FLAME, STS and SPARK spectrometer sensitivity comparison

Flame STS and SPARK (now discontinued) spectrometer sensitivity Comparison


The workhorse spectrometer for Ocean Optics has from the beginning been the “S” bench. There have been various models, S1000, S2000, USB2000+, USB4000, the JAZ, and now the latest, The FLAME spectrometer.  We have sold well over 300,000 units of these great miniature spectrometers.

Now we add 2 new additions to the offering, the STS spectrometer and the SPARK spectrometer. These are smaller and less expensive, but what’s the real differences?

The STS has a smaller version of the S bench, made by automated manufacturing techniques borrowed from the Telecom world. The SPARK has no optical bench, its a filter based system that’s smaller and even less expensive. Its also been dropped as a product by  Ocean Optics.

The heart of a spectrometer is the detector, though. The FLAME uses our standard Sony ILX511B CCD detector, or the Toshiba equivalent.

The STS and SPARK use a new CMOS detector, the ELIS1024 line scan image sensor from Dynamax.  This detector has some really nice features, built in electronics, ability to read pixels non-destructively, and it is very fast, small and low cost.

But how sensitive is it?

Basically the spectrometers separate light into wavelengths, detectors then convert photons to signals, which are reported to the software in completely arbitrary values called counts.

The FLAME seems very sensitive when you use it, because a low light level gives lots of counts. The SPARK and STS seem less sensitive, because they give fewer counts for the same signal.

Counts don’t count! All that matters is how many photons were sampled.  The signal to noise of your measurement will ultimately be limited by the Shott noise, equal to the square root of the number of photons. Of course you add extra noise from the readout of the signal of the detector, and unnecessary noise from poorly designed electronics. But for all of these spectrometers, Shott noise predominates.

The answer to the sensitivity question is revealed by experimentation, exposing the 3 spectrometers to the same light source and expressing the signal as signal to noise ration (SNR) compared to the integration time required to acquire that signal.  Sensitivity then is expressed as how fast can a detector (or spectrometer) reach a certain SNR.

The answer is surprising!

Flame SPARK STS Sensitivity comparison - Spectrecology
SNR vs Integration period for the FLAME, STS and SPARK spectrometers. Spectrecology

The ELIS1204 based Spark and STS are somewhat more sensitive than the FLAME!  What’s more, they have larger well depths and so can collect many more photons in one scan than can the Sony ILX511B. This is called dynamic range, and the ELIS has much higher dynamic range, making it usefull for high light applications that would saturate a FLAME spectrometer.

The cosmetic aspect that seems hard to overcome for users is for the SPARK and STS the signal displayed as a % of full scale is way lower than the FLAME. But as shown above, that’s a graphical scaling issue, not a performance issue.

To be sure, the FLAME optical bench provides a wider wavelength range with better optical resolution than the STS, and much better than the SPARK filter based system, but the relative costs of the Spark (~$500) and STS (~$1400) compared to the Flame (~$3000) warrant their consideration.