Measuring low current

[beerbaron]

Hi, I'm trying to measure the average current going to a LED that has PWM (100kHz). I used to use a 10 Ohm resistor and a 100uF cap in parrallel that I would put in line with the LED. Problem is that 10 ohm has too much impact on the circuit, using a smaller resistor like 1 ohm or .1 ohm would be perfect but then, how accurate is a digital multimeter when it come to 100uV to 100mV...

Any suggestions on how I can pull this off. Is there some good products around for my needs? Should I go with an Op-Amp ... I doubt an Op-Amp would be precise enough for uV and low mV ?

Thanks
Frederic

 

[IRstuff]

That's not really the point, is it? I'd find it hard to believe that your DMM is capable of reading 100 kHz voltages at any resolution.

You should be using an oscilloscope or a data acquisition system.


TTFN

FAQ731-376

 

[Skogsgurra]

The 1000 uF is probably supposed to average the PWM out. And that could work. But you should use a smaller value - and not an electrolytic capacitor. The have too much impedance at 100 kHz.

I would try and use the current range of the multimeter and short out the PWM frequency with a bypass capacitor. Here, you may need the 1000 uF plus an HF capacitor because the resistance of the DMM is rather low.

But, of course, an oscilloscope that measures voltage drop across a 1 ohm (non-inductive!) resistor is a better technique. Especially if your scope has built-in math to calculate average current from the actual waveform. Yes, AVG, not RMS, it is AVG that produces the light.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[logbook]

Of course you haven't said how accurate is good enough for your needs! I'm guessing that ±5% will do. The DVM is the best bet. The scope will just cause more trouble. I suppose the current is small (2mA to 20mA depending on the LED type.)

Use a solid dielectric capacitor as suggested above. You want a DVM with a 20mV full scale setting. If you turn the circuit off and check the zero first the accuracy should be ok. If you are looking for good efficiency in the overall circuit you should chop at a lower frequency.

 

[IRstuff]

The capacitor might cause the LEDs to get fried.

Pulsed LEDs are often run at much higher currents than statically driven devices.

TTFN

FAQ731-376

 

[ScottyUK]

Keithley claim their Model 2000 has a 300kHz bandwidth. Not sure if that is -3dB: spec sheet is a little vague.

Presumably there is a capacitor immediately upstream of the switch transistor / FET? Can you break into the supply there and measure the average current?


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

Sometimes I only open my mouth to swap feet...

 

[itsmoked]

Is this a system or just some lone LED?

If it's a system do what everyone else does. Monitor the system current with the LED commanded off then measure the 'system' current with the LED being PWMed. You will get the actual system current contribution caused by the LED being driven.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[IRstuff]

The Keithley Model 2000 has 300 kHz response, but that only applies to AC voltages and that's only with 35 readings/second:

http://www.keithley.com/data?asset=359

TTFN

FAQ731-376

 

[ScottyUK]

It applies to AC current measurement too. Why would 35 readings per second be a problem to anything other than an automated system?

I'd posted the information on the Keithley meter primarily for your information based on your comment "I'd find it hard to believe that your DMM is capable of reading 100 kHz voltages at any resolution". Perhaps you forgot to include the qualifier "... on a DC measurement range".

As this is a square chopped signal there will be significant harmonics above 1MHz which would render useless the measurement taken using a 300kHz B/W meter if reasonable accuracy was required.


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

Sometimes I only open my mouth to swap feet...

 

[IRstuff]

I'm really unclear why there's such a reluctance to use an oscilloscope. You'd get to see the EXACT waveform with the correct light output at precisely the same instant.

Turning the drive into psuedo-DC is potentially going to change the thermal conditions in the LED significantly, resulting in a different I-V behavior than when it's pulsed.

TTFN

FAQ731-376

 

[ScottyUK]

Agreed, a scope is the most appropriate method.


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

Sometimes I only open my mouth to swap feet...

