How to check an array spectrometer

Authors: Pedro J. Aphalo, T. Matthew Robson and Jonna Piiparinen.

University of Helsinki, Department of Biosciences.


Yesterday (4 April 2013) we did a quick comparison between two spectrometers from Ocean Optics. They were:

1) A Maya 2000 Pro, optimized for UV measurement and with a very recent calibration .

2) A USB2000+ factory calibrated one year ago, which the users had started suspecting that was out of calibration. We must stress that this instrument was not being used for measuring UV-B radiation, only UV-A radiation.

As we were doing the comparison anyway, we decided to estimate the size of the errors in  effective UV-B irradiance estimates made using such a standard instrument with a normal calibration. We also wanted to demonstrate how the functioning of a spectrometer can be roughly checked just by measuring sunlight under clear sky conditions.

We had no replicate instruments, neither replicate locations, dates or times of day, so this should be taken just as an example of how calibration errors and especially stray light and noise can lead to totally meaningless UV-C and UV-B estimates when using single monochromator spectrometers to measure sunlight. Even the “cleaning” of the data, for example forcing to zero spectral irradiance wavelengths that are known to be zero from more reliable measurements and/or the literature, is not enough.

[UPDATE, 2 June 2013] On 31 May we compared the same two instruments after the USB2000+ was recalibrated by Ocean Optics.

Materials and methods

The measurements were done in Viikki, Helsinki (60.226295 N, 25.017178 E), on 4 April 2013, at 10:45 to 10:54 (local summer time), 7:45 to 7:54 GMT. Scans are single scans, with no boxcar or similar averaging. Ambient air temperature was well below 5 C, and the spectrometers were exposed outdoors for about 15 min before starting the measurements.

[UPDATE, 2 June 2013] The second test was done only a few meters from the place used in the first test, on 31 May 2013, at 10:55 to 11:17 (local summer time), 7:55 to 8:17 GMT. Scans are single scans for the Maya, and the average of about 50 scans for the USB, as the integration time was much shorter for the USB, with no boxcar or similar averaging.  Ambient air temperature was well above 15 C, and the spectrometers were exposed outdoors for about 45 min before starting the measurements.

The spectral resolution of the Maya 2000 Pro as configured is better than 1 nm. This spectrometer has a 10 um slit with small aperture (“10s”), and an #HC21 grating, a Hamamatsu S01420 Back-thinned, 2D detector with 200 nm OFLV filter. We used a QP400-2-SR 400um Premium Fiber, solarization-resistant, 2m long, and a high quality Bentham cosine diffuser D7-H-SMA. Each measurement with this spectrometer consisted of three scans: one measurement scan, one dark scan, and one scan with an UV absorbing filter. Raw data from these three spectra are combined and also a correction based on the slit function (measured with lasers during the initial characterization of the instrument) are used to get a spectrum with an improved spectral resolution and signal to noise ratio.

The spectral resolution of the USB2000+ as configured is less than that of the Maya. The detector used in this spectrometer is a Sony ILX511. Each measurement consisted in a dark scan and a measurement scan, the dark scan was subtracted from the measurement scan and the one year old factory calibration applied directly with the SpectraSuite software from Ocean Optics.

A TUV simulation was run for 7:45, 7:51 and 7:54 GMT and 410 du of ozone, the available from NASA for 4 April 2013 and Southern Finland.

Three sets of scans were measured within 9 minutes starting at 7:45 GMT, and within each pair they were measured simultaneously. The local (summer) time is shown in the column headings of the tables. [UPDATE, 2 June 2013] On the second test more scans were measured with the Maya (6) than with the USB (2). One additional measurement was done with the Maya using “bracketing” (a scan with an integration time giving around 55000/64000 counts, and a second one with ten times the value of the first integration time). For each measured scan a new dark scan was measured, and for the Maya also a “filter” scan.

