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

[NiHaoMike]

In a digital multimeter, there's already an ADC and no need to convert it back to analog. It is possible to do it in analog and I actually have opened up an old Fluke that worked that way, but that was back when computing power was expensive. Now, it's cheaper to do it in digital given you need the result in digital.

There's a good reason the cheap Kill-a-Watts use DSP, as do the somewhat better Watts Up meters and even the top of the line Gossens, Flukes, and Agilents. At the low frequencies multimeters operate at, the "DSP" doesn't even have to be a proper DSP. A cheap 8 or 16 bit microcontroller (with clever programming) does the trick. ADCs that operate in the 10s of kHz sampling rates are dirt cheap thanks to their extensive use in audio applications.

[ICanHas]

But your claim that "Additionally, with nonlinear loads, a multimeter must have DSP in order to measure properly." is wrong. There is no conversion to analog. Are you just parroting? The four quadrant real time true power converter provides analog output proportional to real-time watt. There is an apparatus called an integrator which takes the real time true watts information and continuously output the result as an analog signal, which can then go into A/D converter for display reading. Integration is a the process used for averaging.

NHM,
The width of averaging frame is continuously moving forward or it can be anchored down at the starting point and expand continuously. The width is known by the mathematical technical term call the limits of integration. The operating characteristics of integration can be changed with a device such as a potentiometer, an analog apparatus that can vary the resistance by turning the shaft, which then analogly tell the integrator what to do without the quirkiness or lag of digitally cheapened digital PIC software driven microcontroller.

Those cheapo China sourced sampling meters are good under limited conditions and provide pretty good accuracy over a narrow operating envelope.

I think NHM has financial interest ties in pushing open source hardware coupled with royalty costing software.

Analog data is difficult to use for further analysis, so good instruments use a combination of analog front end with digital back end. Analog watt converters and integrators work continuously and produce accurate result even if the load behavior changes every wave cycle or the frequency tend to fluctuate.

With high performance analog front end with digital chart recorder, you can change the sampling rate on chart recorder as needed but real time wattage is always available. This ain't the case with those digital crap.

Plot out the cumulative joules consumed by the motor during the start-up cycle, and the instantaneous power from 0 second until reaching 3450rpm 0.25 seconds later.

Many devices use standby power in bursts and this type of use is also difficult for cheap sampling based devices.

To say "must have DSP" is ludicrous when it is just one of the ways of doing it especially when analog front-end has superiority in many ways. The absolutely the worst one in terms of real time reporting and latency is the virtual DSP that is based on program code... the kind you like to push.

Works good for a steady slow responding loads, but they totally suck for real time for unstable voltage, frequency and wildly varying loads or when you need power reading in real time, such as power input during acceleration or sudden motor loading.

[NiHaoMike]

What's the point when you can use the ADC directly and then do the processing digitally? The higher end multimeters even do the calibration digitally - it allows a completely automated calibration jig to be used rather than have a worker turn a bunch of trimmers.

The space usage of all those extra analog components is not insignificant in a portable device. Contrast that to a slightly more powerful processor that most likely comes in the same package size.

At high frequencies, it does make sense to do a little more in the analog domain, but most multimeters are only rated to a few kHz - hardly "high frequency" by today's standards. Theoretically, the sample rate has to be twice the highest signal frequency to get a good capture (more like 4-5x in practice), but even with a generous 10x oversampling rate, a multimeter with a 10kHz bandwidth only needs a 100kHz ADC - right where the commodity audio ADCs are.

That said, Bob Pease did develop a pretty good analog wattmeter.
RAP's Multirange Wattmeter | Test & Measurement content from Electronic Design

[ICanHas]

What's the point when you can use the ADC directly and then do the processing digitally? The higher end multimeters even do the calibration digitally - it allows a completely automated calibration jig to be used rather than have a worker turn a bunch of trimmers.

Extremely DIY unfriendly, highly non versatile, low cost mass production cost disposable, proprietary China made consumer electronics way.

The art of digital bullcrapping is best friend of China made crap. You could make a ruler where each marker is about 0.1mm apart, but built so sloppily that each division can vary as much as 20%. you apply digital crapping technique of telling 1 count = 0.1mm, 2count =0.22mm, 3count =0.31mm, 4 count = 0.5mm. So, by marrying the factor of BSing for each count, you can get accurate reading as long as the sloppy built piece is stable. Now they're making cameras with slop China made soda botle grade glass and uses embedded pooptems post process lens flaw compensation.

