The interesting decision here sits on the emitting side. Ordinary white light leaves all the spectral work to the camera; this rig moves it into the lamp, stepping LEDs through 16 bands from near-ultraviolet to near-infrared and re-photographing the same section under each one, after which statistical methods combine the frames and pull out contrasts between materials [2][6]. The wavelength information lives in which frame you are looking at. That is the whole argument for illumination as the constraint, and it is worth taking seriously, because a better camera was an upgrade to the second-generation unit rather than the premise of the first [16].
It costs time. Sixteen bands implies at least sixteen exposures of the same face [2], and Kristiansen says the team only really understood what they had once the images were analysed back at the computer [8]. So the goal of work that is "faster, clearer and better in the field" [17] is not really about speed with a trowel in hand. Since a removed layer cannot be put back [15], the value lands where the section is still standing: documentation, and choosing where to take samples for dating and other analyses [18].
The cost figure is the softest number in the account. Roughly 20 times cheaper than the hyperspectral systems previously used for comparable archaeological work, per the researchers themselves [13], puts the prototype at about 5 percent of the comparator [1]. That is a ratio, and the actual price a museum would pay still isn't published. Whether that lands inside an excavation budget or a research grant depends on the absolute figure, which the write-up does not give, and on what a field unit already owns.
The bone result is the cleanest thing in the study. Under ultraviolet illumination, small bone fragments fluoresced and separated from the soil around them [10], and the team got usable signal without the optical filters that fluorescence photography normally needs, which removes a piece of kit from the field case [11]. Earlier work has tied fluorescence in archaeological bone to collagen preservation, and collagen matters when you are picking material for ancient DNA and protein analysis [12]. The study stops short of claiming the system can select those samples for you [14]. Treat it as a triage step that narrows candidates for a person to choose from.
The Sorte Muld test leaves the false-positive rate unmeasured. This was one site, tested with one prototype on one section covering more than a metre of dark deposits, where the layers are thin, irregular and nearly identical in colour [4][5]. The candidate old ground surface, which would mark a break in activity, is flagged by the team as unconfirmed [9]. Nothing in the design as described weighs a boundary the statistics find against a boundary an independent method agrees exists. My view: the illumination-side argument is the durable part of this paper. The number that would settle it has not been reported yet.
