One issue I found with the magnetic paddle wheel sensors is that they may attract and hold small bits of iron from the water. Eventually, they stall and have to be cleaned out. We put a strong magnet trap upstream to minimize this.

 

Would it not be possible to use some form of opto-electronic sensor placed on the face of the meter to detect the movement of these segments - they tend to be very high-contrast.

I second this recommendation. On my water meter (Neptune), there is a small red triangle that spins to indicate water flow, as PhilTilson mentioned, it's very sensitive. A while back, as just a proof of concept, I put a proximity sensor (VCNL4040) up against the clear cover over the gauge and was able to detect the movement of that triangle using the proximity count values. I was able to use the "logic" mode of that sensor, set the close and away count registers, and then simply monitor the sensor's interrupt pin with a microcontroller to determine flow. That could be a very cheap solution if your meter has something you can watch as well.

 

The water meters used for billing purposes in municipal water systems are positive displacement devices that make a clunking sound as the water flows through them, at least where I live that is the case. And any vibration that can be heard can be picked up and amplified and used to drive a counter. I was NOT talking about any of those turbine devices which are much more sensitive but not accurate over nearly as great a flow range.

 

In post #1 the TS mentioned "not expensive", and while there are indeed a whole lot of arrangements available they lack that quality. A spring loaded low pressure relief valve plus a differential pressure gage, transducer, or even a switch, will serve the purpose and be both simple and cheaper, and also less complicated.

You seem to have missed the point.
The original request was for a flow meter at reasonable cost. The person asking has already tried cheap flow "meters" than don't work, yet you are suggesting something else that probably won't work. More money wasted.
I have more than 25-years experience in the water and waste business and have never seen your suggestion employed in flow measurement. If it was so good, it would be commercially available. It would be interesting to hear how you would calibrate and then present usable results.
A secondary displacement meter would do the job, is made for the job, without messing about with gauges, relief valves etc.
As for clunking noises, I've never heard that, unless you have some very poor plumbing.

 

You seem to have missed the point.
The original request was for a flow meter at reasonable cost. The person asking has already tried cheap flow "meters" than don't work, yet you are suggesting something else that probably won't work. More money wasted.
I have more than 25-years experience in the water and waste business and have never seen your suggestion employed in flow measurement. If it was so good, it would be commercially available. It would be interesting to hear how you would calibrate and then present usable results.
A secondary displacement meter would do the job, is made for the job, without messing about with gauges, relief valves etc.
As for clunking noises, I've never heard that, unless you have some very poor plumbing.

Measuring the pressure across a relief valve is certainly a NON-STANDARD approach, it does work to detect low flows, but no, it is not an accurate way to measure flow, and I never said that it would be a measuring system, just a flow detector. An instrument class device is an entirely different thing.
And since I can feel and hear my water meter working, a vibration sensor should work to detect flow. All positive displacement devices seem to have parts that produce a bit of vibration, at least as they age. And most folks who experiment with electronics would have on hand the items to learn if the magnitude of the vibrations were usable. The pressure transducer and check valve have been part of a commercially successful leak test system that has been sold to users for at least 20 years.

Of course you have never seen it, it is not used in water and waste applications.

The TS was looking for a way to detect leaks when everything should be "off", at least that was the initial statement.
The cheap devices described suffered the inability to detect the low flows that the TS needed to detect. The existing water consumption meter was entirely adequate for knowing the exact amount of water used.

 

I know this thread is old but for the purpose of anyone who comes across it I found a solution that was so good it was beyond all my expectations. It achieved an absolute 100% accuracy with the water meter, was able to detect the slowest of flows, was cheap, really simple to install, required zero plumbing and no additional electronics. The solution was to go down to the local plumbing supplies and buy a reed switch. There is a rubber cap on the existing water meter that you remove and the reed switch fits right in. It comes with a screw to hold it in place. This just wired straight to my existing Wemos. You need to find the right part no for the reed switch that fits your water meter but apparently there are only a few water meters installed in Australia and there is a reed switch available for all of them. The solution couldn't have been better really, any external sensor will have some measurement difference from the original water meter. As I only really care how much water I'm being charged for it's effectively 100% accurate right down to the slowest of flows.

The downside is the resolution is one pulse per 2.5L and the water company could potentially remove it in the future if they go to a smart meter. But I have been using it for 2+ years now and no issues at all. It has detected multiple leaks which have just been taps left running slowly or leaky taps. The interesting thing is the eBay meter isn't that bad, it's anywhere from 0 to 8% out over a day but has much higher resolution. This comes in handy for graphing flow rates. The standard water meter is very choppy for flow rates due to the poor resolution.