Some FM tuners provide stereo noise reduction with a switch labeled hi-blend. This function progressively blends the left and right audio channels at higher frequencies. Since stereo noise originates almost entirely in the L−R subchannel, reducing L−R injection, and thus channel differences, lowers it. Separation at lower frequencies remains mostly intact. The main drawback of hi-blend is an unnatural smearing of the acoustic image as instrument harmonics collapse to the center while their fundamentals stay put. In addition, hi-blend does not reduce low-frequency noise, which predominates in stereo, and it rolls off the overall spectrum.
This is a typical hi-blend circuit. The FET switch interconnects the left and right signal paths with 0.01 µF. The capacitor and the differential source and load resistances form a lowpass filter that reduces interchannel differences at higher frequencies.
This is channel separation in dB for the circuit above with no load. Since no hi-blend standard exists, the curve will vary among receivers.
Replacing the blend capacitor with a resistor converts a hi-blend circuit to a flat blend that combines the channels uniformly with frequency. Flat blend narrows the width of the acoustic soundstage, positioning the instruments closer to center, but their spatial images are not unnaturally smeared, the spectrum is not rolled off, and low-frequency noise is reduced. The main drawback of flat blend is that it may provide less noise reduction for a given soundstage width, which is determined primarily by instrument fundamentals. If you do little listening midway between the speakers where the stereo image coalesces, you may be better off with hi-blend.
Usually I install a 10kΩ trimpot as the blend resistor and adjust it for 9.4 dB of separation. This small figure provides a surprisingly good stereo image along with 6 dB of noise reduction, a noticeable improvement. The following table gives other trade-offs between noise reduction and channel separation. The soundstage width figures are for the speakers subtending a 60° angle and are based on Fig. 16 here.
Noise Channel Soundstage Reduction Separation Width 3 dB 15.2 dB 74% 4 12.8 67 5 10.9 60 6 9.4 53 7 8.2 48 8 7.1 42 9 6.2 37 10 5.4 33 11 4.7 29 12 4.1 26