Hacking a Kill-a-Watt for very low power measurements

[ICanHas]

Have you even used a power analyzer? Sampling type digital instrument from that era are excruciatingly slow and quirky when any two of voltage, current and frequency are fluctuating. It gets easily thrown off and drop out like a digital TV under poor reception.

Pay more attention to the definition of the word "must".

[NiHaoMike]

I actually had to design a digital power meter as part of a class project. Even on something as basic as a dsPIC (much slower than even a low end ARM), it's not very difficult to make the code work in real time with a 250kHz sample rate. The antialiasing filters are around 20kHz, giving it plenty of bandwidth for mains power measurements.

I do agree, however, that equivalent time sampling holds lots of traps for unsuspecting beginners. Fortunately, no knowledgeable engineer (nowadays) would use equivalent time sampling at such low frequencies.

[ICanHas]

with nonlinear loads, a multimeter must have DSP in order to measure properly.

It's still possible to do it in analog

Finally, an implied agreement that your earlier statement is wrong.

but it's just not cost competitive with digital nowadays. On a fixed budget, digital does way better than analog for this application.

Cost as seen from which point of supply chain? I agree that the starting manufacturing cost to "make it work" is lower. It certainly would be difficult to offer a multimeter that retails for $6 using analog that approaches the accuracy of $6 DMM. Digital micro controller embedded systems is without doubt cookie cutter automated assembly line production friendly.

Regardless of raw materials and processing cost, something of PM300 caliber digital power analyzers with flexible averaging time are a far fetch from "cheap".

It also allows features like long duration averaging that are difficult to do in analog. In particular, when dealing with time constants longer than a few seconds (plus high accuracy requirements), leakage currents become a big problem when designing analog circuits.

Reference? This is something I have to study more, so I need references. Digitally integrated values have its own problem with wrongly rejecting signal as noise or valuable data misses the train.

Gain in weight of wood by leaving it at a higher temperature and comparing weight at two points is a result of comparing the value between the analogly accumulated values between top and bottom bounds of integration limits.

The gain/loss of moisture is naturally done with an infinite resolution, but digital means of will probably result in accumulation of errors.

[NiHaoMike]

Maybe I should just say "good quality input filters"? Granted, doing it in analog does require a lot of precision components, making it a poor choice for portable instruments where space is at a premium and where the temperature can vary widely. Doing the bulk of it in digital simplifies the analog filter requirement to to a simple antialiasing filter, with the remaining orders done in the digital domain.

Try building a purely analog circuit to measure the average power draw of a load that cycles or otherwise varies over a few hours. Don't be surprised if leakage currents frustrate your efforts. The simplest way I can think of (sticking to analog) is to advance a mechanical counter, as in an analog watt hour meter, then keep track of the time with a mechanical clock.

[ICanHas]

Those features aren't available on consumer market stuff anyways. The instruments that have them come with a hefty price tags anyways.

Maybe I should've said entirely analog front end with digital use limited to something that is not of the "embedded systems" type that contains proprietary embedded software that raises the PRICE by forcing the purchase of hardware that's unusable without purchasing the software license already flashed into the product.

You continue to drill cost the in the same context as what it costs to press per disc while there's only a sparse connection between the cost to press a disc vs the retail price of a movie or a software.

In days of analog instruments that only used digital for the user interface output, the "optional" software provided functions similar to EMBEDDED SYSTEMS FIRMWARE non-sense often listed for several hundred dollars. Sometimes the deluxe version was more expensive. This is hundreds of times the cost of the 640K diskette, ink, paper and staples and box that forms the software kit. The "cost" you speak of is that of these materials and is only relevant to OEMs, not us.

You're simply made to buy a copy of that software that's pre-flashed into the unit with "embedded systems" design, so the price will still be high

[NiHaoMike]

The super cheap (for what it is) Kill-a-Watt does have the ability to average power over what might be a period of days.

There are also quite a few under $500 DSOs out there that could be used to measure power. You do need to get an isolated voltage probe and a current probe to measure mains (or an isolation transformer/other isolated source of AC power), but that's a hardware upgrade.

[ICanHas]

The super cheap (for what it is) Kill-a-Watt does have the ability to average power over what might be a period of days.

As can pulse counting using pulses, although that instrument is susceptible to coasting. Don't get me wrong. I love mines. I've got three of them. They're very good for the limited functions they do and I use it WAY more than my power analyzer, but the claim that "must have DSP to measure power" is unacceptable.

There are also quite a few under $500 DSOs out there that could be used to measure power. You do need to get an isolated voltage probe and a current probe to measure mains (or an isolation transformer/other isolated source of AC power), but that's a hardware upgrade.

And why do I get this feeling that the royalty licensing share of pricing for firmware/code/developer fee, blah blah portion of embedded craptem is more than the COST of the probes?

[jeff5may]

Have you ever thought of writing your own programs? This route is very dollar cheap, and can work on a huge number of devices. No one said you have to buy a brand x super hyper combobulator with embedded digital gimmicks every time you need new functionality. If major industries didn't work out their own problems by inventing proprietary intellectual properties, the world would be a much less interesting and more expensive place.

I'd love to tell you it isn't that difficult, but I hate to lie to people. Starting with nothing and conjuring up some code to accomplish something precise is usually not that easy. It takes sweat equity, skill and determination. The success of a program well written, that actually does what you want it to, is a great confidence booster.

If these virtues are in short supply in your realm, feel free to spend money to have someone else do it for you. If you're weak and poor, learn and grow stronger: you have no choice. If you're rich and lazy, get up off those benjamins and spend some. Just don't complain about it when you feel like you got robbed.