Transferring a conversation with RHL on email here for posterity. Quotes are from bick's original email, while replies are rhl's.

Clearly the spectra need to be sampled above the Nyquist limit, and it sounded like the issue was 'how much?'.

It would have to be the nyquist limit of the narrowest lines, not of the spectrograph. I am concerned that if we smooth the input spectra too much we'll never simulate the narrowest lines that we really see.

Jim suggested R=10000, Robert bid it up to R=20000 with the concern that we need to be quite a bit above critical to make sure we handle the narrow lines correctly. Jenny pointed out that narrow lines won't be a problem for galaxy spectra ... I believe her ... but there are other spectra to consider, of course.

R = 20000 corresponds to 15km/s so it should be OK for everything except stars. Maybe 10000 is enough? Anyway, we need a high enough resolution that the PSF for the narrowest lines we expect come from the instrument not the input model.

How well do we need to specify the data format at this point? Is this format intended to also be the 1D output format of the pipeline (with the exception that the output sampling is that of the spectrograph)?

No, I think we decided against that. The needs of the input are very different from the output (no need for noise or variance; higher resolution).

For spectra with linearized wavelength solutions (or log-linear), the sampling itself is just a WCS in the header, so I guess I'm not clear on why this detail came up as the focus of discussion. I had assumed the simulator would be able to accept input spectra at any reasonable sampling ... if it's under-sampled, the sim should produce the garbage we deserve.

Well, one option for inputs would be a set of sampled spectra + a line list (lambda, power, width). I'm not sure we need to go there. Or we could specify the input as (flux, lambda) tuples with no assumption that the lambda values are equally spaced — that allows high resolution in the lines and low in the continuum.

I don't think this is hard, we just need to define it.

N.b. This is basically a placeholder. I understand (phonecon 2016-10-07) that the LAM team are now responsible for adding emission lines to the simulated galaxy spectra, so it'd be great to get their input.

In the current datamodel.txt we have defined a pfsSimObject. This specifies at present only (flux, lambda). Equivalent width is not included. Should this ticket still remain as a placeholder? Should we revisit including the extra line parameters?