A Fairchild FST3125 used as high-level mixer.
Pat Hawker, G3VA, reported in his column 'Technical topics' in RSGB's RadCom of September 1998, that Colin Horrabin, G3SBI had tested a high-level mixer based on a Fairchild FST3125. This chip is designed to be used as low-cost fast bus-switch and has four independent switches.
This combination of switches can very well be used as a mixer and Colin reported that this mixer had an extremely high IP3 of +44dBm and a moderate conversion loss of 5dB.
The FST3125 switches are controlled with a frequency squarer using a 74AC86, needing a local-oscillator level of at least 0dBm.

I constructed three variants of this mixer. Number one was made using Mini-Circuits broadband 1:4 transformers type T4-1. The transformers are indeed very broadbanded but also quite expensive. Number two was made with broadband transformers using ten trifilar windings on Amidon FT37-43 (mu =850) cores. Number three was made using eleven trifilar windings on low-cost small balun type cores with mu=750 (type 11133203). I do not know who makes these cores (anyone?).

Of these mixers, the conversion loss and IP3 was determined by Cor, PA0CHN. He used two crystal oscillators at frequencies of 3.730 and 3.780 MHz. These two signals were amplified, attenuated, combined, attenuated and amplified to a level of 0dBm. We determined that this setup had internally generated IP3 products of better than 75dB below the base frequencie levels. The output of the combiner was connected to the RF input of the mixer. A signal generator with a frequency of 12 MHz and output level of 0dBm was connected to the LO input of the mixer. The IF mixer output was connected to a spectrum-analyser. The conversion-loss was determined by measuring the difference in signal level of the RF input an IF output at 12-3.730 MHz. IP3-out was determined by taking half the difference in signal level of RF input and IP3 products at the IF output. (This way of determining IP3 depends on having an RF input level of 0dBm, see ARRL handbook).
IP3-in can than be calculated as the sum of IP3-out and conversion-loss. Results were:

mixer	transf.		IP3-in		conversion-loss
  1     T4-1		+41 dBm		6 dB
  2     FT37-43		+43 dBm		6 dB
  3     11133203        +44 dBm		6 dB

These results agree quite well with what Colin reported. The mixer using the low-cost balun cores had the best performance. If you are willing to spend some time, you can easily construct a 'bullet-proof' mixer using low-cost components.
A fellow-ham, PA0MWU, is already using this mixer in an experimental HF transceiver and he is very satified with the results so far.
FST3125 look alikes are also produced by Texas Instruments and Pericom.
Send your comments (email address is on my home-page)