The weather was perfect for such a test, without a single cloud or any haze. We did the calculations using the whole of the spectral data, or after setting to zero irradiances for all wavelengths shorter than 290 nm or those shorter than 293 nm, the known tail of the solar spectrum at this latitude and time of the year, corroborated against the simulated spectrum from the Quick TUV Calculator. [UPDATE, 2 June 2013] On the day of the second test the sky was overcast, consequently no comparison to the TUV simulator was done. Instead, PPFD was simultaneously measured, with a LI-190 quantum sensor from LI-COR (Lincoln, Nebraska, USA).


The table below shows effective irradiances and also unweighted irradiances and some photon ratios, for each scan, as calculated after setting to zero all spectral irradiance values for wavelengths shorter than 293 nm. This step reduced the errors for the USB2000+ very significantly, and for the Maya 2000 Pro, slightly as they were already much smaller. With this approach the Maya 2000 Pro can be used even for estimating effective UV-B doses, with some uncertainties in the case of the generalized plant action spectrum of Caldwell. These uncertainties could be mitigated by repeated scans, and averaging. The USB2000+, if factory recalibrated, should perform satisfactorily for measuring UV-A radiation, but not for measuring UV-B radiation.

Radiation quantities calculated from measured irradiances for wavelenths longer than 293 nm.

The table below shows effective irradiances and also unweighted irradiances and some photon ratios, for each scan, as calculated after setting all spectral irradiance values for wavelengths shorter than 290 nm to zero. This step reduced the errors for the Maya 2000 Pro almost as much as using 293 nm, while for the USB2000+ none of these wavelengths gave usable estimates of UV-B irradiance. There is not much difference compared to the table above for the Maya 2000 Pro, but the performance in the UV-B of the USB2000+ is further degraded.

Radiation quantities calculated from measured irradiances for wavelenths longer than 290 nm.

The table below shows effective irradiances and also unweighted irradiances and some photon ratios, for each scan, as calculated with the whole of the measured data, including the noise in the UV-C band. In this case, neither of the two spectrometers performed satisfactorily for measuring effective UV-B irradiance. For unweighted UV-B irradiance, the Maya 2000 Pro gives reasonably good estimates, while the USB2000+ produces erroneous estimates.

Radiation quantities calculated from measured irradiances for all wavelenths measured.

By comparing the three figures below, one can clearly see the difference in performance of the two spectrometers in comparison to the simulated spectrum (which is limited to the UV band). The Maya 2000 Pro gives, using the correction protocol, a spectrum that is much closer to the expected one than the one measured with the USB2000+ without using any special correction.

Spectrum measured with the Maya 2000 Pro at 10:54.

Spectrum measured with the USB2000+ at 10:54.

Spectrum calculated with the QUick TUV Simulator at 10:54.

[UPDATE, 2 June 2013]

After the Maya2000Pro travelled around five European countries during a field campaign, and the USB2000+ was recalibrated at Ocean Optics.

Radiation quantities calculated from measured irradiances for wavelenths longer than 293 nm.

Radiation quantities calculated from measured irradiances for wavelengths longer than 290 nm.

Radiation quantities calculated from measured irradiances for all wavelengths measured.

Comparison of two Ocean Optics spectrometers, both recently calibrated.

Comparison of two Ocean Optics spectrometers, both recently calibrated. y-axis with logarithmic scale.

Discussion and conclusions

1) One can check the performance of a spectrometer on a spring or summer day with a perfectly clear sky by comparing its output and the derived quantities to those simulated with a model like the TUV Quick Calculator. Of course, you need to enter the correct GMT time (do not forget that summer time affects only your local time, not GMT) and a good estimate of ozone column depth. This type of check is very easy to do as long as you are lucky with the weather conditions.

2) Single monochomator spectrometers (such as all available array spectrometers) need very careful calibration, plus complex corrections for stray light and slit function to be usable for UV-B measurements in sunlight. In some cases, as was the case for our Maya, the optical bench also had to be adjusted for optimal performance in the UV and the aperture had to be replaced with a smaller one to reduce stray light. These adjustments were done at the Ocean Optics factory. After all this work the Maya 2000 Pro is usable for UV-B measurements, but still its performance is not as good as that of a double monochromator instrument.

