Beware of the bullshit!
There seems to be a lot of misunderstanding regarding the shift times that are achievable when using a semi-automatic shift system. Most of this misunderstanding comes about because of the exaggerated claims made by some suppliers at the bottom end of the market.
Let's get this clear right from the start - claims of shift times of 20ms or less belong in the land of fantasy! We've seen these claims made by one supplier on their website, yet when we posed as a potential customer and requested data log evidence, not surprisingly, no evidence to substantiate these claims was forthcoming. So, if they had data log evidence to prove their claims then they would surely make it available, and if they don't have any evidence then where did they get the figure of 20ms? Draw your own conclusions...
The fact of the matter is that a gear shift takes as long as it takes, it's highly variable. If the gearbox is not under any load and the dogs happen to be in alignment (as on a bench for example) then it may be possible to make a shift in 20ms or so, but when the gears are rotating and transmitting torque it becomes a very different story. The effects of transmission wind-up prevent the drive dogs from disengaing immediately, and it takes a period of engine cut before the shift lever will even begin to move. Once the dogs have disengaged, the shift will usually progress very quickly, but the shift time should be measured from the start of the sequence when the engine is first cut, not the time it takes the lever to move once it's started. The actual shift time depends upon the amount of transmission wind-up and the rotating inertia of the engine. The responsiveness of the engine actually has a far greater effect on the shift time than does the gearbox design or method of actuation.
Even if it was possible to change gear in as little as 20mS, it would not actually be desirable because of the increased risk of transmission failures. When we make an upshift, the engine speed needs to reduce to match the higher gear. Depending upon the gear ratios and the engine speed at which we shift, the drop in engine speed may be in the order of 2000rpm. Taking things to extremes, let's imagine the scenario where a shift was performed in no time at all (0mS). When the dog-ring engages with the higher gear, the engine would be spinning 2000rpm too much. The inertia of the car will then force the engine speed to reduce, and a massive shock loading will go through the entire transmission - and the engine crankshaft. Without effective management (in the form of compliant drive couplings or quill shafts), this shock loading would almost certainly result in transmission or engine failures.
The ideal shift time is very closely linked to engine intertia. Too fast and the engine will not have chance to slow down. Too slow and the engine speed will fall to much. Both scenarios result in increased shock loadings. We have found that pneumatic operation, along with precisely controlled engine cuts, results in the optimum shift time. This is borne out by our data log evidence. You can see from the sample log that when we make an up-shift, the engine speed falls at the correct rate to exactly match the next gear. If it didn't then you would see some nasty 'ringing' and spikes on the RPM trace as the next gear engaged.
Finally, a lot of weight is placed on shift times in terms of improved vehicle performance. Exaggerated claims of savings per gear shift are common. This, generally speaking, is yet more bullshit! If a semi-auto shift was performed in, say, 50mS, compared with a manual shift taking perhaps 100mS, then it's easy to make the assumption that each shift saves 50mS of track time. Extrapolating this, you may be lead to believe that during a typical lap with 20 up-shifts, you will save 1 second per lap. This is not the case, for it would assume that the vehicle came to a standstill during each shift, which of course it doesn't. People making such claims are either deliberately trying to mis-lead, or do not have even a basic understanding of mathematics! Any increased performance of a semi-auto shift system comes about due to an increase in the time that the car is accelerating rather than coasting. In this example, if the car was accelerated from a standing start up to 6th gear, it would be accelerating for 1/4 of a second longer than the same car with a manual shift. This not a track time saving of 1/4 second! In reality, unless you have a car that is capable of extreme acceleration, the improvement is hard to measure.