So for the past few months I've been working on a simple circuit which I've (very poorly - please excuse it! I'm not very familiar with this particular CAD software or professional schematic drawing in general) drawn a schematic of and attached. I should point out that in the schematic I've drawn each resistor as having its own separate ground, when this is not the case - all resistors are wired to the same ground.

More or less, it's an optics device which we (my group) are using for biological studies by measuring the transmittance of IR light through a liquid.

It's a pretty basic setup: 6 IR LED emitters and phototransistor pairs, 6 300 ohm resistors and a voltmeter measuring the voltage drop across each resistor. We're using that measured voltage and correlating it with the optical density of the medium. I should also point out that we're using a power supply to ensure that the current across each emitter stays constant at 60 mA.

The problem I've been having is that they're unsteady measurements. For the last 24 hours I've been measuring the voltage drop across all 6 resistors using water as the liquid. While R2, R3, R5, and R6 are all pretty steady and the differences in voltage are normally distributed about a value, R1 and R4 are steadily decreasing in measured voltage, to the point where I can't attribute it to chance or random distribution. I've seen this happen on multiple occasions, leading me to believe something's going wrong.

I don't think it's a problem with the voltmeters - there's only two voltmeters each measuring three resistors, and it seems to take measurements fine for all but R1 and R4.

I don't think there's anything in the liquid causing this either - it's plain water and there's nothing to suggest that it's been contaminated.

I'm open to all suggestions - I'm just not sure what could be causing this besides possibly a faulty detector, though this behavior seems to suggest something else to me.

Thanks!

 

What are you using to drive the set of 6 series LEDs?

hgmjr

 

The LEDs are driven by this power supply, set to a constant current of 60 mA:

http://www.home.agilent.com/agilent/product.jspx?pn=e3616a

The power/ground source for the circuit, however, is just a modified laptop charger.

 

From your description, it sounds like you are building a "single-beam" instrument for determining %T or OD. Single-beam instruments have a lot of problems in calibration and standardization. Have you considered a beam splitter (there are various types) or rotating mirror so you can directly monitor the intensity of the incident beam?
John

 

I hadn't considered that, though I certainly will now!

I'm not entirely certain I see how such an issue would be causing my current problem, though. I wouldn't be surprised if a mirror or beam splitter significantly reduce the noise and measurement variations, but I don't see quite how it would explain what I'm seeing (picture attached of what I mean in case the above description wasn't clear enough).

 

Sorry, I sort of jumped on the design first, without having seen your data. What if the LEDs are aging and output is decreasing with time. Your change is "just" 7.5 perts per 189 (4%), which could easily be due to that factor. My guess is that an emitter ages faster than a detector, so for a rough and dirty experiment, you could put another detector near enough to the emitter to get a fraction of scatter light and see if that response decreases too.

If it is aging of the emitter, then perhaps someone from industry will know how to reduce it. Maybe a lower current or pulsing, say at 10% duty cycle?

John

 

John's suggestion of a differential measurement is a good one and commonly used for situations like this (think Michelson-Morley, LVDTs, differential line drivers, etc.). Another reason for the differential detection scheme is that you can use it to servo power to the source to correct for changes in source output.

A relatively common technique is to "burn-in" the components by running them somewhere around their rated power and at an elevated temperature.

If I were making something like this, my intuition suggests I would take a look at an incandescent source and thermopiles as sensors. Actually, since I have a spectroradiometer that measures into the near IR, I'd use that.

 

60mA continuously is extremely high for an IR LED but part numbers were not shown in the schematic. Maybe the LEDs are heating up too much and fading.

 

Actually, since I have a spectroradiometer that measures into the near IR, I'd use that.

I hope this is not too OT, but the issue of the IR transparency of water comes up frequently in various ways and is related to this thread in that way. It is hard, nigh impossible, to find the near IR spectra (from visible to 1200 nm) on the internet. I suspect the data are available by subscription or in research libraries, but the nearest one to me is an hour's drive each way.

Would you be so kind as to take the spectrum of water, nujol (or any purified mineral oil), and window glass in that range and post a pdf here?

John