3) Array spectrometers, can, just like any instrument, get out of calibration in a relatively short time, as was the case with the USB2000+. One possible reason for drastic changes in calibration is damage to the optical fibre. Even if one cannot do frequent calibrations on an optical bench, checking the measured values against model output or another instrument, or some other known light source should be done regularly. On the contrary, the Maya 2000 Pro, showed during the recent calibration in an optical bench, that the calibration had changed by only about 2 % after one year of intense field use.

4) Including irradiance data from short wavelengths causes large errors in the calculations of effective doses using BSWFs that extend into the short UV-B, and UV-C bands. This is catastrophic in the case of the USB2000+, but also causes rather large errors in the case of the Maya 2000 Pro optimized for UV-B measurements.

5) Irradiance changes quickly through a day, even under clear sky conditions, as can be seen in our measurements and simulations spanning only 9 min.

6) By plotting of a spectrum of sunlight, with spectral irradiance on a logarithmic scale, one can approximately assess whether a spectrometer is working as expected, and whether it suffers from an unusually large stray light or electrical noise problem. In this case the Maya 2000 Pro, is giving almost 4 decades between highest peak and noise floor, except for an exceptional noisy pixel. In contrast the USB2000+ is giving only about 2 decades between the highest peak and the noise floor, and the noise floor increases towards shorter wavelength. (It should be possible to use the USB2000+ for measuring UV-B in sunlight if all the same type of calibrations and corrections as used with the Maya 2000 Pro are used with the addition of integration time bracketing. This last step may be needed because its detector has a smaller dynamic range.)

7) [UPDATE, 3 June 2013] In the comparative figure from the second test, one can see a change in the mismatch between the measurements done using the two instruments at around 400 nm. This is likely to be the result of using two different standard lamps for the calibration. This the usual procedure because of the limited emission spectrum of lamps with good output stability.

8)[UPDATE, 3 June 2013] After the recalibration of the USB2000+, the errors are within what can be expected from normal calibration procedures.

If possible, we will repeat this test after the USB2000+ is recalibrated. [UPDATE, 2 June 2013]  Done on 31 May 2013.

Acknowledgements: We thank the staff of Ocean Optics for all their help in adjusting the Maya for best performance under our measuring conditions, and Lasse Ylianttila for the characterization, calibration and development of the correction algorithm for the Maya. The purchase of the Maya and the development of the corrections for stray light and slit function, was funded by the Academy of Finland (grant to PJA) and this instrument is the property of the Department of Biosciences, University of Helsinki. The purchase of the USB2000+ was funded by the Walter and Andrée de Nottbeck foundation and is the property of Tvärminne Zoological Station, University of Helsinki.

Further reading and references:

Quick TUV Calculator,, used on 6 April 2013.

The source of the ozone data from NASA,

Ocean Optics website with literature on the spectrometers used is at

Aphalo, P. J.; Albert, A.; Björn, L. O.; McLeod, A.; Robson, T. M. & Rosenqvist, E. (eds.) 2012. Beyond the visible: A handbook of best practice in plant UV photobiology. COST Action FA0906 UV4growth. Helsinki: University of Helsinki, Department of Biosciences, Division of Plant Biology. ISBN 978-952-10-8362-4 (Paperback), 978-952-10-8363-1 (PDF). xxx + 176 pp.

Kreuter, A. & Blumthaler, M. Stray light correction for solar measurements using array spectrometers. Review of Scientific Instruments, AIP, 2009, 80, 096108

Ylianttila, L.; Visuri, R.; Huurto, L. & Jokela, K. (2005) Evaluation of a single-monochromator diode array spectroradiometer for sunbed UV-radiation measurements.
Photochem Photobiol, 81, 333-341.

Leave a Reply

Your email address will not be published. Required fields are marked *