Prahlad mentioned that using a snubber will take care of EMI - this is not the case. The snubber is to prevent false firing of the triac/scr due to dV/dT (rapid change of voltage over time). With the semikron parts mentioned, these have a dv/dt rating of over 10,000V/uS (refer to the datasheet for actual numbers - these are off the top of my head!), so unless you are controlling a highly reactive load a snubber is not necessary - HOWEVER! a snubber also protects the drive circuit from voltage spikes, so I always have a snubber circuit. I've used semikron parts for years and they are near indestructible as compared to triacs. For EMI, the worst case condition is firing at 90 degrees where the mains is at peak voltage. The SCR will turn on in about 2uS, so the output rises from 0V to mains peak in about 2uS. This causes an emormous amount of interference and makes the filaments of the lamp buzz. So you need to put a choke (inductor) in series. This has the effect of slowing down the rise time. What value? The more the merrier! For 10kW I'd probably start at around 500uH. Put an AM radio nearby to check the effect. Note for 10kW the choke will get physically large. Check for filament buzzing at lower brightness levels. The chokes I normally are toroidal and use the yellow/white powdered iron compound (see www.amidon.com).

At 10kW we're talking a fair amount of current, so the SCR will need to be heatsinked. Assume a drop of about 2V across the SCR for your calcs. In reality it is around 1V.

As for the number of steps - it depends if you want to fade the lamp up and down smoothly. The eye can detect a phase change of about 15uS, so for smoothness you want a fade resolution of at least 8uS. Also for phase control (depending on the firing circuit) you can't expect to turn on at zero crossing as there is not enough current available to hold the SCR on, so you have to start firing at around 500uS past zero crossing and turn the drive off at around 500uS before zero crossing. This assumes you have a 'hard' firing circuit that applies trigger current to the SCR for the duration of the firing as opposed to pulse firing where you give a trigger pulse to the SCR only to turn it on.


The other method of power control that would be applicable is using PWM and IGBTs. This gets a whole lot more complex as you rectify the mains into DC and use the IGBTs and PWM to regenerate the mains waveform. Using the pwm, you control the amplitude of the regenerated mains to vary the brightness. It would turn out to be more expensive doing it this way for little benefit. Also, SCRs are much more robust than IGBTs.

One method worth mentioning is using IGBTs with phase control. To eliminate the large choke, you can control the turn on time and turn the IGBT on slowly - the killer is the power disipation during this time. You can also turn the IGBT on at the beginning of the cycle and turn of at some point later in the cycle. This is fine for resistive loads only.