Method of Measurement of average current is accurate?

[grigson]

Hello,

Please can you tell if our method of measuring a 500KHz triangular current waveform is accurate?

We are just measuring the voltage across a series sense resistor.


This is the shape of current waveform we are trying to measure the average value of:-

http://i39.tinypic.com/2vwsht1.jpg

So we are trying to measure the average inductor current in each inductor of a dual buck converter.
(we need to do this to assess what actual maximum current the load (FPGA) is actually drawing)



We are measuring the current by using a 2mR series resistor. (its a 1% surface mount sense resistor)
-we measure the voltage across the sense resistor with a Hewlett Packard HP34401A digital meter.
(obviously wires (twisted pair) are taken from the sense resistor terminals to the meter)

HP34401A Digital meter User Guide:-

http://www.ee.buffalo.edu/courses/elab/hp34401.pdf

-the load current, we know is around 15 to 18A , but we need to know it exactly.
-anyway, its obviously a triangle current waveform with a few amps peak to peak and 15 to 18A average.


The buck is actually a dual-buck converter supplying an FPGA

Vin = 12V
Vout = 0.9V
Each FET switches at 500KHz (there’s interleaving)
Inductors are 470nH
Iout(max) = What we are measuring here!!!!

So please do you know if our sense resistor voltage being read by the HP34401A is going to give us an accurate way of measuring the current?


NOTE:
Before taking the measurements, we got the software engineer to code the FPGA with code that will make it draw the maximum current.
The dual buck and the FPGA are actually in the environmental chamber at 80degC when we measure the current.
(obviously we cannot put a current probe into the chamber at 80degrees C……even if we could, we’d still have to break the pcb track and bring a wire through the current probe’s “jaws” which would not be advisable at this high frequency)

HP34401A Digital meter datasheet:-

http://www.metrictest.com/catalog/pdfs/product_pdfs/hp_34401a.pdf

HP34401A Digital meter User Guide:-

http://www.ee.buffalo.edu/courses/elab/hp34401.pdf

 

[LiteYear]

Two obvious sources of error here:

1) the sense resistor. A 1% resistor is not necessarily high precision. It's okay for stability from resistor to resistor, but check carefully the resistor's sensitivity to temperature. It could introduce a lot more than 1% error. This can be compensated for to some extent by putting another resistor of the same type in a similar location, passing a fixed current through it and comparing its voltage to the sense resistor's voltage. Some sort of wheatstone bridge might help too.

2) the multimeter. Make sure the integration period is many times your cycle period so that partial cycles don't influence the measurement too much. Looks like the 34401a will do just fine in DC mode - even in the fastest setting it will integrate over 200 cycles, which ought to give you a decent accuracy average value.

I think the approach is okay. Interested to hear if others have other ideas.

 

[Skogsgurra]

The "shunt's" inductivity is very often an important error source at higher frequencies. Make sure that your return wires(s) are as close to the resistor as possible. Think coax shunt.
I have used a flux gate transducer for this kind of measurements. The one I used has 100 A capability at 5 MHz. Nice little probe. All I remember is that there's the number 520 in the name. Iprobe 520, perhaps.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

I believe your measurement scheme will provide you with the RMS average which should be fine for your needs.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[juggler1]

Hi
sorry to come in on my work handle

Sorry but my rolling contract has been terminated due to the fact that I presented this method of current measurement as being accurate

http://cp.literature.agilent.com/litweb/pdf/5968-0162EN.pdf

…page 2 of the above clearly shows that HP34401A cannot be deemed accurate below 100mV.

-My readings were around 16mV so the accuracy would have been pitiful.

The product was for a very expensive robotic product and they said I nearly jeopardized the whole project by presenting an inaccurate current measurement.

When I asked them why my contract had not been renewed they pointed out that I had not been aware about how ridiculously inaccurate my current measuerement technique (first thread above) would have been.
-they virtually laughed me out of the door…..

