Helena - if you increase the media density you will change your cut point, yes this will reduce the mass of material reporting to sinks, but it may also affect your recovery. You need to make sure that you are cutting at the correct cutpoint for your operation.

You mention operating parameters - have the vortex finder and spigot diameters been checked to ensure that they are in spec? Is the media free of contamination, and operating at the correct viscosity? Have you recently changed supplier or made up a new batch, which may have changed medium particle size? Any changes in upstream processes that may have changed the particle size in the feed?

Have you done or do you have the ability to do tracer tests to check how the cyclones are operating? If the tracers deport correctly and you are confident that you are operating at the correct cut-point then something else must have changed - speak with the geologists too if you can to see if they have seen a  change in plant feed. 

Is the material in the underflow there because the cyclone is operating badly (misplaced) or because you are treating an ore with increased heavies. Check the SG distribution of the underflow material.
Are there any indications of increased medium viscosity?

Generally speaking, increasing density is not a good practice, every case is different but higher density will worsen the separation efficiency. Check what are the operational specs of the cyclone and if you don't have it, ask the supplier.

Maybe you should increase the medium density to reject a higher proportion of light minerals to the overflow. But as with everything there is always a limit to what can be done effectively, as previously says increasing density too much carries the risk of increasing viscosity which impedes separation. Increasing underflow could also be caused by higher proportion of heavies in feed, and /or coarser size distribution of feed.

Increasing the medium density will definitely affect your recovery, maybe the feed material density has varied from the previous you was treating and in cases like this you would need to do the floats and sinks test works so that you have an idea of the cut point in the processing plant.

You may also need to check the density of the medium in the overflow and underflow of the cyclone, If there is a difference exceeding 0.5, then there is need to troubleshoot among the components that makes up the cyclone as well as medium contamination.

If cyclone feed pressure and operating density are within acceptable ranges, you might want to have a look at the media density differentials. By measuring the feed and overflow stream densities you can calculate the density differential. The differential will give you a very good idea as to the condition of your in-circuit media, and it will show if the issue is either media stability or viscosity. High media viscosity leads to low cyclone differentials which in turn can result in an increased recovery of concentrate to cyclone underflow. Don’t forget about media (FeSi) rheology- high media viscosity can lead to an increase in concentrate volume. So check the cyclone density differentials.

Check the cyclone density differentials - feed, U/F and O/F. You may be right at the lower limit of the acceptable range and indicates increased viscosity in the system. Are you checking medium stability on a regular basis i.e. manual settling rates in a measuring cylinder. This will give you a relative indication of whether viscosity/stability has increased. Also are the spigot and V/F finder checked on a regular basis for wear - larger spigot, more sinks. Is there a feed prep screen ahead of the DMS to remove fines/slimes - poor screening efficiency and therefore increased fines/slimes in the dense medium will increase medium viscosity.