"Universal" AC-DC Power Supply for O-gauge Locomotives and Rolling Stock

I've actually done the copper pour on several, and it doesn't have nearly enough mass to be all that useful.  It was a good idea, or at least I thought it was.   Another issue is if you heat up the board a lot, you shorten the life of the other components, specifically the electrolytic capacitors.

Rather than a separate board, a chunk of metal would be a lot more effective.  That can be added to the existing design, or you can even use a commercial TO-220 heatsink.  Of course, in a locomotive or rolling stock, the "chunk of metal" can be the chassis, big and metal!  That was why I oriented the regulator so it could be bolted to something.

I'm wondering why not a switcher?

Just a question, what about the LM2596 instead of the LM317? I think I read here somewhere that the 2596 generates less heat? Remember my knowledge is limited so I really don't know the differences between the regulators, just remember reading that here somewhere in one of the electrical threads. What would be the difference between the two regulators?

More parts, higher cost, higher ripple at low currents, and potential EMI issues.  The module will get bigger trying to use the LM2596.  The 330uh choke has to handle the full current, that makes it larger.  I thought about it, but for most of the stuff I want to do with the supply, I think I can do it with a much simpler and more robust linear regulator.  You typically need bulk capacitance on both ends as well.  This circuit doesn't include the input rectification or DCS choke, and it's already more parts than my design.  The 220uf on the input here is also too low to support a lot of current, that has to get larger as well.

gunrunnerjohn posted:

Note that for full-wave operation, you must insulate the regulator tab from the chassis.

 

Since the LM317T does not have a "ground" terminal per se, doesn't the half-wave mode also require insulation of the metal tab?

Is this to be offered assembled/tested, kit of parts, bare-board only, some combination, or to-be-determined? 

Is that a 1-turn trimmer?  As you know the LM2596 eBay modules have multi-turn trimmers (20 to 30 turns) where one full turn only changes the voltage by about 1 Volt.  With a single turn effecting a change of 10-20 Volts or more, just squinting at the trimmer and the voltage changes!  Well, it's not that bad but you know what I mean.  To that end, if it is indeed offered bare-board, I'd put in an offset resistor in the same way you did for the Hennings LED module.  It costs you nothing but provides flexibility for, say, driving a 12V LED strip where the adjustment range should be bounded by, say, 9-12V DC.  I don't know if there are economical multi-turns that have the identical SMD footprint as what you're using but a pad pattern that accepts different types of trimmers might be nice space-permitting of course.

I like the idea of the aux cap.  But it seems if used it would be in the 100's of 1000's of uF.  J4 looks to be 0.05" or 1mm lead-spacing?  Is the cap meant to be directly installed into J4?

Stan, you are correct.  I was basing my thinking on the fixed voltage regulators that I've been using, of course the LM317 will require insulation from the chassis at all times, good call.

The adjustment is a one-turn trimmer, the same one that's on the lighting module.  I've used it successfully on my lighting module, it works pretty well.  It's also very cheap.

One extra 2 cent resistor wouldn't break the bank, that's a reasonable option to extend the range and make the 1-turn pot less sensitive.

All the multi-turn trimmers I see are larger and/or considerably more expensive.  The addition to the cost of the module is significant, I don't think I'll have that much trouble getting pretty close to the voltage necessary with a single turn pot.  This Bourns Inc. 3214J-1-202E is about as cheap as I find, and they're still $2.50/ea for qty 25.  It's fairly compact if you pick the J-Lead package.  Trying to get pads to line up with multiple styles of pots will probably take a lot of research to find the parts.

As far as the aux cap, I really intended those connection points to be wires.  My thinking is a large cap sticking out of the board will be a lot harder to deal with than having it tethered and separately positioned.  I envision for most uses, it wouldn't be necessary.

gunrunnerjohn posted:

As far as the aux cap, I really intended those connection points to be wires.  My thinking is a large cap sticking out of the board will be a lot harder to deal with than having it tethered and separately positioned.  I envision for most uses, it wouldn't be necessary.

