In a paper cited in comments to this answer to In the 1950's how were radio-astrometric positions with portable dishes so precise they could be assigned to their dim optical counterparts (Quasars)?
The 1963 paper is Accurate Measurement of the Declinations of Radio Sources (click the little PDF icon) and the two excerpts containing the phrase "crystal mixer" are quoted below along with the block diagram.
When I hear "crystal" I think of either quartz crystals for oscillators, or metal crystals (e.g. galena) with oxide surfaces used to make diodes using cat's whisker contacts, but for diode detection in the mixer certainly both vacuum tube and solid state semiconductor diodes existed in the 1960's, though I don't know about the performance at 960 MHz.
Question: What would a "crystal mixer" have been in an 1960's radio telescope at 960 MHz?
Block diagram of an astronomical interferometer made from two radio telescopes separated by 200, 400 or 1600 feet and operated at 960 MHz. The mixers in question are shown just below each dish:
"Crystal mixer" appears twice in the paper; on page 2:
The arrangement of the components of the receiving equipment is shown by the block diagram of Figure 1. The receivers were of the superheterodyne type, and the crystal mixers were connected by short lengths of cable to the antenna feeds without any pre- amplification at the signal frequency. The local oscillator frequency was 960 Mc/s, and the center frequency of the IF amplifiers was 10 Mc/s, with a band width of about 4 Mc/s. No attempt was made to reject the image response of the superheterodyne. Note that the IF amplifiers were split into two sections. The IF preamplifier was located at the prime focus of the paraboloid along with the mixer and amplified the signal sufficiently to allow it to be fed through a long connecting cable to the remainder of the receiver, which was located in the laboratory building.
and again on page 3:
The local oscillator power for each half of the receiver was supplied by a separate klystron oscillator, which was phase-locked by a closed-loop servo system to reference signals of a common, central origin. The high-frequency refer- ence signal power required by the system for a satisfactory lock was about six orders of magnitude weaker than the available local oscillator power required by the crystal mixers of the superheterodyne receivers. The problem of getting phased local oscillator power to the two antennas when they were being used at large separations was thus greatly simplified.