Milton Taylor wrote:
How does the argyll software go about determining the extremities of a printer's gamut?
By examining the profile.
test charts are generated, unless it manages to generate a full range of RGB values that correspond to the printer's most saturated colours, I can't see how it would know?
It doesn't need to know anything about the response of the device to characterize it. The default chart explores the device values full gamut. Any color has to be reachable that way. The recommended chart algorithm evenly spaces test patches in device space. This turns out to be a robust and reasonable approach when nothing is known about the device response, and produces results that are hard to improve on, even if something is known about the device.
correspond to the most saturated colors could not be known, I imagine it would be easy to end up with very few patches that cover certain parts of the gamut that might be the critical parts, especially if the device gamut was rather non-uniform.
Possibly, if the devices gamut maximum points do not correspond to the device maximums. Such a device is not very well behaved though, and can be difficult to impossible to profile. As long as such non-monotonic response is at the end of each channel response, Argyll should cope (but it's not something I've verified in detail).
The practicalities of profiling devices with unknown behaviour are that sampling is the only way of gathering information. There are ways of improving the sampling using a more iterative approach (see the target -A parameter), in practice I'm not convinced yet that they make a huge difference.
I have since found that the profile I generated...which used three A4 sheets, i.e. 3000 odd patches, is still not as wide as the printer manufacturer's profile for the same printer. (I used the VRML viewer to compare the two). In particular, my one falls short by quite a way on the saturated greens and some of the blue hues.
It's hard to make judgements about such things. The difference between a real instrument and using a scanner could easily explain such a discrepancy. (Personally, I wouldn't have tried more than about 1 sheet using a scanner as an instrument). Your printer and the one the manufacturer profiled might not be operating identically. I've also come across manufacturers profiles that simply seem to have been manipulated, and their colorimetric table values didn't seem to be "real".