Need some advice on power management with 3.3v & 5v.

[cbmeeks]

So I've been building small boards lately with various designs (and failing miserably but at least I'm having fun!).

Anyway, I have a few designs I want to build that involve 3.3v micro-controllers. I try to use all 3.3v if I can. 65C02, 65C22, SRAM, MCU's, etc. Even glue logic if I can.

Some of my designs require interfacing with legacy (Apple IIe) equipment that run three different voltages.

In the case of the Apple IIe, it provides (IIRC) three voltages from the slots. +12v, +5v and -5v. I plan on using level shifters between the bus/control pins. But I still need two power sources. +5v parts are easy. I can just tap into the provided +5v line from the slot (along with some decoupling capacitors).

But I'm struggling with providing a clean 3.3v source. My initial thought was to use a UA78M33 linear voltage regulator. It can take an input of 25v or so and provide 3.3v. I'd like to tie the +5v to the input so that I produce less heat. I believe the drop-out voltage is 1.7v. So that would be 3.3 + 1.7 = 5.0v. Sounds perfect. Almost no heat generated. But that sounds too good to be true. Seems like I'm walking a tightrope and any drop of current might start causing issues.

I could use the +12v supply but that's going to be a lot of heat. I guess I could put a giant heat sink on it. But that sounds kludgy. I also thought about maybe using an LM317 to step it down twice. Step the +12v down to +7 or +8v using the LM317 and then using that as an input to the UA78M33. But that means more resistors, parts, math, etc.

The second part of my LONG question is dealing with ground. My apologies for my ignorance, I really should know this but I like to triple check when dealing with power, vintage equipment, etc.

Assuming I get my two voltages, have my micro-controllers hooked up, have legacy NMOS +5v parts hooked up, etc...do I literally tie all grounds together?? I know that sounds silly. But is "ground" just "ground"? Or, do I need to worry about having separate ground planes for different voltages, etc.

My 65C02 designs (and legacy 6502 equipment) will run 1-2MHz tops. My micro-controller (Propeller) could run 96MHz but using internal PLL's with an external 6MHz crystal....not sure if that's relevant but I wanted to mention it.

I'm graduating from simple breadboards to "real" devices. Lots to learn.

Thanks!!

[BigDumbDinosaur]

cbmeeks wrote:

...But I'm struggling with providing a clean 3.3v source.

Take a look at Microchip's MCP1700-3302E/TO regulator. Data sheet attached.

Quote:

The second part of my LONG question is dealing with ground...do I need to worry about having separate ground planes for different voltages, etc.

A single ground should be fine.

[BigEd]

cbmeeks wrote:

My initial thought was to use a UA78M33 linear voltage regulator. It can take an input of 25v or so and provide 3.3v. I'd like to tie the +5v to the input so that I produce less heat. I believe the drop-out voltage is 1.7v. So that would be 3.3 + 1.7 = 5.0v. Sounds perfect. Almost no heat generated. But that sounds too good to be true.

Indeed, I think there's something you've missed. If the regulator drops 1.7V, while providing some amount of current, then it will dissipate the corresponding amount of power, and get warm. The question is how warm is it allowed to get. You're right that asking it to drop less voltage between input and output is going to minimise the heat production. It might even be that with 5V in, less 10% margin, less 1.7V dropout, is still within spec for your purposes. But it might be too low, which is where a lower dropout part starts to look interesting. BDD's datasheet pages 13 and 14 take you though the appropriate calculation: basically if you find you need dissipate little heat, you choose a device in a small package, and if more then you need a bigger one and eventually a heatsink.

Here's what I don't know: if the dropout voltage is only 180mV, does it make sense to first drop some voltage with a suitably rated resistor, then feed the LDO, with of course nice big capacitors nearby both the input and the output? The resistor will drop part of the voltage and produce some of the heat, leaving the LDO to drop less, produce less, and so not run so hot. Anyone?

[GARTHWILSON]

How much current do you need at 3.3V? Approximate minimum and approximate maximum? The dropout voltage will vary with current and temperature, and it's where the regulator becomes unable to regulate. If it really is 1.7V, then going from a regulated 5V down to 3.3V will work, but only because the regulator basically becomes like a 1.7V zener diode.

I have gotten many calls from installers in hangars putting our equipment into aircraft and getting a lot of noise from alternators, strobes, etc. into the audio because they ignored what I said in the installation manuals and just thought ground is ground is ground; but this problem mostly has to do with analog. I split ground planes on our boards when necessary to keep switching power supply noise, and digital noise, out of the audio and A/D and D/A converters. When you do that, you need to run traces over the border only at the one point where the planes join and there's a power bypass capacitor there too. In your case though, it sounds like it's all digital, and you do not need separate grounds or ground planes for a 5V section versus a 3.3V section.

LDO regulators, like their name says, have a very low dropout voltage. They are more difficult to keep stable though. It's easy to lose control of the poles and zeros and wind up with an oscillator; so be careful to follow the manufacturer's guidelines for input and output capacitors. They're there for a good reason.

[Arlet]

BigEd wrote:

Here's what I don't know: if the dropout voltage is only 180mV, does it make sense to first drop some voltage with a suitably rated resistor, then feed the LDO, with of course nice big capacitors nearby both the input and the output? The resistor will drop part of the voltage and produce some of the heat, leaving the LDO to drop less, produce less, and so not run so hot. Anyone?

The total dissipation stays the same, but it will divided between the resistor and the LDO. I wouldn't recommend this though, because the voltage drop over the resistor is fixed for a certain current, leaving less headroom for the LDO the work with. I would just use the LDO, and if it gets too hot, pick a bigger one and/or mount it on a heatsink.

[Aslak3]

I've had good results from switching regulators, specifically the "all in one" 78xx replacements. I'm not sure if they are suitable for your 5V to 3.3V application, but from the datasheet it looks like it would work.

SR05S3V3:

http://www.farnell.com/datasheets/1811823.pdf

One downside is they are not cheap..

[BigEd]

Thanks Arlet. (Might be worth noting: a buck converter is a different approach to a linear regulator, with a more complex implementation using inductors, whereby the excess voltage is not all converted to heat. It's useful when regulating battery power when it's important to waste as little as possible, or when it's not desirable to produce so much heat, or desirable to pull less current from the higher voltage supply. It might be that Aslak3's suggested component is exactly this.)

[Arlet]

Yes, the components that Aslak3 pointed out are buck regulators with controller, inductor and capacitors all integrated into a small potted case. They are a bit more expensive, but very convenient, and they usually stay pretty cool to the touch even when supplying 1A or more.

For battery operated devices, they can actually pay back their investment by lowering battery consumption.

[cbmeeks]

Thanks for all of the great replies! I certainly have my homework.

I'm not sure of the total power requirements yet. Being a newbie, I've just plugged things in and not worried so much about what the draw was. So I know I have some calculations ahead of me.

The circuit isn't anything complex. It will (hopefully) be a Mockingboard clone for an Apple IIe.
But, instead of using the legacy AY-3-8910 chips (two of them), I plan on using one Propeller MCU (SMD) to emulate them both. I don't know what the power draw would be off the top of my head.

I will assume it will be running all COGS full speed when I calculate it.

The other chips in the circuit would be two 65C22's (SMD) and an op-amp to drive some speakers.