The general idea is to discuss the design of a circuit board that will use a 12v power source and provide several power options to run peripherals such as ITX mother boards, cameras, SSC-32, DC motors, etc… all from a single board the same size as the SSC-32 or the Mini-ABB so that it can be stackable. I can lay out the circuit board but we need someone interested in designing the circuit schematic.
I have (very) little experience in designing power supplies, but I might be wiling to give it a shot… What are the specs you were having in mind?
In terms of, how much different supplies, what currents, which voltage ranges, etc etc.
I ordered some step-down converter samples rated at 5V-5A and some other at 10A, yet they are 5v and are simply to have a simple prototype.
It should however be possible to make a step-down converter using a PIC or some other controller, i’d need to compare prices to see what’s more feasible when comparing cost to “useability-factor”
just as a casual whack at a design approach, maybe start with something like an NSC LM3489 which has an eval board (for about $45 at DK) you can use as a guide for layout with some beefier components. It seems like beefing an external switch high-frequency regulator like this up to 5-6 amperes range should not be real tough. It also seems like you could put two to four of them on a ssc-32 sized board and diode them together in pairs to generate VS1 and VS2, then slap a small low drop linear regulator to generate a nice clean VL and you would be off and running. the high frequency lets you use small low profile inductors and there are lots of P-FETs around these days in the low tens of miliohms RDSon with a 4.5V gate threshold.
Hello ScuD,
We need input from other members to see what kind of power requirments thay would like to have in this design. For starters we will need 6v and 12v but current specs is open for discussion.
EddieB, I’ve ordered some LM2677’s, they handle 5A without external FET’s, just for tries though… Maxim will kindly provide me with the part that could spawn 10A but i’ll need some more fets to get that part to work.
High frequencies do indeed mean i could use low profile inductors, but i’d try to go SMT either way.
Anyway I’ve been cracking my head over this all damn day (not much work done, thanks guys! 
) and since I won’t be getting any parts till somewhere next week I think i’ll crack out the protoboard tonight and start coding a PIC.
I am not following the need for a PIC… where are you going with that?
There’s no need for a pic at all, actually.
It’s just that a microcontroller can perfectly well “emulate” a switching power supply, and it’d give me something to do with my time while waiting for samples.
The only benefit it might give is a programmable voltage selection, eg. using it on the same bus as the SSC32 but with different identifier (it does support daisy-chaining, doesn’t it?) so you can select different voltages on the fly.
Offcourse there’s no real need for that in a robot… unless i’m overlooking something.
I think the design should definetly use SMT components. One thing to keep in mind though is costs. You don’t want to have to much on it that makes it to expensive. A pic could be used I guess to monitor the various outputs, or to check battery voltage, or to shut down if there is a problem.
heh. 8)
I would be inclined to do this with jumpers at the time you setup the hardware. A software mis-hap where you tell the regulator to put 9V (with 5A behind it) out on a power rail to a bunch of HS-5645 servos might get expensive fast.
It certainly could be fun to experiment with, however I think you may see a limitation in its ability to respond to transients, and since you are relying on an a/d conversion to regulate the PWM frequency there could be both some quantitization (sp?) and sampling theory issues needing to be addressed. A simpler approach may be to use a switching controller where the error comparator reference is exposed and drive that with a current or voltage from the PIC to set the output voltage level.
Either way we I would think we should definately want to be able to cover multiple output voltage ranges capable of working with standard hobby servos and higher voltage systems like the bioloid servos. being able to easily integrate the two systems could open up a lot of electro-mechanical design flexibility, especially if you could use a BAP to send commands to either an SSC-32 or the other servo systems directly.
Agreed.
Using the ADC’s would impose lots of issues with the high freq. drive of the output coil, which is why I was thinking of using the comparators, but then i’d need a DAC for the reference and i don’t know by heart if the '628 has that on board.
Jumpered power would remove the need for a DAC though, but anyways why use a controller when there’s tons of devices out there specifically designed for such a purpose.
The LM2677 (or something, don’t remember which) is also available in a version with a reference output so you can set the voltage yourself, however it’s an analog output, meaning it needs an external resistor to ground.
Possible to use another controller with a resistor ladder network, but yet again too complicated, a simple ladder network with dip switches or jumpers would work fine just as well.
5vdc @ 1 to 2 amps would handle the logic for most bots of this size. More is probably better.
Dual 6vdc @ 5 amps would handle most Hexapod, Biped, or pretty much any average robotic thing for servo current.
These are minimum currents. If the supply could generate dual 6vdc @ 15amps for VS1 and VS2 and 5vdc @ 5amps for VL it would probably be the most you could design for on a 3" x 2.3" board.
I have nothing but the utmost respect for anyone who can design a switching power supply from scratch. I used to work for a cable company in my youth repairing to component level every module used in the system. At the time we used to describe the way these switchers worked as “magic”. 
I suspect it’s a bit easier nowadays. At least I hope it is. lol
I would say the magic is different. You can buy off the shelf switching controllers with integrated switches and all you supply is the inductor and some low-esr caps. however the new magic is cramming this stuff in the smallest spaces at the highest efficiency possible. I could give you examples but then they would hunt us all down and kill us.
(just kidding, I think, heh heh heh)
No kidding… try debugging circuits where the biggest discrete parts are 0402 all day long 
No stopping technology i guess 
My point was not so much that the parts are smaller, although you are quite right about that, but rather that when you start cramming things together tightly and operating them at high frequencies and high power levels there are parasitics that become relevant. You need to consider them in simulation and your design verification process has to be comprehensive enough to consider a lot more detail than say even 10 years ago.
Indeed, I see problems like small blobs of solder flux ruining an entire HF path every day.
That’s probably why I’m a debugger, and not an RF engineer
Edit: Here’s a simple quick whip-up of how a dip-switch regulated 5A power supply could be managed using the LM2677.
Haven’t done any calculations nor have i checked what components are needed (eg. size of inductor etc), it’s just a lil check on what’s possible with this chip…
Looks straight forward enough. Is there something in the app notes about sensitivity to input impedance or are you just storing all those caps there? 
One thing you might change is to put a resistor across the DIP switch that can not be switched out to prevent an open feedback condition.
Good thinking about the extra resistor, overlooked it.
Nothing about the in- or output impedance on the component, but the caps are needed for the high inrush-currents you get with these switchers, not necessarily so much but then it’d be easier to use SMT caps, plus it increases the current that can be pulled from the capacitor ‘bank’.
But as I said, nothing’s been calculated yet, so it might work just fine with only one cap.
Great work ScuD! It’s a great starting point. It looks like there is going to be plenty of room for parts on this board.
This is only 5A Mike… you need 4 of them at least. 