The following advice comes from Dave at Rallispec, one of the few people whom actually KNOWS differentials and I thank him for his contribution and advice as differential advice from people who REALLY know them is impossible to come by.
To add a couple points of information to this very cool thread:
1. Viscous LSD's are speed sensing. The work by reacting to speed differences between the two axles. They are limited in the time it takes them to respond and therefore they are not an ideal solution. They are designated by their torque reaction per difference in rotational speed. For example the stock unit is 4 kgf-m (torque reaction) / 100rpm (speed difference). Increasing this value increases how quickly it can respond because it will be able to redirect more torque for the same difference in shaft speeds. However, if the value is increased too much then a small difference in shaft speed (such as one that occurrs in a normal tight radius corner) will generate enough torque reaction as to create a noticeable understeer condition.
2. Gear type LSD's and clutch type LSD's are torque sensing. They respond to torque input and/or torque differences across the drive shafts. They do not need to be waiting for a speed difference (as in wheelspin) to be doing their job. These types of LSD's are designated by their bias ratios. Basically what ratio of the torque at the low traction side can be redirected to the high traction side.
3. Torsen, Quaife, and Helical LSD's are all very similar...all gear types that involve complex interactions between the gear teeth to do their job. Quaife and Helical LSDs both use helical type gears. Torsen uses a worm gear arrangement. These LSD's typically have bias ratios in the 1.5 to 4 range I think. However, most are not set up with any preload (although it is possible to build in some preload with these LSD's by shimming the end play of the gears against the case to a certain amount of interference fit). Therefore when the torque at one wheel is zero or very near zero (wheel off the ground or on ice) then the LSD cannot transfer any torque to the other wheel (anything multiplied by zero is still zero!).
4. Clutch type LSD's (Salisbury type...also known as plated diffs) use stacks of clutch plates alternately splined to the axle and the diff case. Pressure to apply the clutch packs comes from a set of pressure rings. The spider gear cross shaft rests on sets of ramps cut into the pressure rings. When torque is applied to the diff the shaft rides up the ramps and forces the pressure rings apart thereby applying the pressure to the clutch packs. Clutch type LSD's set up for racing generally have very high bias ratios (upwards of 6:1 I think). They also can be set up with significant preload if desired. The preload on the clutch plates ensures that the LSD can still do something even when there is no zero traction at one wheel.
5. The DCCD's open torque split is determined by the tooth arrangement of its planetary gear set. Unlike a typical diff's spider and pinion gear arrangement which gives a 50/50 split as long as available traction on both axles is equal, the planetary gears provide a torque split biased toward one end (be it 35/65 or whatever it is designed for) as long as available traction is equal. Attached to that planetary gear set is a clutch pack and a second clutch pack controlled by an electromagnet. The pressure on the clutch pack (which biases torque back toward a 50/50 split) is determined by the current flowing through the electromagnet. Therefore it is a clutch type LSD that can be electronically controlled to alter its operating characteristics.
In terms of tuning the clutch type LSD, your options are as follows:
A. Preload setting -- increasing preload increases its effectiveness in very slippery conditions. However, this preload must be overcome in order to go around a corner (which is why these LSD's when set up with any measurable preload will make noise). It will generate understeer. Preload is set both by a preload spring (some types use a bellville washer but Cusco's RS type uses coil springs) and also by the stack height of the clutch pack (the thicknesses of the clutch plates).
B. Ramp angles -- there are two ramp angles, one that influences on-throttle operation and one that influences off-throttle (decel) operation. When an LSD is described as 1way, 1.5way, or 2way they are talking about the off-throttle ramp angles. 1way types only offer LSD action on acceleration and have no ramps on the decel side. 2way types have equal ramp angles for both accel and decel. With 1.5way have steeper angles on the decel side vs. the accel side so it takes more torque to create the same pressure (basically on decel the LSD is working but less effective). Some manufacturers for some reason use other values such as 1.6way or 1.7way but they are just describing the relative angles from one side to the other.
C. Active friction surfaces (lockup percentage?) -- By altering the arrangement of clutch plates (outer splined vs. inner splined plates) you can alter the number of "active" friction surfaces in play. Reducing the friction surfaces reduces the bias ratio.
In terms of general recommendations....
Front LSD: for track, autox, and street use most people find that a gear type front LSD is most predictable and easiest to drive near the limit. It does not generate much understeer on corner entry due to the absence of preload and the low bias ratio under decel. A 1way clutch type LSD can be just as, if not more, effective when set up properly for the conditions and the driving style of the user. But if it is not set up well then it will cause difficulties for the driver. For slippery conditions such as rallyx, iceracing, rally, etc. it will be beneficial to run a 1.5way plated front LSD as it will help with braking stability. Preload should be set according to available traction and driver preference. For tarmac drivers having difficulty under braking it can also be worthwhile to try a 1.5way LSD with low preload.
Center LSD: I am very interested to try the PPG Torsen center LSD but have not yet had the chance. I suspect it may be an excellent track option. I would avoid using a 20kgf center viscous except for gravel, snow, or ice...or maybe straight line drag racing. On high speed tracks with no tight corners it may be OK but it will be counterproductive on tight tracks or autox. The Cusco Tarmac gear is worthwhile only when you have good front and rear LSD's to back up the fact that it is an open diff.
Rear LSD: There are very few instances for racing applications where I would not recommend a clutch type rear LSD. They are so tuneable that there is no reason not to use one (other than the need to rebuild it every season or every other season). The stock viscous unit used in the WRX and 2000-2001 RS models is essentially useless (the unit is so thin as its packaged off the side of a standard open diff and its capacity is very limited). The factory R180 clutch type LSD is set up extremely loose from the factory and should be rebuilt with appropriate preload and possibly different pressure rings if you want to use it for racing. For a street application the stock Torsen or a Quaife unit is usually preferrable from the standpoint of not needing to be rebuilt and creating very little noise.
I guess if anyone has any questions they can email me (please don't PM me I don't check them very often). I also have a somewhat complete JDM transmission chart if someone needs info on factory diffs, gear ratios, or whatever. Me and Jeff actually cross checked some of our info at one point.