our own!

A Resistor and a Prayer

This seems like a pretty simple device, right? Just put a flashlight on one side of the camera, and a light sensor on the other side, and use the light sensor to measure the amount of time the light passes through the shutter. That sums up my first prototype. A flashlight on one side, and a light-dependent resistor (LDR) on the other.

Since all the cameras I want to test use the same lens mount, I drilled a hole in a camera body cap, stuck an LDR in the hole, and taped it all together.

Conveniently, the LDR was already on a little circuit board with a 10k resistor, forming a simple voltage divider. I put a small voltage across the divider, stuck a phone flashlight behind the camera, hooked up the center of the voltage divider to my oscilloscope, and my shutter tester was complete!

I started with the 1-second setting on my SRT-102, and the results looked decisive:

These are exactly the kinds of results I was hoping to get. I know this camera’s 1-second setting is a little bit slow, and the oscilloscope readout confirmed that. However, as I tested faster and faster shutter speeds, I quickly discovered that this simple approach had some major limitations. Here’s an oscilloscope readout at the 1/250 shutter speed:

That’s not nearly as decisive. It isn’t clear what part of the curve I should count as “shutter open.”

This curvature shows a critical problem with this design: Light-dependent resistors don’t instantly respond to changes in light level. It takes a moment for their resistance to change. In my case, the LDR’s response time means that measurements aren’t very decisive at shutter speeds faster than 1/125. These mechanical shutters tend to have their worst issues at fast shutter speeds, so this severely limits the tester’s usefulness.

Well, That Didn’t Work

There’s an easy way around the LDR problem: just don’t use one. There are plenty of other ways to detect light, and some of those solutions can respond to light level changes in a handful of microseconds.

One of the more common examples of a quick-responding light detector is a photodiode. A photodiode is a diode that generates a small current when it is hit with photons.

So all I needed to do was dig up a photodiode… yet, somehow, I couldn’t find a single one.

I don’t have a massive electronics workshop at my disposal, but I do have a decent stockpile of parts that I have been accumulating since I was a kid. That stockpile keeps my costs down on projects like this, and is a large reason why this blog is possible. I almost always have the parts I need to throw together simple prototypes, or I can at least have something sorta close. This time, though, no dice.

I did, however, find a tiny solar panel (erm, solar cell) that I probably yanked out of a yard light 10+ years ago. Remember, a photodiode is a semiconductor thingy that creates an electrical current when exposed to light If this sounds a lot like a solar cell, that’s because they’re the same thing! The two devices share exactly the same underlying physical principles. Solar cells are optimized for high power output, while photodiodes are generally optimized for quick response times and precise responses to variations in light level.

The solar cell was to big to fit inside the lens cap, so I just taped it to the front of the camera:

This is hardly a surprise, but it didn’t work very well. The solar cell’s response was worse than the LDR, showing slow response even at 1/60:

There’s another critical problem with the solar cell, though. Notice how the cell fills nearly the entire area of the lens mount, while the LDR filled only one small point. This was critical to the LDR’s effectiveness, and it’s detrimental to the photocell.

Rolling Shutters and Measurement Accuracy

Let’s learn a bit about how (most) cameras work. Importantly for our purposes, the following applies to nearly every mainstream 35mm SLR camera ever made.

At slow shutter speeds, the entire shutter opens at once, light hits the film, and then the entire shutter closes at once. Simple enough, but this strategy doesn’t work at faster shutter speeds - the mechanism in most cameras simply couldn’t move fast enough for the “open all at once, close all at once” approach to work.

Instead, when you take a picture at a fast shutter speed, the shutter will only expose a little slice of the film to light. That slice moves across the film plane to create a complete image. The width of that slice, and the speed of its travel, determine exactly how long any particular point on the image is exposed to light. This is called a rolling shutter, and it’s how most cameras produce images at their fastest shutter speeds.

Because the solar cell is “watching” the entire plane in front of the shutter, it can’t possibly take a useful measurement once the rolling shutter effect kicks in. There are only two ways to fix that problem - make sure the light comes only through a very narrow point on the film plane, or ditch the solar cell altogether and try something else. I chose the latter.

