Agilent Seamless Current Measurement - any ideas ?

[dmage]

Hi.
Do anybody have ideas how it is implemented and can be implemented in lab ?

[free_electron]

here you go : read to your hearts content ...

US patent 8,037,327

Patent viewer ...

[dmage]

here you go : read to your hearts content ...

US patent 8,037,327

Patent viewer ...

wow, thanks!
now will read all this stuff ...
come back later with ideas )

[ejeffrey]

So the idea seems to be to measure over a large dynamic range without auto-ranging?  The point being to accurately measure the power consumption of microcontroller/microprocessor that is switching between standby/sleep/full power modes.

The patent link appears to  require a plugin I don't have, so I don't know how agilent does this.  One approach is to simply brute force dynamics range by using low offset, low leakage amplifiers and high resolution ADCs.  This is probably the best way to do it if you want the best possible linearity and accuracy.  Clever ways to do this would be with logarithmic converters or analog charge integrators.  Logarithmic converters have amazing dynamic range but mediocre accuracy.  Charge integrators can give you good accuracy but don't tell you the instantaneous current waveform, only the total energy used in a defined integration period.

Another possibility might be to have multiple ranges amplified digitized in parallel, and have software pick the one that is most accurate for any given conversion point.

[dmage]

So the idea seems to be to measure over a large dynamic range without auto-ranging?  The point being to accurately measure the power consumption of microcontroller/microprocessor that is switching between standby/sleep/full power modes.

The patent link appears to  require a plugin I don't have, so I don't know how agilent does this.  One approach is to simply brute force dynamics range by using low offset, low leakage amplifiers and high resolution ADCs.  This is probably the best way to do it if you want the best possible linearity and accuracy.  Clever ways to do this would be with logarithmic converters or analog charge integrators.  Logarithmic converters have amazing dynamic range but mediocre accuracy.  Charge integrators can give you good accuracy but don't tell you the instantaneous current waveform, only the total energy used in a defined integration period.

Another possibility might be to have multiple ranges amplified digitized in parallel, and have software pick the one that is most accurate for any given conversion point.

hi and thanks. that is almost what im thinking about.

the task is to measure a wireless sensor current while in sleep/work modes accurately, but i not shure do i need emulate seamless technology or similar.
Sensor power goes from 0 to 5V and the peak current 100mA. So, yes we need accurately measure a 0-100mA range with great accuracy in standby time.

now im thinking of an like a BeagleBoard-xM addon board with an 1-4Msps 18bit Analog Device(or TI) ADC with level translators, wich goes directly to the microSD (brute force logger) + ethernet => matlab(gui or just get() function).

the point is: do i need like uCurrent like circuit with the ADC's in parralel with the 2 channels 1-100ma and 0-1ma ranges and then let the matlab do it job ?
if so, can i work with this adc's with 48mhz spi master from the board (will it be fast enough) ?

[Rufus]

The patent link appears to  require a plugin I don't have, so I don't know how agilent does this.  One approach is to simply brute force dynamics range by using low offset, low leakage amplifiers and high resolution ADCs.

I don't have and won't install the plugin either.

Remember this is a source meter so they don't much care about voltage burden. They could use multiple current shunts in series, measure them all and take the reading from the highest value shunt that isn't over range.

They probably clamp the higher value shunts with a diode (or perhaps something smarter) to limit burden and dissipation.

I looked at the specs for one of their models and it is only 'seamless' on the highest 3 of 4 current ranges.

[oPossum]

Patent at Google Patents - no plugin required

[free_electron]

plugin ? they are just tiff files ...

But yes, the gist of this system is that they are using an active load ( sourcemeasure unit )
they program compensation tables and then run those through fast dacs.
what that machine is doing is make an artificial 'zero' point. they don't measure the current directly. they have a reactive negative current generator. and they just feed you the output of the regulation loop.

same principle that a picoampere meter uses, but you need a fast suuply with lots of 'oomph' behind it ... this is not something you breadboard with 2 opamps and a powermos

[Rufus]

But yes, the gist of this system is that they are using an active load ( sourcemeasure unit )
they program compensation tables and then run those through fast dacs.
what that machine is doing is make an artificial 'zero' point. they don't measure the current directly. they have a reactive negative current generator. and they just feed you the output of the regulation loop.

I would say that patent has nothing to do with this seamless current measurement.

This one http://www.google.com/patents/US8115474 looks much more likely and is similar to my suggestion except the current shunts have a common ground and the clamp limiting the voltage across (and so current through) one shunt diverts the balance of the current to the next shunt. The voltage across the shunts is scaled according to the shunt value and summed to obtain a total current.

[free_electron]

could be ... they may also be related ...

that second patent is useing mosfets as an 'adjustable shunt'.

in essence they are tracking the mosfets to get a certain voltage across them. the tracking is done by a control voltage. they close the loop. so the control voltage is a ctually a direct reflection of the curren through the mosfets. they just 'adjust the mosfets' to always have a fixed drop across them.

interesting technique ...

[dmage]

thanks for the answers!
will contunue studyng the problem basys...