These were good technique used to improve highly precise input even better but now they're used as excuse to make sloppily made crap. The new export law in mainland China that lets basically anyone export is causing a domestic side proliferation of uncertain quality, high sample-to-sample variability crap instruments.



"Additionally, with nonlinear loads, a multimeter must have DSP in order to measure properly." You're begging the question. I am contending that your claim is wrong. Do you accept that your claim that power measuring being conditional to having DIGITAL processing is wrong?

When you need rapidly responding real time power monitoring for non-steady loads, Kill-A-Watt and just about every other China engineered crap suck.

It is fine for stable power draw like a stabilized fan motor, light bulb,etc.

[NiHaoMike]

Take one of those cheap power inverters and try to measure the output voltage with a multimeter. You'll find that only the good ones with DSP (or the old digitals with the really good analog front ends) will measure properly. Not even a classic analog meter will read properly.

It is possible to do it in analog (as those old Flukes show), but in practice, a modern meter is going to use DSP. It's simply much cheaper to throw in a processor than it is to mass produce a bunch of precision analog circuits.

[ICanHas]

Take one of those cheap power inverters and try to measure the output voltage with a multimeter. You'll find that only the good ones with DSP (or the old digitals with the really good analog front ends) will measure properly. Not even a classic analog meter will read properly.

But DSP is NOT a requisite, it's just ONE of many ways. Analog or digital back-end is not relevant. An ordinary needle meter or a digital voltmeter can measure the proper value sought after with the proper front end depending on what you're looking for. If you're looking for RMS value, you use an RMS converter or if you're looking for average, you use a circuitry known as the integrator.

Integration is a mathematical function used to gather the cumulative quantity (collected quantity). The 1 and 4 shown in pictures are called limits of integration or a span of 3. A divider is used to continuously divide the collected sum to produce the average value.

If each shaded square meant a joule and there are 150 squares in there and the units for the x value (the x is the side-to-side axis), expressing it as "watts" means it is divided by 3 so it will read as "50 squares/second" or 50 watts. An analog integrator can provide the value continuously.

It is possible to do it in analog (as those old Flukes show), but in practice, a modern meter is going to use DSP. It's simply much cheaper to throw in a processor than it is to mass produce a bunch of precision analog circuits.

It's a mass production, non DIY adaptation friendly, highly mass production friendly, digital bullcrapping and mass China made production optimized setup.

[NiHaoMike]

Take a look at Arduino. It's a very DIY friendly platform that can be used for basic DSP. It used to be the underpowered (but still very useful) 8 bit Atmal platform, but now they have ARM based versions.

I personally prefer PIC/dsPIC since it is far more diverse than Arduino, but it is harder to use. Arduino is still what I recommend for beginners.

[ICanHas]

You're all about the blank hardware's cost this, cheaper materials, open source hardware. Your discussion about programmed ATE calibration discussion and "clever program" driven devices suggest you're advocating mass production friendly platform to host a closed source encrypted firmware that uses indefinite copyright, intellectual property and NDAs instead to oppress innovation unlike patents that will become available for everyone to use.

Let me guess, you see hardware like a blank disc which should as cheap as possible and easily China made while using firmware/software/embedded codes as the main product so it can be produced at low cost, then sold at extremely high cost.

Microcontroller based embedded systems with proprietary has been the sinful tactic brand owners have been using to reduce manufacturing cost, lock down high level calibration settings to their authorized service techs and inhibit competition by copyrighting codes. Very common in industrial and laboratory instrumentation. The backbone is fairly common stuff, but they make it proprietary embedded craptems using codes.

One of the reasons with embedded systems is the malice can be built-in, such as worms, malware in order to make it malfunction to botch the process. Build malware into an ECU, so that if someone chose to install a pirated "tune" or the firmware, it will go out of control and crash.

I think that making it look deceptively simple to copy but with a hidden trap that takes a long time to show up (like the MAC address reverting I mentioned earlier) would keep the pirates from putting much effort into making the copy, thus increasing the chance they'll ship flawed copies.


Probably doesn't apply in your case, but if it has Ethernet or other network interface (and is likely for multiple units to be used on one network), have the copy protection revert the MAC address to some default value a few months after it detects a pirate copy. The pirate would probably copy the code as-is (except for the MAC address), sees that it works, and ship it. A few months later, some very strange problems occur as the devices all revert to the default MAC address.

One way to do that is to cryptographically sign the MAC address and hide the signature away from the MAC address. Put the MAC address in an obvious place so that the pirate would go for it and trigger the hidden copy protection mechanism.

[NiHaoMike]

The community can make their own open source software to go with the open source hardware.

[ICanHas]

But you're a proponent of what I described above, are you not?