I really need to relay on precise measuring of my not yet build spectrometer.
Could you explain what this sentence means? Feel free to take a few lines.
jremington:
Could you explain what this sentence means? Feel free to take a few lines.
Ok. The point is, that I'm building spectrometer as cheap as possible but as efficient as possible. My goal is to get precision at the highest point as possible. When I finish this project, I will bring it to my teacher at university and we'll perform an test how much precise is my DIY spectrometer against the one we have at university (very, very expensive one). But now as you can see, the measured data are heavily influenced by noise. And that's problem, because when I'll measure absorbance and then calculate concentration of my sample it must be as close as possible to real concentration of that sample. I'll say it again. I KNOW that I can't expect miracles, but do everything I can to get the highest precision. Absolutely perfect scenario would be, when I take some sample and do measurement. I'll get for example 220. And each measurements after (with same sample) must be as close to 220 as possible. Then I can say it was measured precisely.
Hope that's the right explanation. 
You may need to modify your circuit a bit. I think you'll find this training module very helpful:
Common Photodiode Op-Amp Circuit Problems and Solutions
But now as you can see, the measured data are heavily influenced by noise. And that's problem, because when I'll measure absorbance and then calculate concentration of my sample it must be as close as possible to real concentration of that sample.
-
Noise is always a problem. To reduce it, use low noise circuit techniques and average some number of measurements together. Particularly with photodiodes, the detector circuit should be enclosed in a grounded metal shield to reduce capacitive and inductive pickup of AC line and other electromagnetic interference.
-
To use your device to measure concentrations by absorbance, the light source will have to be very stable and have a small, well defined, appropriate emission wavelength range.
-
You will need to calibrate your spectrophotometer (photodiode output) by measuring the output for a succession of known concentration standards, and use those data to make a calibration curve.
jremington:
Noise is always a problem. To reduce it, use low noise circuit techniques and average some number of measurements together.
To use your device to measure concentrations by absorbance, the light source will have to be very stable and have a small, well defined, appropriate emission wavelength range.
You will need to calibrate your spectrophotometer (photodiode output) by measuring the output for a succession of known concentration standards, and use those data to make a calibration curve.
- I can understand it, but averaging numbers together is not really precise technique.
- My light source is stable, and I have emission wavelength graph, so I know what I'm dealing with.
- Yes I will do calibrations, that's for sure, but I can't make calibrations when I have such noisy input
but averaging numbers together is not really precise technique.
You have been seriously misled. Back to your statistics class!
jremington:
You have been seriously misled. Back to your statistics class!
Ok, but that would work if you have very little difference between values, but when I measure higher numbers I get like +-50 which is too much when we have 1023 resolution.
I've used 'very expensive' commercial spectrometers and built my own bog standard light detectors enough to be sure that you'll need more than one stage of op-amps and a lot less than a Meg-Ohm in there to get anything remotely decent out of your equipment. Whilst different particular op-amps are better than others for your 'transimpedance pre-amp', 'mid stage amp' and possibly 'third stage boost amp' I think that you'd get marks for making a tangible improvement over the present setup and could do so with two 'bog standard op amps' from your stash, the first used as the 'transimpedance' 'first preamp', possibly with 100k in place of 1M, followed by a second stage x20 with a trim pot somewhere, possibly about 5k 100k configured as a voltage multiplier. 15pF smoothing capacitor on the MegOhm is not enough to squish the usual background noise in a typical university science building. I suggest choosing equal RC time constant on all amplification stages and that equal to one interval going around your arduino measuring loop.
You should not constrain your arduino to one a/d every 300ms. Why not go as fast as possible and sum them?
Find out what the mains frequency is in your region and test with a scope probe on your a/d pin to see whether you have 50 Hz hum. That was easy to see in central London and was very clearly dependent on build quality of the front end circuit. Long looping wires into a MegOhm transimpedance are really bad news. While you have a scope on the bench, just sit there and watch it for an hour one day. I went in before breakfast to try that, and all sorts of intermittent rf nasties flickered up on the screen especially as equipment in other rooms got switched on. Smaller cabling and smaller R_gain on the first amplifier mitigate those.
Whilst I've recommended a couple of things to try to get marks, you are not going to beat the very expensive device (has your department got an "optical spectrum analyser" ?) because those have a smaller front-end with better build quality than anything handmade can achieve.
OP
Monitor the voltage of a simple resistor voltage divider.
What does the graph of this look like.
.
LarryD:
OP
Monitor the voltage of a simple resistor voltage divider.
What does the graph of this look like..
Well, that's good news!
Now using the Amp cct. do the same thing.
.
ad2049q:
I've used 'very expensive' commercial spectrometers and built my own bog standard light detectors enough to be sure that you'll need more than one stage of op-amps and a lot less than a Meg-Ohm in there to get anything remotely decent out of your equipment. Whilst different particular op-amps are better than others for your 'transimpedance pre-amp', 'mid stage amp' and possibly 'third stage boost amp' I think that you'd get marks for making a tangible improvement over the present setup and could do so with two 'bog standard op amps' from your stash, the first used as the 'transimpedance' 'first preamp', possibly with 100k in place of 1M, followed by a second stage x20 with a trim pot somewhere, possibly about 5k 100k configured as a voltage multiplier. 15pF smoothing capacitor on the MegOhm is not enough to squish the usual background noise in a typical university science building. I suggest choosing equal RC time constant on all amplification stages and that equal to one interval going around your arduino measuring loop.You should not constrain your arduino to one a/d every 300ms. Why not go as fast as possible and sum them?
Find out what the mains frequency is in your region and test with a scope probe on your a/d pin to see whether you have 50 Hz hum. That was easy to see in central London and was very clearly dependent on build quality of the front end circuit. Long looping wires into a MegOhm transimpedance are really bad news. While you have a scope on the bench, just sit there and watch it for an hour one day. I went in before breakfast to try that, and all sorts of intermittent rf nasties flickered up on the screen especially as equipment in other rooms got switched on. Smaller cabling and smaller R_gain on the first amplifier mitigate those.
Whilst I've recommended a couple of things to try to get marks, you are not going to beat the very expensive device (has your department got an "optical spectrum analyser" ?) because those have a smaller front-end with better build quality than anything handmade can achieve.
I had to read your post maybe five times 
Thank you for such a long answer. The problem is, I'm very early into electronics and something I understood, something not. Can you please link to some similar projects or some schematics? and yes I really want to decrease resistor value, because I though about digital potenciometr... to increase sensitivity to some wavelengths.
LarryD:
Well, that's good news!Now using the Amp cct. do the same thing.
.
So you have no detector (just resistors and the amp) in the circuit for the graph in post #32?
.
LarryD:
So you have no detector (just resistors and the amp) in the circuit for the graph in post #32?.
Sorry wrong graph. Here's with 500k
My 10W (mains powered) LED bench light and the power adaptor for my laptop generate so much noise that they can't be used if I'm doing anything with high amplification op-amps. The results from the ADC have an offset of 400+ and noise of +/- 50, and there is little difference with changes in the input I'm trying to read.
Does that sound familiar?
Martin-X:
My 10W (mains powered) LED bench light and the power adaptor for my laptop generate so much noise that they can't be used if I'm doing anything with high amplification op-amps. The results from the ADC have an offset of 400+ and noise of +/- 50, and there is little difference with changes in the input I'm trying to read.Does that sound familiar?
Well it depends. When I have my arduino connected with pc my precision is +-10. But when I switch to external DC it can be worse then +-30.