The electro-optic effect has a potential bandwidth of a several THz. The SI-traceable primary standard for the time-domain voltage impulse response \(v(t)\) is based on this property. Electro-optic sampling (EOS) systems, realized in several national measurement institutes worldwide, typical have bandwidths between 200 GHz and 500 GHz. The EOS principle and key design criteria for the PTB EOS system are described together with the sources of impairment and the necessary de-embedding to correct these errors. We describe the key results from a recent international intercomparison of EOS systems, using a 100-GHz photodiode transfer standard. The NMI participants’ EOS systems are different designs, independently developed. The intercomparison results provide confidence in the agreement between the systems. Dissemination of the calibration from the EOS primary reference to industry and academia is an essential part of the traceability chain, providing confidence in the results. We describe the advantages, limitations, and impairments associated with different choices for transfer standards with coaxial connector geometry.