OK.  If you see the aux cap as a rare requirement, what if you placed a pair of thru-hole pads near the 150uF pads.  In other words, for the rare application that needs more than 300uF, it seems the on-board caps can be skipped and just install a thru-hole 470uF, 1000uF, whatever.

Here's another reason to make the J4 pads bigger/more accessible.  I get your point about not making a board that depends on the life-cycle of an eBay DC-DC switcher module.  So if you are not going to offer that "other" design that overlays a specific eBay switcher, one could use this design and NOT populate the LM317T circuitry and use J4 to feed an eBay DC-DC switcher.  As you know, most of the smaller DC-DC switchers have tiny (e.g., 10uF) input caps whereas the LM2596 designs typically have 100uF or 220uF.  So somewhere in all this is using this new board to make any eBay DC-DC module into an AC-DC module; that includes step-down, step-up, etc..

Hmm...  I like the idea of just putting the cap connection as pads to use instead of the standard caps, that's good thinking.  I also will put in the resistor.

I still have the boards for my two variations of the eBay switchers, so I'm set for those.  If you were stringing one of switchers from this board, the whole assembly gets a little cumbersome.  I know you're the king of the pile-o-eBay-parts solutions, but somehow it offends my sensibilities.

Here's what those changes would look like.

OK, I'll try a different tack.  Imagine the following.  What if the application requires 2 separate voltages.  For example, there was a recent thread about a 1.5V DC battery-eliminator horn power supply.  A step-down switcher is clearly the way to go for such a huge drop with relatively hefty current.  But say the application also needs another voltage for lighting or whatever.  So this board could provide the lower-current, higher-voltage output using the LM317...and if you made it convenient to access J4 you could run unregulated but filtered DC to a 99 cent eBay switcher.  In other words I'm still lobbying for 2 more holes (or some convenient way to run unregulated DC off-board).

IMO the thru-hole cap should be rotated 90 degrees to neatly fold in the footprint of the board.  Also, is that an 8mm aux cap?  IMO probably better to provision for a wider body (at least 10mm).  I realize if folded it might cover the trimmer, but I figure you set the voltage once, then fold the cap.

And now for something completely different.  It occurs to me that if I install a small value (10, 22) resistor in place of L1, this board can drive an isolated-rail relay with regulated 5VDC, 9VDC, 12VDC.  The resistor of course for inrush current limiting to demote arcing.  So I'm imagining you overlaying say 1206 pads over L1.  THEN, since it appears that the bottom of the board is empty, what if you placed pads for an SMT relay!  I realize the board targets rolling stock.  But if you look at what's out there for off-the-shelf isolated-rail relay modules, I'm not sure any have enough capacitance to perform suitable dirty-wheel/rail chatter prevention.  I only say this because pads cost you nothing...and you did ask for suggestions!

You have connections to the unregulated power now, those two phantom capacitor connections.

I didn't rotate the cap, though that occurred to me, simply because the connections interfered with the existing cap pads. 

Well, you could solder a suitable resistor to the L1 pads the way they are, a 22 ohm resistor could be any reasonable power level, it isn't going to drop much voltage.

As for a relay module, how do you envision this working?  Are you powering the whole board simply to pick the relay?  You'd certainly have chatter control with 300uf of capacitance and a relay with a 10-15ma coil!

Thanks for the explanation, I see the differences now. Now to go back and read the rest of you and Stan's discussion here. I enjoy reading these, even though I don't always understand them.

gunrunnerjohn posted:

You have connections to the unregulated power now, those two phantom capacitor connections.

...

As for a relay module, how do you envision this working?  Are you powering the whole board simply to pick the relay?  You'd certainly have chatter control with 300uf of capacitance and a relay with a 10-15ma coil!

