On Tuesday, November 12, 2013 6:00:06 PM UTC-6, steve...@gmail.com wrote:
Perhaps there is a solution to increase output current with short circuit protection with LM7805. Use BD534, PNP Epitaxial Silicon Transistor to handle current ... I do not know if it may need heat sink for BD534 transistor. ... I presume when it run at less than 3A should not product too much heat or possible a small heat sink. What do you think about this method?
Linear regulators turn (Vin-Vout)*amps into watts of heat, no matter how you build them, so the heatsink would still dump 9 watts regardless of tech. Also a heatsink of given thermal resistance (degrees C per watt) will always run the same delta T for a given # of watts no matter the tech so that won't help either.
So yeah its dropping 9 watts in a TO-220 and a bare TO-220 is usually like "fifty" degrees per watt hanging in the breeze, so ten or so watts at fifty or so degrees per watt would theoretically imply like five hundred degrees or so without a heatsink? Of course just bolting to a PCB without any metal heatsink would help.
One problem with the BD534 is Mouser claims its end of life. Also Mouser claims its $1.09, which is more than a second (almost 3rd) 1 amp 7805 regulator would cost. So you can get 3 one amp regs for the cost of 1 eight amp pass transistor (and you still need all the other parts), but you only need 3 amps not 8, so ...
I couldn't get the datasheet from datasheetcatalog to work but I got the sheet from Mouser and it lists a max junction temp of 150C which is a good 25C higher than the regulators we've been talking about, which is awesome because if its merely as good as the LM323T thermal resistance that means it can drop an additional 8 or so watts, which at a 3 volt drop means an additional 2-something amps more than a LM323T before it overheats at its higher temp. Of course 150C is smokin hot.
I did a search on PNP at Mouser and found a TIP42 from Fairchild, sixty cents a piece, not end of life, also TO-220 package, it too can run up to 150C junction, and it too does not specify thermal resistance from junction to case, but I looked at the watts / degree graph and did some estimation and if I got it right they seem to be claiming a thermal resistance of 0.5 C/W which is awesome compared to a linear regulator IC, which makes sense... the only thing on the die is a giant transistor not all that regulator "stuff" in the way. So where this does help is the regulator transistor junction at 9 watts and 0.5 degree/watt would run about 5 degrees hotter than its heatsink, and the junction in one of those won't fry until it hits 150C, so that means any heatsink temp at 9 watts below 145C is "OK". Well a 145C heatsink is insane, but at least the transistor wouldn't mind, anyway.
I didn't bother looking at the TIP42 beyond thermal, maybe the voltage ratings are hopeless. I believe it had a 6 amp Ic so thats good enough.
Really the problem with using many transistors and ICs to create a complicated large power linear supply is it becomes cheaper to install lots of smaller regulators. I'm not saying a TO-92 (looks like a transistor) next to each IC makes economic sense, but that's certainly the far opposite extreme of one giant supply. Could end up with something like two 7805 one running the top half chips and the other running the bottom half. Or a 7805 next to each "row" probably not needing much heatsink with modern chips.
There used to be a series of linear low dropout FET based regulators in TO-220 around 25 years ago, that aside from the grapevine reporting they were legendary oscillators, they could be slapped in parallel because they had built in current limiting so each one would supply its perfect 1 amp or whatever or until it overheated, so if you needed 5 amps you just slapped 5 in parallel, until they started oscillating, or so the stories went. That would be a handy chip to have around for this. LDO regs are a different kettle of fish which I'm not as familiar with. All I remember about the oscillator rumors was the current limiters interacted with each other.
TLDR - Cheaper and simpler to use three 7805, one for each row of chips...