2

I am working on a project that requires measuring absorbance of a solution.

The circuit consist of an emitter led, a photodiode attached to an opamp with \$10^7\$ gain (TIA). The solution absorbs light from the led and causes different voltages corresponding to different concentrations. This works well for large concentration solution.

For example the voltage obtained with solution of:

  • 0 mol/l (no solute) is 9.95V
  • 100 micromol/l is 8.92V

But for concentrations <=5 micromol/l the voltage is also 9.95V, which indicates that the change in voltage is so small that I cannot measure it.

How can I improve sensitivity? I considered increasing the gain, but it can lead to a maximum of 100V detected. An opamp's saturation is around +/-15V.

MarkU
  • 14,413
  • 1
  • 34
  • 53
Totally New
  • 353
  • 1
  • 4
  • 12
  • You may need to design several gain ranges or else an auto-ranging system or else use a logarithmic amplifier topology. What's the highest micromol/l and smallest micromol/l that you need to measure? And how many decimals of precision do you need? – jonk Jul 23 '17 at 08:38
  • highest concentration around 10,000 micromol/l and smallest around 3 micromol/l . A less than 6% error is expected – Totally New Jul 23 '17 at 08:42
  • By "error" are you talking about traceable accuracy to a standards organization? Or just precision into the measurement's value? How repeatable must this be if you build two of them? (Are we talking about scientific replicability here?) Note that your dynamic range is more than three orders of magnitude and getting close to four (and more, if you count your 'error' in the mix at the low end.) – jonk Jul 23 '17 at 08:49
  • I mean the precision into measurement. – Totally New Jul 23 '17 at 08:53
  • The voltage between different turn of measurement (for example : first time , 2nd time)... also be expected not to be so much different . – Totally New Jul 23 '17 at 08:53
  • And will you be making measurements from 3 to 10,000 in the same sitting during a single experiment? Or would you be able to approximate the range and set the gain manually for a much smaller range of measurements in a sitting? – jonk Jul 23 '17 at 08:54
  • Note that you are talking about something like 150 nanomol/l resolution at the low end of your scale. Is that about right? – jonk Jul 23 '17 at 08:54
  • Yeah in a same sitting , i would like to measure (in micromol/l) : 3,5,10,20,50,100,500,1000,10000 – Totally New Jul 23 '17 at 08:56
  • 1
    Looks like (1) you will need to find a way to remove the offset and (2) perhaps use a DC/transimpedance logamp. The logamp will probably have enough dynamic range and also enough precision into the signal. Have you looked at the LOG101, LOG102, LOG104, and LOG112 yet, for example? The DDC112 might be yet another approach, if you can consider it. It's different, though. But I like it a lot and I used it extensively for a wider dynamic range than you are considering and at the kinds of equivalent gains you seem to need. – jonk Jul 23 '17 at 09:18
  • I have thought of log opamp a while ago but it is not available for my country . Since the price is also expensive ($16/unit for LOG101, i should use 2 opamps). Maybe remove offset is the only left option – Totally New Jul 23 '17 at 09:29
  • 2
    If 16 dollars is too much, I fear you will have a hard time getting accurate, repeatable measurements at 4 decades of resolution. – Joren Vaes Jul 23 '17 at 09:37
  • @MarcusMüller Joren A 16$ opamp maybe turn into $50 opamp within local shops. Even if mouser and digikey could do sth , a 16$x2 + ship fee is also expensive for us student to consider – Totally New Jul 23 '17 at 09:48
  • 1
    @TotallyNew Have you considered looking at requesting samples? I've (rather to my suprise) had Analog Devices ship me, for free, >75€ parts (usually 2 of each). This might be an option worth considering, especially since manufacturers tend to like students. – Joren Vaes Jul 23 '17 at 09:51
  • 2
    @JorenVaes has a very good point. Manufacturers **especially** like students that use their measurement-related ICs in a Lab equipment to do small series of measurements. Turns out, students later do the same when they're Phd candidates or in industry, and then they might be buying a crate of these ICs. But really, if your research hinges on $32 and the shipping fees to anywhere in the world, speak to your professor. – Marcus Müller Jul 23 '17 at 09:54
  • @Joren & MarcusMüller: thank you for the idea and all of your kindness . I am trying to request sample . Our professor actually fund us ~$35 (seems odd when researching with $35 in US or elsewhere , but this is a huge amount in my country) . That is why i do not want to spend too much . But if request sample is impossible , maybe we will ask our prof to help. – Totally New Jul 23 '17 at 09:58
  • @TotallyNew that seems like a reasonable course of action. Point is that you might not really end up much cheaper if you did this on your own design – doing something that a IC manufacturer can do with laser-trimmed electronics is often very hard to do with discrete components, and at these resolutions, noise and interference are not to be underestimated, so you'd probably have to run at least two prototypes on self-designed PCBs if you did this yourself. – Marcus Müller Jul 23 '17 at 10:22
  • 2
    @TotallyNew also note that the LOG101 family isn't the only logarithmic amps in existence – Maxim has the [MAX4207](https://www.digikey.com/product-detail/en/maxim-integrated/MAX4207ETE-/MAX4207ETE--ND/1701868), for example, which usually is much cheaper. (but it's an amplifier that measures currents, not voltages – but this might actually be an advantage in your usage scenario! – Marcus Müller Jul 23 '17 at 10:42
  • 1
    The solution might not be a circuit change. Can you use a longer optical path through your sample? For example, use a bigger sample, or use mirrors so that your optical beam passes through the sample multiple times before reaching the detector? – The Photon Jul 23 '17 at 14:17

4 Answers4

4

Spectrophotometers often use a reference beam and a sample beam. The idea is that rather than having to detect these small differences directly in the presence of noise and other (e.g. temperature) fluctuations, only the difference with respect to the reference beam is measured.

So, build two identical sets of your apparatus, and in one of them put a cuvette containing only the solvent. Then use an instrumentation amplifier with large gain to amplify the difference between the TIA outputs of the two devices. Instrumentation amplifiers are fully differential amplifiers, so suitable to measure this small difference in the presence of a large d.c. offset.

Of course, since the two absorption cells will not be completely identical, you will still have a d.c. offset to deal with. But it will be much smaller than otherwise, and you can probably either calibrate it out, or compensate it by adjusting the gain and offset for the reference TIA.

Oleksandr R.
  • 1,644
  • 11
  • 14
3

Posting a schematic of your circuit may help in providing more insightful answers.

One approach would be to reduce the current to the illuminating LED and increase the op amp gain when the solution concentration is low. The reduction in light reaching the photodiode will increase headroom for amplification before the ouput approaches the rail. This allows you to effectively increase your low concentration resolution.

The additional gain can be obtained with a basic op amp following the TIA. This will keep your costs down.

This will of course require a second calibration procedure for the lower concentration range.

Glenn W9IQ
  • 5,526
  • 11
  • 17
3

Instead of a Log-Amp, I'd use a 24 bit sigma-delta ADC. Here are a few examples.

These achieve extremely high precision for voltage measurement. You also need an accurate voltage reference, though.

Also, keep in mind that your LED and photodiode are temperature sensitive, therefore you will need calibration.

bobflux
  • 70,433
  • 3
  • 83
  • 203
0

Implement an optical chopper, switching between 2 paths with standard and the test specimen. Then use AC_coupled amplifier to boost the squarewave output.

This, as experiment, lets you determine what is possible.

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46