But if one chooses to install the aux cap, that uses up the pads to tap into unregulated power.  So part of the "universal" idea is as a pre-regulator to the 99 cent eBay switcher with the 10uF input caps.  So when running on a Z-4000 set to 22V AC you can have a problem; as discussed in your other thread they can have max DC inputs of 24V - 28V depending on the switcher chip used.  If the 317 is set to 22V DC or so, it can safely limit voltage.  When AC voltage is lower, you get the 317 drop-out voltage.

As for relays, I'm mostly thinking of the isolated-rail triggered 2-board solution using a ~$1 eBay relay module.  Size is not an issue and ideally the board would have mounting holes but that burns board real-estate.  So yes, the "universal" board is used for its AC-to-DC, anti-chatter with inrush limiting, and voltage regulation.

As for the SMD relay idea, it's just a lark since/if pads are free.  For example, I see really cheap SMD relays for less than 50 cents:

This foot-print seems to be common for SMD relays; if the pads are placed on the back of the board with finger pads coming to the edge of the board, an ambitious DIYer can have a 1-board AC relay activated by track voltage whether in a piece of rolling stock (on a UCS track) or on a layout (isolated rail triggered). 

I realize slicing and dicing the "universal" concept eventually leaves nothing!  But I think you're still at the measure twice cut once stage. 

stan2004 posted:

But if one chooses to install the aux cap, that uses up the pads to tap into unregulated power.  So part of the "universal" idea is as a pre-regulator to the 99 cent eBay switcher with the 10uF input caps.  So when running on a Z-4000 set to 22V AC you can have a problem; as discussed in your other thread they can have max DC inputs of 24V - 28V depending on the switcher chip used.  If the 317 is set to 22V DC or so, it can safely limit voltage.  When AC voltage is lower, you get the 317 drop-out voltage.

I'm sure the enterprising folks can solder a wire to the cap connection on the back side of the board, we are talking about folks that know which end of the soldering iron gets hot, right?

I'm not understanding the pre-regulator, you're sending the supply current through the LM317 and then to the switcher?  That seems not all that efficient.  In order to get any meaningful current from the switcher, you'll be pushing the LM317 pretty hard.

Look at the size of those relays Stan, they would fill most of the back of the board.  The reason I use the ones from Digikey is they're small, and I know that in a year or two, I can still get them.

The outputs for an external relay seems more realistic and more universal.

stan2004 posted:

I realize slicing and dicing the "universal" concept eventually leaves nothing!  But I think you're still at the measure twice cut once stage. 

Yep, we're still talking about ideas, I'm not going to copper that quickly, I want to think about this one longer than I did for the ill fated 150ma regulator version!

Seems like adding all the additional voltage taps and caps is getting away from the intended simple small board

Matt, I'm not making it any bigger, and the basic function remains the same.  If I can inject some options without changing the size, that's a free bonus.  I just want to be selective about what options I inject.  I think if I laid it out for a relay, I'd probably use the TE Connectivity IM series relays, they're really small and come in various coil voltages and contact ratings of 2A and 5A.

I use these on a couple of different projects.

gunrunnerjohn posted:

I'm not understanding the pre-regulator, you're sending the supply current through the LM317 and then to the switcher?  That seems not all that efficient.  In order to get any meaningful current from the switcher, you'll be pushing the LM317 pretty hard.

Look at the size of those relays Stan, they would fill most of the back of the board.  The reason I use the ones from Digikey is they're small, and I know that in a year or two, I can still get them.

