TMT has defined the accuracy to be achieved for both absolute and differential astrometry in its top-level requirements documents. Because of the complexities of different types of astrometric observations, these requirements cannot be used to specify system design parameters directly. The TMT astrometry working group therefore developed detailed astrometry error budgets for a variety of science cases. These error budgets detail how astrometric errors propagate through the calibration, observing and data reduction processes. The budgets need to be condensed into sets of specific requirements that can be used by each subsystem team for design purposes. We show how this flowdown from error budgets to design requirements is achieved for the case of TMT's first-light Infrared Imaging Spectrometer (IRIS) instrument.

IRIS is a near-infrared (0.84 to 2.4 microns) integral field spectrograph and wide-field imager being developed for first light with the Thirty Meter (TMT). It mounts to the advanced optics (AO) system NFIRAOS and has integrated on-instrument wavefront sensors (OIWFS) to achieve diffraction-limited spatial resolution at wavelengths longer than 1 micron. With moderate spectral resolution (R ~4,000 - 8,000) and large bandpass over a continuous field of view, IRIS will open new opportunities in virtually every area of astrophysical science. It will be able to resolve surface features tens of kilometers across Titan, while also mapping the distant galaxies at the scale of an individual star forming region. This paper summarizes the entire design and capabilities, and includes the results from the nearly completed preliminary design phase.

IRIS (InfraRed Imaging Spectrograph) is one of the first-generation instruments for the Thirty Meter Telescope (TMT). IRIS is composed of a combination of near-infrared (0.84--2.4 $\mu$m) diffraction limited imager and integral field spectrograph. To achieve near-diffraction limited resolutions in the near-infrared wavelength region, IRIS uses the advanced adaptive optics system NFIRAOS (Narrow Field Infrared Adaptive Optics System) and integrated on-instrument wavefront sensors (OIWFS). However, IRIS itself has challenging specifications. First, the overall system wavefront error should be less than 40 nm in Y, z, J, and H-band and 42 nm in K-band over a 34.0 $\times$ 34.0 arcsecond field of view. Second, the throughput of the imager components should be more than 42 percent. To achieve the extremely low wavefront error and high throughput, all reflective design has been newly proposed. We have adopted a new design policy called "Co-Axis double-TMA", which cancels the asymmetric aberrations generated by "collimator/TMA" and "camera/TMA" efficiently. The latest imager design meets all specifications, and, in particular, the wavefront error is less than 17.3 nm and throughput is more than 50.8 percent. However, to meet the specification of wavefront error and throughput as built performance, the IRIS imager requires both mirrors with low surface irregularity after high-reflection coating in cryogenic and high-level Assembly Integration and Verification (AIV). To deal with these technical challenges, we have done the tolerance analysis and found that total pass rate is almost 99 percent in the case of gauss distribution and more than 90 percent in the case of parabolic distribution using four compensators. We also have made an AIV plan and feasibility check of the optical elements. In this paper, we will present the details of this optical system.