Multiranged current measurement - multiple shunts in series with diodes?

[daqq]

Hi guys,

I'm designing a testing device that's supposed to, among other things measure the current consumption of the DUT, preferably without switching any relays or anything. The DUT has several power modes, the extremes of these are one where it consumes around 1 Amp, the other is a deep sleep mode, where it consumes around 3uA. Both of these must be tested, to know whether or not there are any leakages, or anything fishy.

The schematic in the attachment seems ideal for this - when the current is in the upper values, the diode for the 1k shunt will activate, since the voltage on the 1k resistor would be well over ~0.6V (or 0.2ish for a schottky diode). A differential measurement would connect the shunts to an ADC with a Mux and amplifier. This would measure the voltages, check which ranges should be ignored and then choose the proper range.

This would limit the shunt power as well as giving enough resolution for all orders of magnitude that need to be tested. Also, no range switching would be involved at all.

Is there any drawback to this kind of solution?

I've only found that the limit voltage of a diode will vary over temperature, but that can be simply fixed by using a lower discrimination limit.

Thanks,

David

[capt bullshot]

You get a quite large variation of the load supply voltage (R3), don't know if this is OK for your DUT
Otherwise, this should work fine. Maybe you can compensate for the voltage drop by using a high impedance sense line.
I've seen this concept in a power supply to provide a high resolution low current measurement, but there the low current shunt gets shorted by a MOSFET if not used, to save the voltage drop and associated power dissipation.

[daqq]

My apologies, R3 represents the DUT - the tested device. Should have been more clear. As to the MOSFET, good idea, should work.

[alm]

The Agilent 663xx series communication power supplies uses FETs to bypass the low range current shunt and has remote sensing beyond the shunts. See snippet of the block diagram.

[T3sl4co1l]

That works, but it suffers from max N-1 diode drops (for N ranges), and each range is only precise up to a voltage drop that's a fraction of a full diode Vf (otherwise the diode current is a substantial error i.e. 0.1% or the like).

Wide dynamic range suggests logarithms.  You could use a carefully calibrated diode and nothing else, and get about 7 decades in one fell swoop.   The readout is already in scientific notation, so to speak, too!

Tim

[danadak]

PSOC has a 20 bit A/D on it with a .1% reference. That's a dynamic range
of 120 db. Note actual performance has to include doing a complete signal
path error analysis. Depends on your goals, absolute or relative accuracy,
PSRR and T effects......


You can do signal averaging and/or DSP on it to improve results.

Very simple to setup, in fact there are projects already written you could modify.





This example is AC coupled, ignore that, you can do DC measurements just as easily.
As well as differential measurements to get rid of ground drop and other common mode
issues. Note the PSOC delsig, when input buffer is bypassed, can achieve 100 mV CM
range outside the rails, perfect for shunt sensing in some applications.


Regards, Dana.

[daqq]

Thanks for the replies guys!

The actual voltage drop in the current measuring mode on the shunts/diodes does not really matter as far as this test is concerned. That is, the DUT is OK with a few volts of drop.

I have been considering a high precision ADC - have just one range and amplify as needed. But the way of more shunts is just simpler - no switchable amplification needed, no particularly exotic ADCs...

[m98]

There are many non-exotic ADCs with an internal PGA. Microchip MCP3421 for example.