Using the LM317T as a preregulator to an eBay DC-DC switcher allows operation at full 22V AC as might be seen if someone bumps the handle on a Z4000.  We are talking about a "universal" AC-DC converter rather than one that has the Lionel warning about not operating above 19V AC (or whatever).  The LM2596 DC-DC modules can handle 22V AC but the smaller DC-DC modules cannot.  So even though the LM317 would lower overall converter efficiency, if dropping track voltage down to 1.5V, 3V, 4.5V, 5V DC or other common DC voltages, if drawing any reasonable amount of current (say, 100 mA or more) the loss/dissipation in the LM317 is more than compensated for by the efficiency of the switching step-down converter.  In other words, the LM317 is more used as a over-voltage limiter than as a regulator per se...15 cents (plus a couple 2 cent resistors) is a small price to pay for "universal" O-gauge performance...in my opinion that is.

As for the SMD pads for a common style of relay, that's a cost-free option as soldering to pins of a thru-hole relay can be annoying.  From what I can tell the board is about 1/2" x 1" or 12mm x 25mm (or so).  So yes the 50 cent relay would take up most of the back of the board...but it's simply a mounting convenience as I think it's easier to tack wires onto surface-mount pads that to pins.  The smaller Axicom relays run $3-4.  I don't know if they still sell what they used to call "surf boards" as in surface mount boards but these were simply convenience boards where you mounted surface mount parts and gave easier soldering access to the contacts on those parts.  So the back of board is really a "surf board" adapter that allows easier soldering access to a relay.

Besides the mounting of the relay, you have to have additional board area for the pad to solder the wire to.

True.  So maybe the current crop of 50 cent "full-size" SMD relays is too big but I figure SMD relays are only getting smaller though I don't know if there's standard footprint that is worth committing too.  So getting back to the "universal" aspect of this board.  The idea is you have some unused board space on the back.  So like the surfboards (which apparently are still sold), the idea is to come up with circuit topologies that might be useful in a piece of rolling stock used in conjunction with AC-to-DC conversion and voltage regulation.  Then put down some SMD pads to make mounting parts and soldering to them easier.  If the SMD relay is impractical for surfboarding, then oh well.

That's not a problem, I just don't know what kind of "surfing" might be appropriate.   Guessing at what might be useful is like a needle in the haystack.  Do you have any thoughts as to what you are thinking of?  Maybe a 555?

I have a finished sample of my power module.  I did some bench testing, and it works very well, not that it should come as a surprise, it's pretty simple.  I was able to draw 100ma with no additional heatsinking at 18V input into a 5V load on the outputs.  With a heatsink, I was using a large pair of forceps on the regulator tab, it got to 250ma easily.  At 250ma, I could see I was close to "out of gas" with the 300uf of on-board filtering, so additional filtering would be required.  All tests were done in "full-wave" configuration, using the half-wave common ground, the outputs would be lower before ripple became an issue.  With the current component values, the voltage adjustment range is about 1.5V to 12V, a useful range.

Here's the finished boards and the schematic.

Actually, I meant this to have the ability to be bolted to the chassis for additional heat sinking, that's why the regulator was positioned in that manner.  If you don't need additional heatsinking, I'd personally encase it in heatshrink.  When I heatshrink these, I punch a hole in the heatshrink to access the voltage control on the board.

gunrunnerjohn posted:

Actually, I meant this to have the ability to be bolted to the chassis for additional heat sinking, that's why the regulator was positioned in that manner...

As mentioned earlier, if you do this you have to insulate the metal tab from the chassis since the metal tab of the LM317T is not "ground" or at the chassis/outer-rail voltage.  This requires insulating washers and such which can be a nuisance to deal with.  Hence you can use the LM317P which costs a bit more but the package/mounting-hole is non-conductive.

In practice, I find that is can be inconvenient to just a nuisance to find a location on the metal chassis to drill a hole to mount the LM317 to provide heat-sinking.  I fully realize it may seem tedious but I still stand by my alternative option of using a 99 cent (free shipping) eBay DC-DC converter.  If 100mA (at 5V DC) is all you can get without resorting to some kind of heat sinking, adding a DC-DC module might be a mechanically simpler solution.  In other words, set the GRJ module to, say, 12V DC output and use a 99 cent DC-DC converter to drop it to 5V DC.  You can then get that 250 mA without needing a heat-sink or bolting anything to the chassis.  As I mentioned earlier, this would also provide 2 regulated voltages in a piece of rolling stock... for example 12V (or so) to drive 12V LED strips, and 5V to drive digital electronics.