This is Fine

I did a little more digging, and came across an old sensor kit that I rarely use. Inside that kit, there was a board labeled “flame sensor” with a component that looked an awful lot like a photodiode. As it turns out, this weirdly-named board is just a photodiode and a couple op-amps used to generate a digital signal. That’s exactly what I wanted to build, and it’s already assembled on a little board!

This would be wonderfully convenient, except the flame sensor’s photodiode is an infrared photodiode - it’s only sensitive to IR, and (of course) the cheap sensor kit doesn’t specify what wavelength. I rifled through my spare parts bins again, and found a handful of seriously mangled IR LEDs. I’m not sure where they came from, but I’d guess that I yanked them out of TV remotes when I was a kid. I got lucky - these IR LEDs trigger the flame sensor!

This arrangement turned out to be a substantial upgrade. Besides being far more decisive than the LDR, I can also tweak the position of the light source without blinding myself. Using the mount from an old tripod head, plus a steel bracket I had laying around, I created a nice little jig to hold the camera and light source steady:

Stiff solid-core wire holds the two LEDs in a consistent spot behind the camera’s shutter. I can bend the wires so that the LEDs line up properly for different camera models. I can also adjust the wires to test different areas of the shutter, but this proved to be much more finicky. The IR photodiode didn’t pick up the IR light when the LEDs were positioned in the extreme corners of the shutter. To remedy this, I fitted the lens cap with some aluminum foil to reflect some light towards the photodiode. I didn’t expect this to work, but it actually did!

Unsurprisingly, this arrangement worked much better than the LDR. Here’s an example of the XD-11’s shutter at its 1/250 setting:

The rigid mounting jig combines nicely with the photodiode to make a brutally simple, yet very effective shutter speed tester. For whatever reason, it can’t register 1/1000 on most of my cameras. This is a “flame sensor,” after all, and fires don’t usually last for exactly one millisecond. This is a really frustrating limitation, though.

Some general observations I made while testing my cameras:

• My Minolta SRT-102 was professionally serviced in 2022, yet it’s still not very accurate. It’s within a third of a stop from 1/4 to 1/250, but it’s slow by as much as a full stop at 1/1000. The longer 1/2 and 1-second shutter speeds vary wildly, but average out to half a stop of overexposure.
• The XE-7 is extraordinary. As far as I know, neither of my XE-7 bodies has ever had any professional work, yet these 50 year-old cameras are accurate within a tenth of a stop from 1/500 all the way up to 4 seconds. I can’t measure the 1/1000 setting reliably.
• My XD-11 is really good, but not quite as dead-on as the XE twins. All shutter speeds are within a fifth of a stop until you get down to the 1/500, which is slow by a third of a stop. I can’t measure the 1/1000 setting reliably.

Problems And Next Steps

As was already mentioned, this lowly “flame sensor” device seems fully incapable of detecting shutter speeds any faster than about 2 milliseconds. This means that I can’t properly test the 1/1000 setting on any of my cameras. That’s a big deal, because on many old cameras, that setting has its own internal adjustments that work independently from the other shutter speeds. I can’t fully calibrate a camera on my own unless I can measure the 1/1000 shutter speed.

The light coming from an LED isn’t a perfectly straight ray - instead, it spreads out all over the place. As such, my shutter speed tester isn’t really measuring any specific point on the shutter. It’s measuring some small area of the shutter. When the rolling shutter comes into play, different parts of that area are exposed to light at different time intervals by design. As a result, these measurements aren’t very trustworthy.

To improve the situation, I will need to redesign the shutter tester so that the light sensor can “see” only the tiniest possible area of the film plane. To make that work, I’ll have to flip the entire contraption backwards. Instead of positioning the light source right next to the shutter, and the light sensor far away by the lens mount, it’ll have to go the other way around.

To make a circuit that can reliably measure shutter speeds at 1/1000 and beyond, I’ll also need better IR LEDs, better photodiodes, and greater visibility into the circuit’s behavior. That’s for a future blog post, though. In the meantime, you can help fund my rabbithole-spelunking activities by buying me a coffee.