I don't completely agree with that article, I agree that people do tend to undersize things (I sometimes do on the bases that it not only saves money and I probably won't use it at full power very often) but there's over sizing things to the point where it becomes unnecessarily expensive and bulky. I would choose a 15V transformer minimum and no higher than 18V at this current.
A 60000uF capacitor, used at 20A, and discharging during almost a half cycle at 50Hz (10ms), will drop the voltage by almost 3.3V
LTSpice says 2.55V of ripple at 15Vrms with a 20A constant current source on the output; this is assuming the transformer is perfect which I know is not true.
We need to calculate the equivalent series resistance to model this correctly. Suppose the 15V 30A transformer regulation is 5%, the off load voltage will be 15.75V but if it's loaded with a resistive load that draws 30A then the voltage will drop to 15V, the equivalent series resistance will have 0.75V across it at 30A, therefore ohms says it will be 0.75/30 = 25m
hm:.
Accounting for the series resistance and the rectifier voltage drops, LTSpice says the ripple will be 2.2V, the peak will be 18V and the valley is just over 16.6V, if the regulator drop is really just 2V then we have 0.8V to spare. I do agree that if the mains voltage is on the low side it could still be a problem but this normally isn't the case in the UK. I don't think I've ever measured a voltage below 228V, it's normally 235V to 240V.
On the other hand, if you use a regulator with a lower drop, and/or a larger filter capacitor, then you can slightly ease the transformer voltage requirements. This can be very useful to keep the filter capacitor voltage rating requirement at 25V, since otherwise you would be forced to use a 35V capacitor, which is much larger and more expensive. A lower transformer voltage is also an advantage from the efficiency point of view. After all, the complete excess voltage has to be burned off by the regulator, causing a huge power loss and requiring a large heat sink!
This is precisely my argument, and is also why I would use N-channel MOSFETs rather than bjts. Also a 120,000:mu:F capacitor 25V is about the same price as a 35V 60,000:mu:F capacitor and you save money on the heat sink, transformer and rectifier.
Another issue is what kind of pass elements to use for the regulator. MOSFETs are not a good choice, because they are much more expensive than bipolar transistors for a given minimum voltage drop and power dissipation. So, almost every power supply uses bipolar transistors. NPN transistors are usually preferred over PNP ones, because they are cheaper for a given performance, and there is wider selection. So far, so good. But most designers place their pass transistors on the positive side, in emitter-follower fashion, adding a Darlington driver (or two Darlington stages). This is a very bad choice for several reasons: One, each transistor connected in that way produces a minimum voltage drop of 0.6 to 0.7V. A three-stage arrangement, as is often needed, would have a minimum drop of around 2V, plus the drop caused by any equalizing resistors! Also, the transistor collectors, which are connected to the cases, are at the unregulated positive voltage, and thus require insulation from the heat sink and power supply case. The necessary mica insulators add a huge amount of thermal resistance, making it much harder to cool the transistors properly.
I don't know when this was written but today MOSFETs and bjts are around the same price; not only that but MOSFETs have the following advantages.
- Lower dropout voltage.
- Better transient response.
- They don't need any drive current.
- You safely parallel MOSFETs without series resistors thus saving additional parts and voltage losses.
The only real disadvantage is they can be destroyed but high gate voltages but that's pretty easy to fix by adding a zener across the gate and source.
I do agree that it's better to use negative devices, PNP bjts and especially P-channel MOSFETs are more expensive for the same rating then their N-channel/NPN counterparts.
https://ludens.cl/Electron/Ps20/ps20.gif
What's going on with the :mu:A741? I would use a real op-amp like the TL081 that will give half decent transient response and would consider using a proper little regulator IC like the LM723.
I have built MOSFET regulators before with dropout voltages well below 100mV. A good quality MOSFET design should give good performance at this current even with a 14V transformer and a low mains input voltage.