I have given thought to what the O-gauge world needs now and it is a low-cost remote-control option to activate rolling stock.  In other words, get rid of those UCS or activation track sections that constrain where you can activate a piece of rolling stock.  Yes, you can buy into the ERR commander modules but that can run you $30+ and of course you need to have TMCC.  Instead, how about using GRJ's "Universal" power supply module to generate 12V DC.  Then use this to power an eBay RF remote relay module (less than $5 shipped).  The RF control can come from either a handheld fob (less than $5 shipped) if you don't use command control...or from the same fob connected to a ASC or AIU if you already use TMCC or DCS for remote control.

Well, I do have the other higher current option if you want to go that way, remember these?

AC-DC Power Module for Model Train Applications

It's the term "universal" that is the fly-in-the-ointment.  Those modules don't play well with a Z-4000 set to full-throttle (22+ VAC) - perhaps unusual but within the realm of reality.

semi-universal?

Nothing is "universal" and the things that claim to be universal usually don't work that well. I can see a bunch of applications for Johns tiny little board for lighting all kinds of things like marker lights on locos and cabooses/cabeese.

I must admit, though I'm sure many will see a use for my creations, my primary motivation is first to solve a problem that I have.  That is where most of my existing products came from, the Super-Chuffer, the Chuff-Generator, and the passenger car lighting stuff.  After doing some of these the hard way, I decided I needed a better method.

Matt, you'll like my next creation for just lighting a couple of LED's.  I have an even smaller board that just has a choke, diode, cap, and an LM78xx regulator.  I run across a lot of stuff where I just need 20-30 ma or so and I'd like to do that conveniently.  I still want to have DCS compatibility, but I don't need a lot of current.  C1 is an optional part that's used if the load is a long distance away, you don't necessarily have to include it.  This will be another thing that you'd just heatshrink and stick in a convenient place.

Well, you could stuff a CL2 into the socket and it would work, so that would be a perfect match for the LED loads.  I didn't think of that, but it's a great idea, same pinout!  No mods required, just pick a different part.  You do have to reverse it on the PCB, but that's not a big deal as long as you're aware of it.  VA is the input, and VB is the output.  Since the center lead is no connection internally, no problem.  Nice catch!

New schematic, the same except with the additional notes.

There was an instance where I wanted to install a 25mm case fan in a unit to cool the electronics because heat-sinking was not an option, this would be perfect

Again, strictly for the sake of discussion, if you are going to offer the LM78Lxx smaller board as a bare-board or unassembled kit (allowing end-user to select a CL2 or 78Lxx), how about a LM317LZ?  I figure you'd have to add one hole and one 0805 resistor which I think can fit without changing the board size.  The resistor wouldn't be used if U1 is a 78Lxx or CL2.  The LM317L would then allow the end-user to select the current rather than being limited to the CL2's 20mA. 

The CL2 is nice but still hasn't hit the eBay market so you pay 30-40 cents each at Mouser or DigiKey.  The LM317L is widely available on eBay for 5-10 cents. The 0805 resistor is a few pennies. 

Yep, but I'd have to spin the board again.   The CL2 just came out of left field and was free to fit into the existing board.

I truthfully wasn't thinking of an adjustable power supply when I did this, I figured I could do the "adjustment" using the different voltages of LM78Lxx chips.  I figured that for my uses, the 5.0V, 6.2V, 8.2V, 9.0V, 12V, 15V voltage selection seemed sufficient. 

Of course, I also don't get down to the 20-30 cent level here either.  Anyone that's actually trying to use one of these isn't trying to save 30 cents, or at least most folks wouldn't have that on the top of the list.