Flutterby™! : Power supply follies 2004-03-01 00:46:47.777943+00

Power supply follies

2004-03-01 00:46:47.777943+00 by Dan Lyke 5 comments

I'm finally building that DCC model railroad throttle I've been threatening to do, and I wanted to have a power supply that will deliver decent current without costing an arm and a leg, and I need 5v for Vcc and 12v for the signal, so I'm starting with a transformer, a bridge rectifier, and a monster capacitor, and working up from there.

I am unused to dealing with real current. It is not good to short things that have real current behind them. Luckily I've only made funny flickering lights and popping noises, and not fried anything yet. Yet.

Wouldn't a PC power supply be perfect for that? Or do you need a lot more juice than one of those would put out (model railroads are not something I know a lot about)

Yes and no. The power supply on my bench that I use for testing claims to give 10A at 12v, which would be more than the power supply I've built. However...

Switching power supplies like that work by partially filling up a capacitor through a resistor from a higher voltage, and measuring the voltage of the capacitor and pulse width modulating the higher voltage to figure out how much to fill up the capacitor.

This works great for relatively constant loads, you keep the capacitor filled up, you have little wasted energy, the circuit generates little heat, life is good.

The problem is tht motors are not relatively constant loads, in fact they're also generators, so they can induce large spikes back into your circuit. So cheap attempts at figuring how much to fill that capacitor without overfilling it can get really confused. And model railroads have two long parallel bars, which also contribute to capacitnceissues.

Aaaand, I'm not sure quite how this will play out with the H-bridge I'm using, but compounding the motor transients is the fact that I'm inducing a square wave AC on the rails to send the DCC signal.

All this is the long way of saying "I'm not sure, but I have my suspicions". In my motor controller at work, when I was using a smaller switching supply I saw some huge problems, and I'm wondering now if similar things in the larger supply caused me some occasional glitches that I could never track down.

yeah, switchers really prefer nice constant loads. A lot of the work that goes into the controller circuit is to prevent certain pathological feedback cases from causing the whole thing to blow up.

Digikey has a couple of pages of linear supplies, however getting 10A is going to be expensive (12v 10.2A is about $115), and you really don't need your 12v regulated anyway.

On the other hand, they have *hundreds* of pages of capacitors, and a honkin big cap will be a lot cheaper than a moderately-large 12v linear supply. With a 1F capacitor you can make pretty much anything look like a constant load..

It's just like decoupling capacitors: you want to put them as close to the load as possible (you're effectively shuttling energy back and forth between the inductor [the motor coils] and the capacitor). You may also want to consider a big diode of some sort to keep the back-emf out of the switcher. All the usual issues of driving a motor or relay, you're just protecting the PS controller instead of the drive transistor.

Because the signal/load voltage has a huge allowable range, I'm just using a transformer, a honkin bridge diode, and will put a load on this puppy and keep upping the capacitor values 'til the ripples go away. The 5v supply needs to drive a couple of LEDs and the microcontroller, so an LM7805 is overkill for that, and right now I'm just using an LMD18200T H-bridge (with current and temperture sensing, hardly worth building tht one myself) for the track driver.