To both 2112 and RadekG
The idea is exactly the opposite - use simple diodes rectifier and transistor current limiter.
I am not a noob and know how the rectifiers work and in the calculations above 220 * 1.4 is simplified for 220 * sqrt(2).
Many people said that it won't work and that's why this (crazy / hobby) project - to prove it does!
I have a working design at 24V DC and the built-in current mirror in the chips is working and balancing (almost) perfectly the voltage. I have made extensive test with the chips previously and this project will be a miner around the 55nm Bitfury chips, not a standard miner design. The chip survives 1.5V with proper cooling and it's voltage drop depends mostly on the current passing through it.
So transistor current limiter should be more than enough and 1.5V per chip is 420V AC, which won't happen and at that voltage the current is over 3A, while i will limit it to 1.5A or 2A max. This equals to about 0.88V per chip, because it's internal resistance when hashing is 0.44 Ohms based on my tests.
There are two other problems concerning me - the SPI speed (my next post) and the ripple/capacitor size (the next next post)
Yeah, it will most certainly work, when designed correctly. I'm still perplexed why you were concerned about few percentage point of RMS voltage variation (220-235V or 205-265V) while your main concerns should be ripple and protection against spikes.
Ripple itself on a single-phase system would be around 40% (25%-50% if I remember correctly) unless you drastically over-design the ripple filter and your rectifier diodes conduct only during few percents of the mains cycle. I'm unclear what is the purpose of your design: (1) show that it is possible using ridiculous over-specification or (2) an experiment in a proper power electronics design that could be practically scaled up and deployed later.
In case (1) you just don't worry about inductive spikes in the mains voltage because you ridiculously over-specified capacitors.
In case (2) you have to have proper protection against spikes. I don't have specific information about Bitfury's chips, but in metal-insulator-semiconductor chips the over-current protection is typically not enough. The breakdown is not only thermal, it involves electromigration of metal over insulator and insulator under metal. Electromigration is both irreversible and cumulative. In normal industrial practice you will need parallel over-voltage shunt to extinguish the inductive spikes coming from the mains. Relying on over-current protection is only acceptable for bipolar-junction devices where the junction breakdown is reversible and non-cumulative.
In case (1) you just don't worry about power factor and accept that the diodes conduct only during top few percents of a cycle.
In case (2) you'll have to use 3-phase power and polyphase rectifier design to have acceptable power factor and diodes that conduct current around 60 degrees of the 360 degrees cycle.
Again, I don't have a specific knowledge about Bitfury's chips and their communication protocols, but some power electronics designs deal with 100% ripple by resetting the devices at 100Hz or 120Hz during the zero-crossings of the mains power. It probably wouldn't pertain to Bitfurries, but you still may consider modulating your mining clock rate during each half-cycle of the ripple. Over-clock when the chip are over-volted on the average and under-clock when the chips are under-volted on the average. The required additional power in the controller is next to nothing when compared with the power losses in the full, actively stabilizing regulators. Adjusting the clock rate several hundred times per second shouldn't be a problem for properly designed controller.
I'm sorry my first response took you for a noob. I simply didn't consider the case of (1), a design like one would do to settle a bet. In the past I had to deal with many people who were digital design experts but relatively ignorant in power electronics design. They would advocate type (2) designs with flaws obvious to anyone who had to experience beyond stabilized/regulated benchtop power supplies.