OK - I'll take a little time here. Listen carefully. I'll try and be useful technically and otherwise. Good (hopefully) technical comment below, but 1st a very little general advice. People here are generally helpful, but not always. They don't always get the point and some are ruder than others. You happen to have been exchanging words with one of the ruder ones - but he is also extremely competent technically and well experienced in power supplies. You are far better off having him as a friend than an enemy. You have reacted to his rudeness with rudeness and in the process of accusing him of not listening and of offering stupid advice you have failed to listen fully and have said technically stupid things about his (quite possibly useful) advice because you didn't read what he said carefully. Exchanging sarcasm with Olin is seldom profitable for any concerned. The best you get is nothing and more often than not you get two or more disgruntled people. There are indeed a few one time posters who never come back but most newcomers do in fact stay. Some people here have trouble reacting well to newcomers but we try to stop them abusing them while they get used to the friendly family atmosphere :-). In fact, a premium is placed on politeness and treating each other well here. Levels of abuse which are an accepted part of many lists are frowned on here and active steps are taken to "discourage" such. (Try not to find this out actively ;-) ). I suggest we all start again and try and be nice to each other. Your original proposal had merit. It is not especially new and has its good and bad points, but it had the makings of a good idea. I suggest that anyone who has been criticising what they THINK Bob has been suggesting have a look at his original proposal. It was competently described and commented on and it seems technically sound. He didn't describe the end application but he didn't need to. While it's always useful to know what people are trying to achieve end to end, in this case he said he was keen in implementing a particular system. It's entirely valid to suggest alternatives BUT trashing the original concept without being aware of what it was is not an especially good idea. [[SUMMARY: PWMing mains AC with an AC switch in series with a transformer primary presents variable voltage low frequency AC input to the transformer. Advantages are no rectification loss, prospect of OK overall efficiencies, excellent power factor, reasonable prospect of open loop control, retrofitting to existing mains transformers, .... . Still needs low frequency mains transformer. ]] If I was trying to meet you spec from scratch I would try either something like what you have suggested OR a transformer and rectifier (or a SMPS) providing about 2.5 volts DC followed by a buck regulator. The output could be very precisely controlled by feedback. Let's have a quick look at what Olin suggested. He proposed a full wave rectified unsmoothed DC supply followed by PWM into a transformer. The act of PWMing will provide the AC component that the transformer seeks. The transformer will be able to be a very small one compared to one used on 50 Hz due to the high frequency of the PWM. When turned off the transformer primary will "flyback" and will need a recirculating diode across it. Lo and behold - you now have the SMPS you were trying to avoid :-). Efficiency should be able to be reasonably high (= lowish heat) as rectification and switching is done at high voltage and low current. What he is suggesting is similar in end result to yours but has the very significant advantage of a small high frequency transformer and it also allows a simpler switch. Let's try and address you original circuit. You can chop the AC as proposed with 2 FETS *BUT* they cannot be in parallel antiphase due to the body diodes - they must be in series - gate to gate, source to source, input and outputs from the 2 drains. Switching is achieved by PWMing the gate to source as required. The whole switch floats across the whole AC range so the gate drive needs to be floating. There are good ways of doing this BUT for experimenting it's better to use an "EASY" way first and come back and improve it technically once the rest works OK. An easy way is to use a 9v volt battery plus an opto to drive the FET gates. The battery will last for weeks of use and a floating isolated supply (or other solution) can be implemented to replace it in due course. If you have a conventional mains transformer AND you PWM the mains supply AND you filter the output to remove the PWM component BUT not the 50 Hz AC component THEN you will get a variable amplitude AC waveform which the transformer will happily accept. This could be a very nice solution but still has a standard mains transformer. Switching efficiency is goodish, and there are no rectifiers to dissipate power. The PWM filter need not be excessively large. The transformer itself will form a reasonably good filter in its own right due to its significant inductance. Let's jump to the PIC question and then look at other things. If you use 10 bit control then you have 2^10= 1024 steps. If you want 48 kHz then you get about 50 million possible steps/second which is VERY fast PWM. It's also humungously fast for FET or other switches. I would be surprised if you need either 48 KHz frame rate or 10 bit resolution BUT in the absence of knowledge of the final application I may, of course, be wrong. I have a design where inaudibility is important and I am PWMing hundreds of watts in an electromechanical device that attempts to be a speaker. I find 20 to 25 kHz is high enough for most people. If you are working with something that heats as its main task and it is in air and if it is working into a heat sink (person, wood, liquid, other) and if it is moving and if it is fed via leads that constitute a significant percentage of the load THEN variations in convection/conduction/radiation, air currents, thermal paths, working medium and the rest all work to ensure that far far less than 10 bits accuracy in the outcomes is possible. Real world engineering often doesn't need such. As we don't know the application .... . I am not familiar with all of the latest PIC range but some modern processors (eg AVR ATtiny26 have a phase lock loop controlled PWM where the PWM frequency is locked at many times the processor clock frequency. Without looking at the data sheet AFAIR the PWM can be at up to about 50 MHz data rate ! Switching a power FET at such rates is another matter. I would not recommend much above 1 Mhz fastest switching time unless you are desperately unable to do it any other way. Feedback control will often make raw control precision less important as you can alter the drive to suit what the output is doing. If you want to try the basic idea you can use far far far lower than 10 bits accuracy to trial the basic concept. An analog 555 oscillator would do just fine to drive the series connected back to back FETs into the transformer primary. If I was doing this I'd start with low voltage :-) If you are able/willing to tell us the actual application is it would help. Sounds something like either a poker-work type heater or a person heater/therapy unit. Russell McMahon -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist