This came up the other day on Photrio where @RalphLambrecht wondered about why his Gossen light meters drain their batteries while they’re supposed to power down automatically. Now, I happen to have one of those Gossen meters powered by a 9V battery, so I figured I could have a peek inside to see what might be wrong with his.
For clarity’s sake – there’s nothing wrong with my Gossen Variosix F. At least not that I know of. I got it a few months ago from another Photrian – @Philippe-Georges. It’s presently my most-used hand-held meter and I prefer it for (quasi-)spot metering over my Sekonic L-358 (yes, the one with the annoying but tolerable contact problem with the spot attachment!)

Still, curiosity got the better of me. Ralph’s problem report states that he stashes away his meter with the battery installed, and then after a while (weeks? months?) he tries to turn on the meter only to find it dead.
Assuming that self-discharge of the batteries isn’t the problem, what could possibly be wrong with the meter? It’s important to note that these Gossen meters turn themselves off automatically, and will wake up once the meter-key on the side is pressed.
I’ve always liked these “soft-on, auto power-down” devices. It somehow feels so much nicer that the device wakes up without having to physically flick a switch – and that it’ll save you the embarrassment of finding the batteries empty after forgetting to flick the switch back off once you’re done with the machine. I’ve in fact built a few devices that work this way, so I’m aware in general terms of how something like this can be implemented.
The basic trick to these circuits is that…wait for it…they really never actually turn off, entirely. They just go into a state in which they draw a very, very tiny current (barely any at all, ideally) but that still keeps a part of the circuit ‘alive’ so it can respond to the press of a key. So the challenge is to minimize this ‘keep alive current’, which generally is done in two ways: (1) minimizing the number of physical components that are kept powered on when the device is ‘off’ and (2) minimizing leakage currents in/through those components.
Conversely, the failure modes fundamentally trace back to these two mechanisms as well. Overly rapid battery drain can be due to too large of a part of the device remaining active in power-down state. This is an engineering problem, and it’ll be characteristic of the device, regardless of its age, usage etc. The other part, i.e. leakage currents, can be more variable and be influenced by e.g. environmental and storage conditions, and aging of components.
With this in mind, let’s dig into this meter for a little bit to get a general feeling for where the potential problem(s) might be. Thanks Phillippe-Georges for giving me a crucial hint for opening the meter – I had pried around the sticker on the back earlier, but it wouldn’t budge, so I had given up on that route. Turns out it was the entrance nonetheless:

I’ll post some pictures from the inside of the meter. When thinking about this issue, I Googled a bit, but didn’t find any interior shots of this type of meter. I’m sure they’ll be useful to someone out there, some day. (In case that’s you, reading this 8 years from now as you’re working on one of these that happened to have a leaking battery in it – here you go buddy!)

This is the first part we come across; it’s actually the amplifier board for the photo sensor. I suspect Gossen decided to make this a separate board mostly for noise issues. It’s spacious and seems to be laid out to minimize leakage currents and noise in the high-impedance circuitry.

The amplifier/signal conditioning board connects with a single-row pin header to the mainboard; in the photo above I’ve lifted away the analog signal board to expose the mainboard and CPU underneath. Based on general look & feel and component choice, it looks like a design from maybe the early 1990s – and it would have been modern for its day.
What’s immediately apparent – and actually something I was kind of expecting – is the use of tantalum electrolytic capacitors in many places. This, in fact, is the prime candidate for the battery drain problem. I’ve circled a couple here; there may be one or two I missed, also on sides of the PCB’s not shown:

