If we're using something other than direct current (DC) generators for wind or hydro power, or if we can't use variable frequency (unlike the mains) AC directly to advantage, we have to get involved with diode rectifiers sooner or later. As soon as we start talking about batteries, inverters and such, we need DC.
Unfortunately, diodes, like many things in the real world, are less than perfect in doing their job. They have power losses that generate heat - lots of heat. This heat has to be dissipated to keep the device from destroying itself since they have a maximum internal temperature rating, usually about 125-175°C, before meltdown. More on this later. Common silicon diodes exhibit a forward voltage drop (Vf) that is dependent on the forward current (If) and temperature. We often see numbers of 0.5-0.7 volts per diode quoted for Vf while actually, it can get up to 1.2 volts (peak) for high current applications like we use. Being a little conservative (not enough) and using the commonly used 35 amp diode bridge (MB352) as an example, an If of 15 amps per diode times a Vf of 0.95 volts = 14.25 watts per diode which results in 28.5 watts (they conduct in pairs for half the time) of heat to get rid of for the package of four diodes. Not trivial. In a 400 watt 12 volt system, that's over 7% loss, just in the rectifiers. Don't count on 35 amps being available under all conditions. The data sheet says that if you can't keep the case temperature under 55°C, you have to reduce the current. At a case temperature of 115°C, you can run only half the current. Losses explain one of the reasons why low voltage (12 V) is not often recommended for high power availability systems since more current is required for a given power level. Higher current - higher losses. Vf also requires that we use more turns of wire on our coils to achieve a given DC output voltage, which results in higher coil resistance and additional power loss. Diodes are a mixed blessing.
Diodes come in several flavors, each with it's own advantages. It's clear that a lower Vf is desirable and fortunately there is a diode type that has about half the Vf of the common type. The Schottky diode is available in ratings suitable for our needs but as usual, there's no free lunch. They cost a bit more and are somewhat limited in their reverse voltage rating and have higher reverse current. They're suitable for 12 and 24 volt systems but for 48 volts and above you have fewer choices. They come as single units and also as dual units with their cathodes tied together internally. You should be able to run these duals in parallel, if desired, with excellent current sharing. Schottkys are available in a wide variety of package types to suit many mounting configurations. Unfortunately, I haven't yet found a nicely packaged four-diode bridge (or even better, a six-diode three-phase bridge) with suitable ratings so we have to use multiple units. I don't understand why they're not commonly available. So, with a diode type that can cut our rectifier losses in half, reduce the amount of heat sink required, and reduce our stator losses, why aren't they being used by everyone? They aren't that much more expensive so it's a mystery to me. I've been using them for quite some time and haven't found any serious disadvantages.
Can we do even better than Schottkys? Sure, it's called synchronous rectification (SR) and doesn't use diodes at all. Instead, high power FETs are used as switches and are synchronized to the AC waveform to turn on and off at the appropriate times to achieve rectification. The adventure into SR isn't for the faint-hearted. It's significantly more complex and requires a fair bit of electronics but oh boy, what a payoff. With current-day FETs the losses are reduced to next to nothing. Heat sink requirements are minimal - a few square inches of aluminum takes care of it. I'm still working on a prototype and it looks promising. I suspect that before long, when more commercial SR modules are available, that we'll all be using them.
Now, let's attack the thermal problem. The heat generated in a diode is right at the internal semiconductor junction, a quite small area. Our job is to transfer that quantity of heat to the ambient air to be dissipated without exceeding the maximum temperature rating of the junction. If we don't - meltdown! Be careful in assuming what this limit is. Various manufacturers of the same part number have different ratings. In checking this rating for a number of different MB352 35 amp bridge rectifiers, I found maximum junction temperature ratings ranging from 125 to 175°C - a 50°C spread. Check the data sheet! And while you're at it, look for the thermal resistance from the junction to the case. Of the half dozen or so checked, I could only find it on two and they were different. Thermal resistance is specified in units of °C/Watt and for the MB352 is 2.2 to 2.5 °C/W. What this means is that for each watt of heat we move from one end (the junction) of this thermal resistance path to the other end (outside case surface), we will have a temperature differential of about 2.3°C at the two ends. You do remember the difference between heat and temperature, don't you?
To refresh your high school physics, there are three ways to move a quantity of heat from one place to another - by electromagnetic radiation, by conduction through a solid or by convection via a moving fluid (air is also a fluid). We will mainly be concerned with the last two. The device internal path from the junction to the outside case is through solids so the method is conduction. To continue the process, we mount the case on a heat sink via an interface that's also solid - more conduction. Then there's conduction through the heat sink itself and finally, the heat sink surface area is exposed to the ambient air so the final transfer is convective although, there is also radiation at this point, particularly if you paint it flat black. In our imperfect real world, each time we make a transition we incur additional thermal loss represented by more thermal resistances. We end up with three major ones to consider - the already mentioned junction to case, the case to heat sink and the heat sink to air. These are effectively all in series and determine the final result.
The formula for junction temperature (Tj) is: Tj = Pd(Rj-c + Rc-s + Rs-a) + Ta, where Pd is the heat to be transferred in watts, Rx-y are the three thermal resistances and Ta is the ambient air temperature in °C. For an example of Pd = 25 watts, Rj-c = 2.3, Rc-s = 0.3, Rs-a = 2.0 and Ta = 25°C: Tj = 140°C, uncomfortably close to the limit. If we had a hot day, it could push it over. But wait, we have another consideration. The data sheet says that if we have a case temperature of greater than 55°C, we have to derate the current. Using this example, the case temperature works out to 82°C (140 - (25 x 2.3)) which means we have to limit our output current to about 28 amps. Personally, I would never run a MB352 under these conditions. So, when is a 35 amp bridge NOT a 35 amp bridge? The answer is often, if we don't do it right.
If we want reliability, being conservative is the answer. Don't push components (particularly semiconductors) to their max ratings. Use more heat sink than you think you need. My personal rules-of-thumb for power devices are to limit expected operation to 50% of the ratings. I haven't replaced very many over the years.
Of course, this is far from an exhaustive treatment of the topic. I'm sure some of you are asking things like "how much heat sink should I use?", etc. ON Semiconductor has an excellent manual available online that really covers it all from the basic physics on up. The Rectifier Applications Handbook, HB214/D at 272 pages (1.9mb) is in .pdf.
http://onsemi.com
Stay cool!