 

[Skogsgurra]

Yes, IR. Agree. The capacitor idea isn't very good after all. It will be a short to the HF.

The reluctance may come from unavailable scope?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Skogsgurra]

Duh...

I have to go back on that. The cap does only short HF around the measuring shunt. And that's what it is supposed to do.

Current limiting for the LED is already there. So, no probs with the cap.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[logbook]

problems with the scope method:

The scope is going to want to ground one end of the sense resistor. The circuit might not be this way up.

There is going to be a shed loadf of common-mode noise on that point anyway so you would need to do a null reading with probe and ground lead shorted together to establish the "zero" level, which may not be zero and might therefore need to be subtracted.

Actually the capacitor helps give you the mean level, so the DVM bandwidth is irrelevant. You don't need or want any bandwidth in the DVM. A 20ms integration time (or longer) would be ideal.

The capacitor/resistor pair is not supposed to be the current limiting part of this circuit. Hence the capacitor does not blow up the LED!

 

[IRstuff]

There are a number of ways around the scope grounding issue:
> Use a differential scope
> Use an op amp based unity-gain buffer to covert to single end
> Use a current probe with the oscilloscope.

TTFN

FAQ731-376

 

[logbook]

Yes IRstuff, all that is true. But it is an expensive way of going about things!

 

[IRstuff]

Perhaps, but if one were trying to design a PWM circuit, would one really be sans oscilloscope? That would be tantamount to debugging a Windows program without the display.

Obviously, everything is relative, but a cheap oscilloscope is only around $330:

http://shop4.outpost.com/product/321869?site=sr:SEARCH:MAIN_RSLT_PG

add an op-amp with 5 or 6 resistors for around $3.

Or go really cheap:

http://shop4.outpost.com/product/5230047?site=sr:SEARCH:MAIN_RSLT_PG,

but this guy doesn't quite have enough bandwidth.

TTFN

FAQ731-376

 

[jimkirk]

Back to the original post, the ICs in most cheap DMMs have an inherent range of +/- 200 mV, higher ranges are done with resistor dividers, so readings of tens of mV, even a couple of millivolts (2 mA through one ohm) are generally no problem. I have a $20 Radio Shack meter which I use for this type of thing all the time. With a typical "3 1/2 digit" meter, that will get you 100 uV (100 uA with 1 ohm) resolution. Even cheap ones use dual slope integration techniques or better, along with auto-zeroing. I've never had a linearity problem doing this, but you can easily set up some voltages and currents to check that out.

As per Skogsgurra, to get the average reading, put a resistor in SERIES with the volt meter, and the capacitor across the meter itself. This will allow the PWM current to switch normally through the LED and your current sense resistor, and apply a filtered average to the meter.

Something like 1k will isolate your capacitor from the 1 ohm sense resistor, and a 0.22 uF will filter out the 100 kHz nicely. Use a good plastic cap. Polypropylene of you can get it, otherwise mylar shold be adequate. You might also put an NPO ceramic cap of several hundred pF across the plastic cap to filter out the highest harmonics.

I've nothing against using a scope, but for a quick, easy measurement, DMMs are dandy. Especially if you're in an isolated place and don't have a nice battery operated scope around like some folks do.

 

[Madcow]

How precise does it need to be?
Build the same circuit but use a dc sorce.
Adjust the led until its as bright as the
other one , then read the current.
I personally would use a scope if one is
handy.

 

[beerbaron]

Thanks guys for all your inputs. ±5% accuracy is allright for me.

I have a USB card that cand read up to 2000 samples per second.

I wanted to use the digital multimeter because it is floating and therefore a lot easier to avoind ground issues.

Madcow: LEDs light intensity can vary a lot between each of them the human eye is all but good at detecting light intensity. Anything under 2x the eye will not see. So, the eye would not see the difference between 2000 mcd and 3000 mcd.

Jimkirk, that sound like a good plan I'll check for that.

Thanks everybody
Beerbaron