Really confusing campaign. My understanding of low power electronics was that it enables battery powered equipment, not that they save the planet (in the end batteries are a necessary evil). If we want to save the earth from CO2 gasses, then we should probably all stop driving a non-electric cars and stop using coal or gas in power plants.
Moreover, I don't think much people will do a low-power optimization cycle for anything non-battery powered, unless there is a heat issue. However, heat issues are often resolved with better cooling.
The "power analyzer" looks like a very similar to a project I am working on:
https://www.eevblog.com/forum/chat/what-are-u-working-on-right-now-%28electronics-related-and-with-picture-please%29/msg331101/#msg331101Shunt is 10 ohms (0.1%). I use the AD8253 INA, which is programmable gain 1-1000x (per order of magnitude).
At 1000x gain I get 50nA/LSB. At 100x gain I get 0.5uA/LSB. The AD8253 has 4MHz -3dB bandwidth of 550kHz at G=100. The noise (because I forgot an anti-alias filter
) is about 10LSBpp. So 16-bit / 10LSB = 12.7bits roughly usable bits.
INA gain error is typically 0.04%. As the shunt is 0.1%, and I have a resistor divider (0.1%) to connect the front-end to the A/D (which is a bit of a pity to throw away so much gain), I get roughly 0.24% DC accuracy.
Unfortunately the MCP3911 has a PGA gain accuracy of 4%, and it's auto-zero feature (which can't be turned off) screws the offset cal up every reset. So it has to go in Rev2.
However, it does work:
https://dl.dropboxusercontent.com/u/207647/PA/330k%20ohm.pngConnecting a 330k ohm resistor (3.3V rail) at 1000x gain over 10R shunt. Nice 50Hz pickup / body resistance. No filtering done. I still think this image is quite noisy.
100x gain is probably the most usable range.
https://dl.dropboxusercontent.com/u/207647/PA/Noisefloor gain 1000x.pngNoise "measurement" at 1000x gain. 0dB = FS(1300uA).
Not sure whether I should open source this project.
Anyway, I really wonder what kind of performance this probe/analyzer achieves. It says it has a dynamic range of 1:40000, which sounds to me 15+ bits of resolution, but at the same time the DC accuracy is 2.5% (order of magnitude worse than my power analyzer, or several orders of uCurrent Gold).
So 1:40000 is way overkill for an 8-bit or 12-bit scope resolution measurement, but at the same time measurements can be off by 2.5% for a multimeter.
Am I missing something?
The picture shown could be from a single low-ohm shunt, but with different gain front-ends. You can clearly see the noise at the microAmp graph, not so much on the mA chart. So as one clips you can view the other one for a large scale picture, and use the other for low-current measurements.