@Benm:
Actually if we follow like what you said, we can just calibrate the sensor when cylindrical lens being applied (the lens must not have AR coating, for flat spectral rejection/reflection)
The thing is, i'm sure the lens itself will be expensive and if the lens is gone then you'll end up with miscalibration.
For now, the sensor's coating burning off can no longer being a limiting factor, only the thermal limit of the PN junction is still being a factor for now.
Higher power to be measured is impossible for coated surface, as the 6 watt represents peltier Vout saturation, which giving you extra skew (about -10%).
Are the 10mm ones able to handle more power than the 6 mm ones?
As for the losses caused by the lens, they surely are a factor, but that could be adressed in software - just make an option 'lens on/off' somewhere to compensate for the loss.
It would be cool if the upper range of this system could be extended a bit so it works with all modern laser diodes, up to perhaps 20 watts or so. Another way to do this is by using ND filters, but many of those will note cope with such optical power densities well either.
Another option would be to come up with a partial mirror aligned such that it reflects 90% of power or so into a beamstop. I'm sure this would be delicate to set up, but at these power levels options become very limited.
It would be great if one could make a small Peltier TEC based LPM good up to 20 watts, but the TEC seems to be the limiting factor. Everything suggested makes this more complicated and higher priced. Isn't this a case of "it is what it is?"
@Benm:
We don't need that expensive setup actually, that's why on OP/Post#1 I have coating option No.3 (No coating at all)
The 10mm sensor has not been tested, it is because of the slow response that if must use laser with prolonged period which may cause a new problem to my laser.
All the gas lasers that i can use here only outputs a weak beam that cannot characterize the TEC.
@paul1598419:
It can be done actually, by choosing for non coated surface (coating option no.3), this making the sensor surface act as broadband rejector rather than broadband absorber, yielding smaller TEC voltage generation while widening the dynamic response.
You'll get a higher measurable power albeit at worse resolution. Also, as the TEC do not use coating, it will produce a very high damage threshold, not to mention the maintenance will be easy.
Unfortunately, some may think that if i sell uncoated surface, it just like i'm selling incomplete product as we had a placebo/doctrine that laser sensor must have coating to absorbs more light. I guess you can catch on what i'm trying to say :beer:
Yeah, it never occurred to me to leave the TEC the way it comes. I suppose that would work, though what does it do to the linearity of the TEC's output voltage? Do you have to add a curve fitting equation to allow for that? I'm only asking because i don't know the answer to that question.
I'm using a linear approach, as i cant use curve fitting equation in the microcontroller.
Thus, just like digital multimeter, this LPM has a different uncertainty at different range.
Also by using linear approach, the user can recalibrate (or updating the calibration) by themselves with only requiring to enter one value.
Long ago i tried to program curve fitting on the microcontroller but it require user to enter value at different point which eating large program space used for other features.
Because of this, i have omitted the use of curve fitting. I can just write down the uncertainty at certain range :beer:
The important question is: does the uncoated TEC reflect the same amount across the whole visible spectrum? If it doesn't then calibration would have to be per wavelength.
It's the surface absorber characteristic of some sensor (the green one, thermal broadband).
As we can see that it doesn't absorb the same amount across whole visible spectrum.
But every user of this sensor will just ignore that fact and uses linear approach (because the manufacturer tells them to do so (?)). What do you think?
I'm sure you can guess what sensor it is, if you read some explanation in the OP
Don't take this badly, but I don't think that's good enough. There's no point at all in using a sensor as LPM if you don't know (numerically) it's spectral absorption characteristics, nor, as I understood from OP, it's heat/voltage relationship curve.
The main difficulty in building an LPM is calibrating it properly. If it isn't spectrally flat nor linear - from what I could gather this is neither - you'd need many data points at different powers and wavelengths to have proper calibration.
If you want to develop a proper curve compansation for it and the MCU hardware isn't good enough you could do it on the datalogging program.
If all you need is a rough power comparison tool you might as well buy a cheap TEC and attach it directly to an Arduino. Would cost less than $20. No point in using a fancy 22-bit ADC if the initial data isn't accurate.
As far as your comment on the Ophir heads go, they are spectrally flat according to the datasheet. Calibration is done at an specific WL to improve accuracy to 3%, but as has been demonstrated by dozens of comparisons on these forums it is quite accurate on the whole visible spectrum. If that graph is of the Ophir you will notice the difference in absorption on the 400-1064nm interval is less than 5%.
It depends on the application how important this is. No coating will be entirely flat in the whole visible range (let alone deep into IR), but this doesn't have to be a problem.
If it varies by say 5% in the 400 to 900 nm range, that'll be fine for most users, as long as it is consistent, and similarly linear over the power range.
A result that just says 'this laser is 100 mW +/- 5%' is still good if you compare between pointers or adjustments. You can, for example, determine that one type of lens gives an output improvemnt of 5% of it then reads '105 mW +/- 5%'.
I think most people will use the LPM to check that a laser is 'in spec', so something sold as 200 mW is actually around that power level (190 would probably be okay, 150 probably not).
Another thing that might be popular is for people to figure out how output power varies with input current. This also doesnt require high absolute accuracy as long as measurements are reproducible.
Main thing I am going to worry about is consistency within wavelengths.. if I test a 450nm vs 450nm I'd hope they test consistently.. obviously it'll be hard to compare exact numbers between 450nm and 520nm etc this isn't a 300$ Ophir head.. it's a 75$ setup and I feel that's pretty reasonable vs the 300 I was thinking of spending on abother usb lpm.
I will still be getting abother lpm later to compare to.. probably an Ophir head and custom setup but I will atleast have something that gives me ballpark numbers until then.
@Atomicrox:
I actually do likes when someone asking these. Because i've seen there is no question/discussion as detailed as this about other DIY LPM on their respective thread AFAIK. The user just accept it as they are.
And as you can see this has the most disclosed specs available :beer:
And what you are saying is actually a good things.
Especially as like you said there's no point at all in using a sensor as LPM if you don't know (numerically) it's spectral absorption characteristics.
But some things actually bothers me about all my reference sensor is the fact that i cannot find any data regarding its spectral response. The one i used for calibration sure consistent each other within their specs (two sensor with the black smooth surface, 150C-A-.1-AX). But clearly i observe that it reads higher than those gray coated sensor (like 20C-A.1-Y, Ophir 150C-A-.2-FLASH, or Ophir L30A-19mm-BNC). Honestly i don't even know what "FLASH" is, are they calibrated it using flashlight? lol
I've contacted ophir HQ (Israel) actually, but they never responded my email. Not sure why, but some said it's because my country's diplomatic issue (?)
Using ADC 22-bit itself is not an entirely fancy things for me
It's because i don't like using opamp which will introduce extra offset and/or creating more trouble over time, as semiconductor like that will change over time, e.g the offset or some drift over temp. And the important thing is, the cost for 22-bit ADC is nearly the same as those 16-bit ADC, but the 22-bit ADC has slower speed which is still reasonable speed for us.
And yes, you can use arduino (10-bit, 5V Vref), but for sensor 300mV/W for example, you'll get the resolution about 16.29mW per step.
TL;DR: I have tested the Hyperion sensor (coated) and compared it with my reference sensor, it's consistent at these common wavelength: 405nm, 445nm, 450nm, 465nm, 532nm, 635nm, 650nm.
