Skip to main content

You are here

Trouble reading mechanical power meter

48 posts / 0 new
Last post
Trouble reading mechanical power meter

I have been following along with the Jeelabs blog for a fair while now and wanted to finally dive in and have a go at replicating some of the power monitoring projects that jcw has completed.

I am trying to read power off my houses electricity meter which is the kind with a rotating disk, there is a glass cover over the meter which I don't not want to remove which I believe makes looking at the dashes on the face of the disk next to impossible.

The edge of the disk (probably 2mm thick) has a short black region once per revolution and I was kind of hoping I would be able to sense that.

To start testing this I bought a few KTIR0711S IR sensors. However before even heading outside I am running into the issue of when connected up as per the schematic in the datasheet I am only getting a 20mV or so swing when going from nothing in front of the sensor to a white sheet of paper.

Looking through these forums some other members have managed to use QRB1134 sensors, before I try and track down where I can get one in Australia without costing $20 shipping I wanted to get peoples thoughts on how well it would pick up the black mark with comparatively small area on the edge of the disk.


Hi Hugoagogo,

I have done a few experiments with a similar meter without the success I hoped for. It is possible detect some of the time but I haven't found it good enough to implement yet. I think the glass in front of the wheel gets in the way and using IR make it hard to "see" the focus points etc. I also wanted to run on batteries with a low power solution. I have wondered about using the internals of an optical mouse but nothing yet.


Hi, All. Wanted to do a similar thing. Two things coming to my mind:
1) Is the beam of the used LED is narrow enough? It is not for nothing the lens in usually used.
2) To decrease frequency of checks and save battery, I was thinking about reading the grey level of current digit on second and third number disk, not a narrow disk. Every time the level changes, it should indicate a number change.

Any thoughts?


Yea I dont think I can monitor the numbers as mine has multiple small dials (not this exact one but like this

I was thinking I might be able to use a LDR and a few leds to provide illumination instead of the infer-red, as i think it may have a slightly longer working distance. I am going to make up a little test rig so i can play inside without having to keep heading outside

Has anybody tried using this visible light option (also the leds could be turned on and off to save power)


Hi Hugoagogo,

Indeed, I have a "proper" old style meter with large dials and one number per window.

Another option [which was in my plans to test]: every piece of cable has its own resistance, although sometimes pretty small. Yet the flowing current should drop some millivolts on that even short cable. If you could measure the cable resistance pretty exactly, it is easy to calculate the current.

Otherwise you could just wait for a period of higher load, capture both voltage and changes in meter counts. This would allow you to approximate the used power.

And, as you know, JeeNodes with battery power should be just perfect to implement such a voltage logger.

When one does not require the 100% accuracy (I doubt if everybody actually needs it), it looks as a feasible solution.

Also we could design a correction step (once a month), when the real meter readings are checked against the calculated one. A correction coefficient (like x1.0001) could be introduced then.

PS: Sorry could not try it myself yet because waiting for deliveries.

PPS: Just came to my mind, that it would be nice to capture the exact voltage too. Should design a voltage divider then.


That sounds very scary with mains voltage. A safer way would by a current transformer clipped to the cable. These don't give meter accuracy but I you live to see the results.


Hi JohnO,

Indeed, this is the reason it is in my "fallback" list.

Good idea about a transformer, but I'm thinking about actually use it in a different direction - to INCREASE the voltage deviations. Remember, that the suggestion is to connect in PARALLEL to a existing SINGLE copper wire inside your meter installation.

Or you were talking about measuring the mains voltage, not current ?



Another thought closer to your idea:

Have you tried to vary the angle of both your IR LED and the photoreceptor ?

It could be a case that the major portion of light coming to photoreceptor as reflection from the glass, not as reflection from the rotating wheel. Would explain the low signal you are getting and changing the angle might help.


I'm talking about using a current transformer clipped around the AC live cable. Make assumptions about the voltage - or measure it. Then calculate the power used in a similar manner to the many commercial offerings. For instance:


Sorry about the long delays I have been pretty busy with work.


I cant change the angle as mine is a little package where both are integrated into the same part. I may try going and buying some individual parts and seeing if they work better. Also videos like this are not exactly encouraging,

Also I am not very keen on using the current shunt or clamps, I would prefer to stay away from having to deal with the mains power.


CT clamps don't involve actual connection to the mains power. This is the typical way to measure this stuff but I would like to see the same reading as the meter - like you I guess.


It might not be a direct connection but it is less accurate and involves running new wires near high voltage lines (which are not easily accessible in our meter box)


Dear Hugoagogo, Not everything is easy, indeed. But there're some physical limitations.

Of course I would suggest you then to ... stop bothering with Arduino, install a simple wireless webcam and program it to take a picture of meter every now then. And use a good OCR (like your own brain ;) to convert some of them into numbers...

