Page 3 of 3

Posted: Wed Jul 02, 2008 4:58 am
by MrAl
Craig wrote:
The best choice is a switching regulator, built from a Simple Switcher
IC chip available from National Semiconductor, or another type
of switcher chip.
This sound like a good idea, however, I have looked and looked for a 20-24 volt, 3 amp+ voltage regulator but have been unable to find one. Do you know the part number for one, or point me in the right direction?
Hi again Craig,

Just to recap, i was recommending a switcher to reduce power dissipation
in the heat sink and of course to waste less power.

Since the current is relatively low (3amps) there are switching regulator
IC's available that are pretty easy to use. National Semiconductor makes
several IC's called "Simple Switchers" that are basically an IC plus
a few other parts. They even have a web page that allows you to design
the circuit right in the web page (they call it Web Bench) so you can do
a design without even having to calculate anything (much).
All you have to know are things like Min supply voltage, Max supply
voltage, load current, and output voltage.
The LM2670 handles up to 3 amps, and the LM2677 goes up to 5 amps.
Note that these devices do not require any external pass elements such
as transistors (MOSFETs, Bipolars). They do require an inductor of
suitable size, and at least two capacitors of suitable size and a couple
resistors (a pot for variable output) and a high current Schottky diode,
all which are available on the web for prices that are not too bad.
Just to note, these are not power supply 'controllers', they are complete
regulators similar to LM317 but they are switchers instead. The other
type of IC's out there would be classified as 'controllers' only, in that
they need external pass transistors (at least one) and these days
you really only need that if you have to handle large currents like
20 amps or 50 amps where you may not be able to find a chip that
does it with all the active elements on the chip itself.

Here is a comparison of the circuits for linear and switchers:


Posted: Wed Jul 02, 2008 11:55 am
by Craig
Thanks for the info. I will look into these options, and probably order some samples of these IC's to check them out. I tried my local supplier and they don't have any of them, other than the LM350.

Would I need a heatsink for the LM350?

Also, the bridge rectifier that I bought has a metel top to it, do I need to be careful where that touches inside? I was thinking of screwing it to the chassis some how.



Posted: Wed Jul 02, 2008 12:49 pm
by MrAl
Hi again,

For your app the LM350 will need a heatsink yes.

I found this heat sink


at Digikey for about 4 dollars.

It looks big enough for the train motor if it draws about 1 amp.
Let me know if this is not correct however. This assumes the
TO220 case and you dont mind drilling a hole for mounting.
They make other heatsinks with holes already drilled and
just for TO220 packages too if you like that better.

You can use a smaller heat sink if you want to use a fan with it.

It looks like your bridge rect can be mounted to the chassis because the
data sheet says it's isolated up to 2500v, but just to make sure do an
ohm meter check from each lead to the metal case.

If you need to run two train motors you should use two IC's and two

Good luck with it, and please let us know how it works out.

Posted: Mon Jul 07, 2008 4:00 am
by Craig

For the time being, I decided to build the power supply using two LM7824 voltage regulators. For the two 24 volt outputs, one is on all the time when the power supply is turned on, but the other one has a switch on it so it can be turned off when not in use.

We tried the unit out on the railroads and everything seemed to be working well. The two green LEDs are pretty useless outside, we weren't able to tell if they were on or off, and they stay on for about a minute after the unit is turned off. I assume this is because of the capacitors being used? Is there a way to have the LEDs turn of instantly?

One weird thing that does happen is it takes about 5 seconds after power is applied to the train before it starts moving. Is this also because of the capacitors? Do they need to fill up before they will release any power?

I have applied for some samples of those simple switcher IC's you mentioned (LM25576MH, LM2676T-ADJ, LM2670T-ADJ, and LM2677T-ADJ). As you can see in the pictures, I put all of the voltage regulation components on a single board. I figured this would be the easiest way to do it, so when/if I get those samples, I should be able to simply replace that board with the new one and be on my way.





Posted: Mon Jul 07, 2008 5:18 am
by MrAl
Hi Craig,

Usually a cap doesnt take that long (5 seconds) to charge up but then
again i dont know what value you are using for these caps. Is there
any series resistance present too? Where exactly is the LED connected?

It is very common for the caps to stay charged however, for a relatively
long time after power is removed if there is no load on the supply.
The way to beat this is to connect some resistor (500 ohms perhaps)
across the output of the supply so it 'bleeds' off the cap when power
is turned off. If the caps are big this wont work either though, so
you may have to use an active bleeder, which could be as simple
as a dpdt power switch where one side shunts the output to ground
through a low value power resistor (like 25 ohms) when the power
is turned off. If a dpdt switch is too inconvenient, a transistor that
turns on when there is power in the caps but no power coming in
through the rectifiers is another way to shunt the power (through
a 25 ohm resistor of course).
If it's just the LED you are worried about and not really the output
power itself, you can always connect the LED to a second rectifier
to the AC output of the transformer (instead of the dc side) and
this will mean it will turn off as soon as the AC power goes away,
even if the full dc voltage is still present.

Posted: Mon Jul 07, 2008 5:22 am
by sghioto
The discharge time on the caps is why the LEDs stay on so long. You could install a "bleeder" resistor across the caps to speed up the discharge, say maybe a 1K ohm 2 watt but that still leaves about 10 seconds and waste power. Better idea is to use a DPDT switch and disconnect the DC output from the regulators when off.
I think the train motor on start up is exceeding the 7824 current limit and the chip is shutting down enough to protect itself until the train gets moving. Going to need the 3 amp regulators.

Steve G

Posted: Mon Jul 07, 2008 5:57 am
by Craig
Hi Al,

Sorry for not posting the details. Here is the schematic of what I built. I have ommited the AC-DC conversion for clarity.


Posted: Mon Jul 07, 2008 8:31 am
by MrAl
Hi Craig,

Are you still seeing the 5 second turn on delay?
I dont see any logical reason for this from the schematic.
I have used 78xx devices many times in the past and never
saw a delay by one of them.
Maybe a lot of leakage inductance in the transformer,
that's my only guess so far.


You could measure the dc voltage on C5 as you turn on the power (after
a full discharge) and see how long this voltage takes to rise to
the full level (check that first). This might help explain a little.
Those caps are 4700uf each right?

What happens when you power up with J1 open, wait 10 seconds,
then close J1...does the LED2 light right away or is there a
5 second delay like this too?

Steve, good idea. Craig, are you trying to start up with the
train running? The motor stall current is much greater than
the normal run current...can be 10 times higher.
Do you have any 3 or 5 amp regulators to try out?

If you answer these questions we might be able to help a little.

Posted: Mon Jul 07, 2008 11:57 am
by Craig

I will try measuring C5 as you suggest. I will also try powering the train with my 3A test bench power supply to see if the delay is still there. It might turn on that it doesn't have anything to do with the power supply at all.

This is the order in which we start up and notice the delay (note that the power supply and the speed controller are two seperate items).

1. Set the speed control to the lowest setting
2. Turn on the power supply (the fan on the speed control comes on, LED light on the power supply comes on)
3. Turn the dial on the speed control to full speed
4. Wait 3-5 seconds
5. Train starts moving.

If I close J1 after the power supply is up and running, the LED comes on instantly.

I'm not starting the power supply with the train on full, however, cranking the dial from nothing to full (I assume) would have the same affect of drawing a larger number of amps.

I don't have any other regulaters to try, only that other power supply. I have ordered some of those simple switcher IC's you mentioned, as well as the coils, and will try that out when they come in. I just noticed on the weekend that there was an article about them in the June issue of N&V. How convenient!