 

[MacGyverS2000]

I see for samples <100mV, the range is 10mA with an accuracy of 0.005% of the reading and 0.01% of the range.

What's the problem?

I don't see any current measurement method introduced, I only see a datasheet for the VOM.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[MacGyverS2000]

Ack, my bad... thought that was the first post in the thread. Let me go back and read it all.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[jimkirk]

Maybe too late, but I'd add a couple resistors in series with each node of your current sense resistor and a capacitor across the free ends of the resistors to filter out your 500 kHz, then measure the DC voltage, which will be proportional to the average current. Don't use the AC scales, the meter doesn't have the bandwidth or accuracy, and you said you wanted average current anyway, not RMS.

 

[IRstuff]

There seems to be a whole bunch of troubling questions:
> You refer to the measurement as "average curent" whereas the meter is measuring "true rms." That distinction alone would cast doubt on your understanding of the problem.

> You stated that the switching frequency was at least 500kHz, yet the meter is only specified to 300 kHz.

> You presented the datasheet initially, and you knew both the resistance and current range which would have resulted in a 6 mV delta between the high and low points of the current, and the meter was rated only to 4% of reading=~1.2mV and 1/2% of range=0.5mv, which would have resulted in ~1.7mV uncertainty, when the delta voltage range was only 6 mV to begin with, resulting in a 30% measurement uncertainty.



TTFN
faq731-376

 

[Skogsgurra]

Somewhat confused.
Are Grigson and Juggler1 the same person? I get that impression, but can't say for sure. In that case...

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[juggler1]

yes grigson and juggler1 are both me. (in and out of work)

So anyway, i did have an RCR filter just before the wires went into the meter....ie two series resistors (one in each line) and the shunt cap of 3uF.

SO that was filtering the ripple out.
As IR stuff said, its hopelessly inaccurate

 

[LiteYear]

No, no, you need to use the meter in DC voltage mode, as I stated above. The whole point is to rely on the meter's input integration, not on its ability to do a true RMS calculation.

Then surely 0.003% + 0.003% = 0.00048mV + 0.003mV = 0.00348mV is an acceptable error? What am I missing here?

 

[IRstuff]

15A*0.002mohm = 30 mV
18A*0.002mohm = 36 mW

So, 3.48 mV is half of the range the OP was trying to measure

TTFN
faq731-376

 

[LiteYear]

15A*0.002mohm = 30 mV
18A*0.002mohm = 36 mV

Yes, correct voltages but note that the sense resistor is 2mOhm, not 0.002mOhm.

So, 3.48 mV is half of the range the OP was trying to measure

Yes, but 3.48mV is 3.48% of the range, not 0.003% of the range. That spec sheet claims the multimeter is rated to an accuracy of 0.003% + 0.003%, not 3% + 3%. I would expect 3% from a crappy handheld multimeter and <0.01% from a high performance benchtop meter.

 

[roydm]

For a simple shunt just take a few feet of wire and pass a known current thru it.
For example 90 inches of #12 passing 1 amp.
Now measure the mV at two points near each end, say 80 inches apart, you will get just over 10mV
So now just move the meter leads a bit closer until you get exactly 10 mV (about 76")
Now you have a shunt that will give you 10mV per Amp.
Scale it back to 1 mV per Amp and your two points are about 7.6 inches apart

Don't make your meter connections at the end where the current's connected or you will get an error due to the contact resistance.

A #12 shunt should be able to measure 50 Amps for a short time, I used one to measure the current draw of my starter motor (about 160 Amps)

You can do this with any old scrap of wire and an appropriate known current source, I use something like a 12 Volt headlamp which draws around 4.5 Amps. Measure the current directly with my Fluke then mV.

 

[ScottyUK]

At 500kHz I bet the results are getting a little suspect...


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[Skogsgurra]

Scotty
Not only suspect - downright wrong due to the shunt's inductivity.
I know you know, but others seem not to understand.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.