PS: I'm going to play with CT soon - make a small induction sensor from couple of wire rounds on a power cable, put diod + resistor and see what is the output.


@PavelE - a C.T. solution holds promise (that's what commercial power meters in the 100KW and above range use) but from the 90 degree current/field relationship, coiling the wire around the wire is ineffective.

Placing coils around the wire will work. If you don't want the amplification effect of using some iron core material, you can look at the air-cored Rogowski method but be prepared to deal with accurate but truely tiny signals.


@martynj Thanks, this is why I'm going to test [which construction would be the most effective, with or without coil, how many wire rounds etc]. But, we are talking about currents of single Ampere and I only need some tens of millivolts to be inducted.


Just tried to measure the signal from wrapping one wire around the power cable (approximately 30 rounds) of the boiler (10A current). Signal was much less (more than 10 times) than noise measured about 1-2 mV. So my idea does not work, now need to try the CT as was recommended.


Agree. Giving up with building my own CT - would just get one from EBay.


I've just found this: perhaps I'll try again to 'watch' the wheel spin.


Thanks for that link, I am in the middle of exam week now but I have 10 of those TCRT5000 sensors making their way to me to try, all for the very affordable price of $1.80 including shipping (ebay).


Now I understand you... You just cant' resist them ;-)

But this parallel setup requires a reasonable lens/condenser to direct the beam.

My idea was much more simple 1+1(or 2+2) sets of diodes (LED+Photo) and a removable board, which I could drill holes in. In such a setup it is very easy to experiment with different directions/angles.

I wonder how effective TCRT5000 are on the extra low current. Maximum 200 mW (100mA) according to specs is a lot ... With distances of 1-2 cm, do you really need that much power ?


In can't see why it shouldn't work, you just need to use a higher value resistor for the pull up, the values you were looking at were absolute maximum.


Yesterday after a good idea of JohnO I've finished my design for reading meters: CT for current, Hall-effect for Gas and water. Now starting the implementation.

@Hugoagogo Could you try with your mobile phone (most of them have a compass application) to see if there any changes in magnetic field when the disk rotating in your electricity meter? When there are, it is pretty easy to measure using a Hall-effect sensor and pulse counting.


My latest failure to 'see' the black marks on the spinning aluminum disk in my electricity meter is just completed. Tried using the TCRT5000 triggering an interrupt rather than reading an analogue value. The TCRT5000 works well on the bench, particularly detecting white surfaces but I get nothing through the glass window of my meter.

It gave me the chance to experiment with two pin change interrupts on the T84 at the same time. I used PCINT0 & PCINT1 to make it easy. It is quite surprising what the T84 can do in 4.4k bytes.

I will probably be cheaper to have my meter replaced than all these failures!

Ah well, you live and learn a little more.


@JohnO Why do you are not considering CT yourself? There is a chance, that a new meter unit would be digital and not necessary better than the current one ...

PS: If you would go for several CT, it could make sense to get many of them to save money for several people (like including myself ;-)


Don't all the digital ones also include a flashing LED?


My point was more about "do not change what is not broken". At least I'm not going to change my meters till absolutely required to do so or would find that they are very inaccurate ...


I already have some CT's but have a yearning for meter accuracy. I also don't warm to sampling solutions when digital accuracy appears only a step away. I also guess its fun to play.


I hear that they do - I don't think I'll really switch meters - maybe move house instead ;-)


Hmmm - the siren song of "accuracy". I'm sure you will eventually crack the problem of peering through the glass and count those disk rotations with great precision. But how accurate do you expect that analog meter to be? Worn bearings, fading drag magnets, sticky gears presenting a variable load...

But wait! Aren't there regulations protecting the consumer from inaccurate metering?
Hmmm - yes and no. Yes, often a tight specification that power registered must be within 1-2% of power consumed. But with wear, how to keep meeting the rules?

  • Ignorance - which consumers are capable of measuring true consumption to 1% and challenging the Utility company?
  • Forced skew - since the regulations are really about overcharging, skew the initial calibration so that it reads low by the allowed error band. As it drifts out of calibration (usually by under reading further anyway) you have the whole error band to traverse before risking an over reading. Lost revenue? Nah - push up all prices 2% to compensate....

Bottom line - it is unlikely that even the most careful observation of the Utility meter will represent actual consumption without several percent error. It is a tool, built by the lowest cost bidder, to extract money on a pseudo-equitable basis; not a lab-grade instrument.


Ah, but depending on the purpose of metering at the consumer end, the accuracy may be important: in the end, JohnO gets billed based on the readings off his meter; and no matter how (in)accurate that meter is, he can exactly monitor how much power he will be billed for by accurately tracking that meter.. :')


Premium Drupal Themes by Adaptivethemes