(Reposting an excerpt from a forum post I've just made.. figured I should keep it here as well.)
I've been planning for awhile to feed the 5V rails from a 7805.. good, reliable, solid technology and cheap as dirt; a lot of the newer LDO regulators are mosfet based and seem to die if you look at them funny.. ie: the beautiful little TI TPS7133QP PDIP for instance is a pretty ncie little chip.. but it maxes out at 10V; you pass more than 10V and poof, it goes hot and shorts internally. A 7805 takes anything about 8V or more and gives you 5.. the more you put in, the hotter it gets until melt down, but it delivers 5 till then; it can take a beating.. 20V or something, and with a big heatsink good to go. (Yes a switching regulator doesn't make heat, but I'm a touch worried they ripple a bit.. add some caps, probably fine; maybe a coil, a friend suggests.. ) .. anyway, 7805 is tough, I like it, and I have a pile of them.
(I should concern myself with component size .. go surface mount and small size items, but I'm still trying to stick with DIP and SOIC where possible etc, keep things easy to work with; next project I'll shrink more, but I want zikzak nice and easy to work with, easily hacked and modded and repaired. Even still, with STM32F on there we've got a fair number of SMT parts.)
My PSU seems to send 10.1V for a brief instant on powerup (or more, depending on its settings), and that kills the TPS7133 .. damnit! So I've been looking at options to handle that; I'm sort of assuming a 9V power brick is a common zikzak power source (though I hope to offer banana hacks and headers as well for ease of working on the bench), and assuming it'll have a similar risk of spike on startup from the brick; so I've been thinking of tossing in a 9V zener diode (say) up front to trim off some of that potential excess voltage.. or other tricks, as I kept thinking I was feeding that raw power into the 7805 to get 5V, and also into 7133QP or L78L33 to get the 3.3V out. (Problem with L78L33 is 100mA limit current .. and the STM32F4 eats up to 93mA at full consumption; I'm not driving it full, and not using many GPIOs, but still.. don't want to brownout at the wrong times, either.) So L78L33 probably out, despite my getting a 20 pack of them coming from china in a few weeks for a buck.. and other variations of the 7833 seem very hard to find. I suppose everyone has switched to 1.8V or switching regulators or whatever, though 5V is so common 7805's are still around and cheap (and big…).
If you are curious, I have very little to worry about for level shifting .. the STM32's I'm using can handle 5V inputs on their GPIO pins, even when running at 3.3V; no voltage level adjustment needed for the most part. I'm hoping not to include any trickery like that, such as voltage level buffering, or using a leveler.. at all. The RAM and such needs 5V, pretty much everything on the board is 5V, excepting the sTM32s, which are 5V friendly.
So back to TPS7133 …. worried about popping more of them (at $3 ea!) but had a _duh_ moment today; TPS7133 can take 2-10V and deliver 3.3out; for some reason I didn't see I could feed it low volts .. (I wonder if it upscales from 2 to 3.3? one would think from the basic datasheet..) – anyway, so I have a solid 5V from the 7805, so I can just drive the TPS7133 from the 7805, and voila .. stable 5V rail, and fork off the side to get the 3.3V rail. Seems to run great on breadboard, with the 7805 and caps taking that initial spike from the bench power supply no problem. No more worries about the spikey from the power supply killing the TPS7133 and destroying the whole board.. the 7805 and caps will protect the entire thing. It might be wise to run zener 3.3V and zener 5.0V after their regulators just in case of regulator meltdowns, but that just seems unnecessary.
I've got DC jacks in the pipe from china (and a pile of other little bits, see the blog post a week back); right off though, power section is DC jack, couple LEDs, couple caps and resistors, and the TPS7133 with its two caps. Thats not a lot of stuff, but with through hole components, thats a good square inch or square 1.5“ of space used up, and not even gotten anywhere. I've not worked out it, but I keep fealing like zikzak is going ot be a 6” square board to keep things airy. Elecrow 4“ square boards are $15ea or $25 for 5 of them. 6”x6“ works out like $70 for 5 of them.. big jump there. So I shall endeaver to maek 4”x4“ board, or maybe 4”x6“ or something.
Zikzak is currently planned to actually be two boards .. the base board which is RAM, cart slot for ROM, IO chip (avr probably) and the STM32F4 (or F2) for GPU, plus all the supporting stuff (on/off toggle for power, ps/2 keyb jack, etc etc) and a big pile of headers all over it, including the cpu-board header ('shield' style.) Then the cpu board which will be one of those credit card sized pcbs, including its own decoupling caps and any extra pins it may need. (ie: anything not used by zikzak can have its own headers on the cpu board, so you can jumper off to use them.)
cpu board figuring to be avr8 for simple testing of the zikzak main board, and it'll be able to run gamesd or whatyever of course; and eZ80 cpu board for awesomeness; and stm32 cpu board for supwer power.. dual 186MHz ARM chips would be monsters ..
Just waiting for all those jacks and bits to show up so I can measure them and make the CAD bits.