Science-Surplus Spectrophotometer Review

[∞] Science-Surplus sells reformed Spectrophotometers initially used as medical devices. Their spectrophotometer is a SMA compatible fiber-coupled compact unit based on a Czerny-Turner design with a 1800 lines/mm reflection grating. It features a Sony ILX511 linear CCD sensor sampled by a 16-bit ADC. The spectrophotometer ships with a 5V power supply (US plug), a null modem cable and a short SMA ended fiber. I got my unit for $199 on e-bay directly from Science-Surplus with an additional $70 international shipping cost and 52€ custom charges for Belgium. I had to add an extra 10€ for a US->EU plug and 25€ for a RS232->USB converter.

In this review, I will compare the spectrophotometer sold by Science-Surplus with [»] the one I’ve built from Thorlabs parts.

Clearly, the spectrophotometer from Science-Surplus is way cheaper than the version I proposed (about 1200€) even with all the additional charges. I guess this has to deal with the fact they use reformed medical devices because the mirror and grating they are using costs much more than the price asked when bought new.

My first trials revealed that the spectrophotometer was blind on a large portion of the spectrum. When opening the unit, I realized that this was due to some plastic scraps stuck on the CCD sensor. After blowing a bit a dry air on it, the whole spectrum range opened to me. The inside of the spectrophotometer (after removing the cover plate) is shown on Figure 1. It is a Czerny-Turner design where the light comes out of the fiber (on the right of the picture) through a SMA-threaded connector. The light passes through a slit (integrated in the SMA connector) and is collimated by a first mirror on a 1800 lines/mm grating element. The light spectrum split by the grating is then imaged on the CCD sensor using a second mirror. All the elements are kept in place by sets of screws that you may loosen to adjust the elements position.

The unit I got was pre-aligned but the 200 nm span was not the one I was looking for (it recorded the blue-green region while my experiments require me to study the green-red region). I then took a few hours to re-align it and play a bit with the Czerny-Turner setup. And believe me, it was a real pain to deal with the alignment procedure even after several hours of practice: the CCD sensor is extremely sensitive and even in a dark room with a piece of clothes on top of my hands I could barely record a non-saturated spectrum. It makes tuning extremely difficult because once the mirrors or the grating are in positions, you have to find the locking screws in the dark hoping that you don’t make any single move on the element you are trying to adjust. And that is the second major trouble with the alignment because any slight variation in the positions of the mirrors will dramatically impair your spectrum. This has probably to do with the fact that the system is not apertured and so very sensitive to focus changes. Finally, don’t even try to change the imaging mirror tilt angle or you will lose your spectrum and spend two hours trying to recover it! The CCD sensor pixels height are extremely small (0.2 mm) and it is very difficult to have an horizontal spectrum that covers the whole sensor area. On that point, the spectrophotometer design I’m using is much easier to align.

Concerning the spectra, a Neon sample is given on Figure 2 compared to the custom spectrophotometer spectrum region I had on disk. Basically, the two spectra are comparable in terms of resolution apart from a region on the right of the spectrum where the custom version lens begins to introduce aberrations. This is delicate because they are using a 1800 lines/mm grating while my setup used only a 1200 lines/mm grating. In theory the Science-Surplus spectrophotometer should have a better spatial resolution but, in practice, it doesn’t. Even after hours of alignment I did not manage to do any better than this and the fact that Science-Surplus mentions a 1 nm resolution (compared to a <0.4 nm in theory) lead me to think that their setup is working below optimal conditions. Since the alignment does not help and the sensor doesn’t seem to be low-pass coated, I believe that the limiting element might well be the entrance slit.

Also, they mention a 16-bits (1:65536) ADC with an average RMS noise level of 50 counts. This is true but should be understood with care. During my experiments, I recorded a dynamic range (ratio between the highest signal and the lowest one) of about 1:32 compared to 1:74 for my camera-based setup. This means that, in practice, it is not possible to distinguish two signals whose relative intensities are larger than about a 30 folds. This is about half of what I got with my CMOS 8-bits camera and a tenth of what you can do with Thorlabs $1500 dedicated sensor. For quantitative analysis, this is going to be problematic. After looking at the CCD sensor datasheet, it appears that this is clearly due to the integration time that is bound to a minimum 50 ms for the Science-Surplus spectrophotometer. Sony quotes a ~1:260 dynamic range when using an integration time of 5 ms. Longer integration times make larger dark signal levels and lower dynamic ranges. On that point, we can then consider that the electronic the are using is a bit too lazy and could take high benefits of upgrades.

On the other hand, I have also found the sensor to be extremely sensitive even with at the lowest integration time. Added to the fact that the dynamic range is quite small, it makes the sensor saturate with a lot of common source light and neutral density filters may quickly become necessary when using this spectrophotometer. But this sensitivity may also be a good thing to quantitatively detect extremely faint signals such as in Raman experiments. So it can be advantageous depending on the kind of experiments you are running.

Still concerning the sensor, I have noticed a few hot pixels (see Figure 3) and some disparity between pixel sensitivities. The “noise” you can read on Figure 3 is actually not fluctuating in time but only the difference in sensitivity between the various pixels. I should also mention that the relative peak intensity do not match the result I got from the custom spectrophotometer. Since the sensor seems to have been designed essentially for bar code scanners, you should pay attention to the eventual non-linearity of the response. But since I have already recommended to keep the unit for quantitative analysis, this should not be much of a trouble anyway.

Finally, the software is quite nice and works well so you should not have any problem with it. The only thing that I would have liked to see would be an automated procedure for the calibration of the wavelengths for known source. So far, you have to export the spectrum in Microsoft Excel (or any other equivalent software), find the peaks, assign them wavelength using a database and make a 3^rd order polynomial fit that you can write in the Science-Surplus software. It is not a big issue still since you don’t have to do it often.

In conclusions,

Since both last issues could be fixed by an electronic upgrade, I highly encourage the vendor to have a look at it because this could dramatically raise the spectrophotometer value. Even though, this spectrophotometer is of great interest and, if you are experimenting with light spectra, this is a must-have!