No definition for `type` field enumeration in `NXSample`

[paulmillar]

Commit dc2710d1e adds the definition of the NXsample.

This definition includes a field type (of type NX_CHAR) that takes values from a controlled vocabulary. The valid values of type are:

• sample
• sample+can
• can
• calibration sample
• normalisation sample
• simulated data
• none
• sample environment

The most pressing issue is that the semantics of any of these terms is undefined.

This leads to a number of questions:

• What is a "can", in this context?
• Why is being conveyed by having a NXsample of type "none", compared to simply excluding any NXsample group?
• What is the difference between having a single NXsample of type "sample+can" and having two NXsample_component (one for the "sample" and one for the "can")?
• Why is "sample environment" a valid type of sample? This seems to be a different kind of thing.

On this last point, is it valid for an NXentry to have two NXsample groups, one of type "sample environment" and one of type "sample". If so, how does NeXus describe the link between them?

I'm also not sure that NXSample is the ideal place to indicate that the data is the result of a numerical simulation (perhaps NXentry would be better?). However, compared to the other issues, this is a relatively minor concern.

[paulmillar]

As noted by @prjemian in his comment on another issue, the choice of name for this field (type) is also problematic because it conflicts with other use in the NXDL language.

[prjemian]

In this context, can refers to the sample container. It comes from neutron scattering, such as a powder diffractometer, where the powder was poured into a metal can. For such instruments, a can is constructed from metal (such as vanadium) that is transparent (or very low absorption) to neutron radiation.

Here's an example picked at random from a web search:

The Si was loaded into a vanadium can and measurements were carried out ...

[paulmillar]

Thanks @prjemian, that makes sense.

I guess type=can indicates data taken with an empty can, while type=sample+can indicates the can was loaded with the (powder) sample, right?

Would an NXsample group with type=none provide a positive statement that no sample was present. By contrast. omitting the NXsample group would provide no information regarding the sample: maybe it was there, maybe not.

[prjemian]

I guess type=can indicates data taken with an empty can, while type=sample+can indicates the can was loaded with the (powder) sample, right?

right

[prjemian]

NXsample group with type=none

Such as a small-angle X-ray scattering measurement with no sample in the beam (common in metallurgy).