accurate measurements at uA range using 9227

[thanvg]

Dear all hi,

please assist me on my problem which seems straightforward but has led me to a deadend.

I am trying to measure current consumption on a low power device which, in sleep mode, draws around 3 uA.

I would like to do this using the NI 9227 module. I have the one with the four pairs of screw-type connectors.

Performing some initial tests I ended up with measurements that have too much fluctuation. Please have a look at the attached images. They are taken from the Signal Express and from the NI max, respectively. Please, note that measurements were taken with no component connected to the 9227 DAQ. (I also tried shorting the two leads of one of the inputs without noticeable effect).

So, I have a system with accuracy of 24 bits which should give a step of 60 nA and I cannot measure uA.... I would be amazed if this was just ambient noise getting captured by the 9227.

Has anybody had such an experience with the 9227? Or, if possible, could someone just try and disconnet their device-under-test and just do a current measurement without load?

It really seems strange to me, I was expecting that I would be able to perform such a, reletively, simple task with the 9227.

Thanks in advance,

Thanos

[Henrik_Volkers]

the spec tells me 1.785µA/LSB and a noise of 400µA_rms

the internal shunt resistor is 12mOhm ... with 3µA means 36nV to measure 🙂

and Offsetdrift is +-51 µA/°C

and 24bit delta-sigma ADC doesn't mean a 24bit resolution !!  the ENOBs arn't specified 😉

I would use a TIA for such kind of measurements .....   And the 3µA continous or pulsed?  

Another way is to use a capacitor of known value as a power source and measure the voltage ramp down.

[thanvg]

Hi there Henrik, thanks for the great advice!

Well, actually, I admit taking into account the specs selectively, ignoring both the absence of ENOB spec and the huge rms noise spec.  Also, I entirely failed to notice the offset drift over temperature, which would cause enormous problems (at fixed temp you can calculate it out, but 51 uA for a Celcus degree...this is too much).

Please, if it is not difficult, advise on the TIA solution. I want to measure the current that a portable device draws from a battery or a power supply (continuous current) without introducing a huge voltage drop on the device. I am not sure how can I configure a TIA-employing setup to perform such task, the sample schematics I have found online are mainly targetted to sensors.

Also, I am not sure I can employ the capacitor solution, since I do not want to interrupt the operation of my device, which would be the case after some point of discharging. (not sure if I correctly got the idea though).

Greetings from Greece!

Thanos

[Henrik_Volkers]

A TIA circuit is like this one

Source is the LTC6268 datasheet a very fast , very low bias current OP from linear. Maybe too fast for your application , but you can add a lot of Cf since you don't need that bandwidth.

the output voltage is U_out = I_in * Rf   so for 1nA and 1Meg you get 1mV (the feedback resistor will add noise so I wouldn't choose a bigger one (beside other thinks ...leakage currents etc.)

next would be a second stage with an amplification of maybe 10 to 100 .......

take a look at the datasheet and I strongly recommend Jim Williams appnotes from linear .. I don't have the time for searching, but he did investigations on low current measurements..

And look for the Keithley appnotes on low current&voltage measurements.....  

the other  idea would be to use that OP as a very high impedance voltage follower

Use a low leakage capacitor that you load with the supply voltage, if you switch of the supply (reed relay) the current of your circuit will discharge the cap. this voltage decrease is a funktion of the capacity and the current.  if you make a measurement with and without circuit you migth cancel out residual bias and leakage currents......

However building low curent,voltage,leakage circuits is not that easy...  clean and dry PCBs , PTFE standoffs , good thermal conditions, guarding traces & shielding  .......

[thanvg]

Hi there Henrik,

thanks for taking the time to present me with such a detailed description.

I don't think I can employ your suggestion at the moment, since I was looking for a fast, plug-and-play solution with the existing equipment. It seems that this is not possible and, when possible, I will have to invest the time to follow your suggestions and prepare my test-bed.

However, still, it is not obvious to me how I am to connect this circuitry in series between a battery and the mobile-DUT, transparently (as I would with a multimeter). Please excuse my ignorance but it's been a while since I was last involved with actual electronics...

Many thanks again,

Thanos

[Henrik_Volkers]

Think of a shunt resistor between GND and your device. To measure the current, you measure the voltage across the shunt resistor.

For low currents you get low voltages (U=R*I) .

You can increase the shunt resistor to get a higher voltage, but that voltage (burden voltage) is 'lost for the supply voltage'.

Now you take a OP in a TIA setup: The (burden) voltage between input and GND is (nearly) zero , the input current of the OP is also nearly zero , all current will flow through Rf (say a high value current shunt) , the output voltage is Uout = - Iin * Rf.   

 The joy of engineering is to do what you want with the stuff you can get:

What do you want?  Current range and rel. uncertaincies.....  

What do you have?

And finally:

Do you need it permanently ?  (You can rent such devices, no invest 🙂 )

Traceable? 

[Henrik_Volkers]

And just got another idea:

If you have a DC current (not pulsed)

Add a relay to short cut the 9227. measure 1s with and 1s without shortcut. mean each 1s measurement (choose samplerate for best line noise reduction) .    that will compensate the drift, and migth get you down to 0.5 µA .....

Now take a 1.5 cell and four 1Meg resistors  as a current source , by combination of the resistors you can get a good and quick estinate what you can resolve 🙂

EDIT: Look for a relay with low contact resistance  and low thermal EMF .....    you want to shortcut 12mOhm !!  But still worth a try

EDIT2: Have to think twice on it...    even better: take two DPDT relays,   measure the current in both directions and measure the open contacts (0A!)  while shortcutting the current sense connectors to your device 

[thanvg]

Dear Henrik thanks again,

you are so prompt in sharing knowledge that I will take advantage of it a bit more. I am attaching the simplest of schematics, based on your advice earlier. I still can't put my finger on this: ideally, I would like a zero resistance path from my load to the ground (so that no voltage loss is observed). The TIA solution, send the whole current through Rf to the output, which is connected where? Here is where I am missing something it seems...

Thanks also for the great real life advice of calculating out the offsets of the 9227 (and any similar for that matter) device. Elegant path around the problem....!

Thanks a million,

Thanos!

[Henrik_Volkers]

You should add a Cf of at least 10pF (better 100pF) , because amplifier tend to oscillate while oscillators don't 😉

what you are missing is the DUAL power supply with GND connection for the OP (have a look at the datasheet of the OP you are using, +-15V for the old models , modern ones +-2.5 to +-5V  ) 

two 9V blocks or two 12V lead acid ones (said to have the lowest noise)  for the first try

AND decoupling capacitors as close as possible at the OP!!    100nF are fine....

The TIA allow you to measure the current without burden voltage, but all current your DUT consumes must also be supplied (well, the negative current) by the OP.

Read the datasheet i linked , some good advices on layout and stuff is noted there.

Have nice weekend

[Henrik_Volkers]

You can try this

http://www.eevblog.com/projects/ucurrent/

http://www.adafruit.com/product/882