I poked around a bit with the multimeter here and there and measured 5V on some of these – a common logic level power supply back then (and even today). The question of course is whether the Gossen people actually kept those tantalum caps powered up in the meter’s power-down state.
The circuit board is evidently too complex to trace quickly, but I did decide to do a little more poking around. Based on my own DIY projects, if I had done a power scheme as used here, I would have used something like a P-channel MOSFET as the ‘mains switch’, with its gate pulled up to battery voltage with a high-value resistor. Then have the physical button pull the gate to GND so that the MOSFET conducts, which allows the CPU to wake up and then have the CPU keep the system on even as the button is released. This is actually a simple, common way to do this today. The question is: did the Gossen people do it this way, too?
P-channel MOSFETs are common now, but they weren’t in 1990, and insofar as they were used, they might not have been an attractive proposition. I’d have to look into it, but I wouldn’t be surprised if they had high resistance even when fully on, maybe high leakage currents as well, and perhaps other disadvantages, too (reliability, cost!?)
I took the obvious route and looked for how the measure-button on the side (which turns on the meter if it’s in power-down state) relates to the circuitry on the mainboard. It took some fiddling around, but I found it after a while:

The measure-button is in the semi-circular bulge top left of the photo above. Its button contact traces to the top of four wires soldered onto the main PCB. On the rear of the PCB (well, actually the side that faces the user if the meter is held in their hand), a trace runs to a 1k MELF resistor to the left of the CPU. The other end of that resistor has another trace diving underneath the CPU and ending up at one of its pins at the top.
At this point, I stopped looking, because it gave me enough of an idea of what’s going on. The concept is apparently that the CPU remains powered on permanently, but probably in a low-power state. In this state, it can be ‘woken up’ by the press of the measure-button. At this point, the CPU will likely start whatever peripherals necessary (such as the LCD) making the system come alive.
The essence of this mechanism is that the CPU is basically powered up all the time – although its actual power consumption is likely kept very minimal: probably in the nA range. A cheap microcontroller today easily achieves something like 10nA current in low-power sleep mode; I expect something comparable would have been feasible around 1990 as well. This would effectively give the meter an infinite battery life as long as it’s turned off – even if the CPU is actually powered on in reality (or a small part of it, at least).
If not for leakage currents. And this takes us back to those tantalum caps. These are used as buffer capacitors around the board to ensure the electronics (including the CPU) have a solid/dependable power supply even under conditions of momentary current peaks etc. And if the CPU is kept alive even in power-off state, this means at least some of the suspect tantalum capacitors will be powered on just the same.
Why is this a potential problem? Well, in the order of things in the cosmos, there’s some ground laws, such as: we all die some day, until that day we pay taxes, and tantalum capacitors start to leak at some point. Tantalum capacitors look(ed) great on paper: they have high energy density and so they’re very attractive if you want to build a compact device – like a light meter. So they started to be used en-masse especially from the 1980s onwards. By today, we know that ultimately, tantalum capacitors all fail (and generally much earlier than e.g. aluminum electrolytic capacitors). The failure mode is that they start to leak current, and once this begins, the process is self-reinforcing and you may find tantalum capacitors leaking dozens of mA’s virtually all of a sudden.
I know, because I spent quite some time analyzing a situation like that not too long ago!
So, now what? As remarked at the start, I haven’t noticed any problems with my own Gossen meter. So frankly, I left it at this since I didn’t feel like meddling with a perfectly functional bit of equipment. But hearing that some of these meters have developed batter drain problems – it comes as no surprise knowing what’s inside them.
If one of the battery-hungry meters was mine (or somehow became my problem to deal with), I’d probably start by simply replacing those tantalum caps and then see if the problem goes away. Am I 100% sure that this will fix it? Nope, but the odds are big enough to try that before spending lots of time on further diagnosis.
As a bonus, here’s some more shots of the analog board to mull over. There’s some nice components on there I don’t think you’d find anywhere anymore, today. The metal cans, for instance, are double/matched NPN (MP311) and PNP (MP351) transistor pairs, and the black TO92 case in the middle is an AMS9491 voltage reference (1.235V). Note the neat guard trace around one of the pads & pins in the bottom right corner in the photo below – nicely done!

Not to mention a CA3240 MOS opamp – probably the bees knees back in the day.
