Removing slide shoes from operating cars (formerly: "Reed switch"

gunrunnerjohn posted:

Maybe it's time for a compact track powered DC converter.  The hot ticket would be a switcher so  you could draw reasonable power from it without large heatsinks.

It's just that you can't buy the parts for less than the price of those completed eBay modules!  I've occasionally go around in circles on this with no conclusion...or else I would have already egg'd you on to build something to sell.  My latest half-baked thought was the 70 cent DC-DC buck using the MP1584 (instead of the chunkier LM2596) or similar stepdown switching IC:

These switch at 1 MHz so several times faster than the LM2596 hence a smaller inductor.  They don't have the large input cap  so I'm imagining how easy it would be for you to make a motherboard that this module sits on.  It would have the DCS inductor, bridge, electrolytic input cap.  These chips don't have the input voltage handling of the LM2596 or the linear LM317/78xx parts so while it pains me, the input might need clamping via a LM7824, 24V power zener, or the like.  I can't get over the thin form-factor relative to the LM2596 modules.

The problem with those modules is they keep changing them.  The only way to build something around them would be to buy a large lot and work off of the inventory.  I have two different small modules, one half the size of that one, they're no longer offered.

I looked at my p.17 and it had the optical parts you referred...so I thought we were looking at the same catalog!  But I see mine is Sept 2015.  I figure it ought to be a few pages away...

When you used the term "packages," I pictured a package containg a LED and a sensor.  I have no way, looking at the chart on p23 or VAP on the Jameco website, to figure which items would work together.

gunrunnerjohn posted:

 

The problem with those modules is they keep changing them.  The only way to build something around them would be to buy a large lot and work off of the inventory.  I have two different small modules, one half the size of that one, they're no longer offered.

Part of the fun is hitting a moving target.

So looking just at eBay for an apples-to-apples:

Buy 50 completed modules for 60 cents each....or buy just the 50 chips for a bit more!  Go figure.  Clearly this new-math is above my pay grade.  Anyway one might argue you could fab a hundred (or whatever) motherboards and inventory same qty of completed modules...and for next batch pay the re-tooling (if needed) for the next module style...and still come out ahead.

Bottom line is I believe such a widget would be useful.  And given the somewhat unique requirements of O-gauge AC such as the voltage range and the pesky DCS inductor, I can't imagine anyone else building this.

RJR posted:

When you used the term "packages," I pictured a package containg a LED and a sensor.  I have no way, looking at the chart on p23 or VAP on the Jameco website, to figure which items would work together.

I think I see where you're coming from.  Yes, there are packages that contain both the LED and the sensor used in, for example, the MTH tachometer circuit where an infrared beam reflects off the black/white stripes on a spinning flywheel to control speed.  Those chips have 4 legs as you previously observed, 2 for the LED, 2 for the sensor.

But in your coal car application the LED and sensor are physically separate so they are just 2-legs each.  The photo I showed with the sidelook style "package" (Fig.2 drawing in the Jameco table) is what I showed in the picture with the parts laying on top of a UCS electromagnet.  Are you planning to pursue this approach (optical beam)?   Jameco seems to only offer 1 LED and 1 sensor in the Fig. 2 "package" style and in your application they should work together.  Clearly there are more choices for the Fig. 3 package style but I don't know if the geometry would suit how they would be mounted on the track (LED) or on the bottom of the coal car (sensor).  If you use other than Jameco (such as DigiKey or even eBay) there might be more selection. 

stan2004 posted:

Bottom line is I believe such a widget would be useful.  And given the somewhat unique requirements of O-gauge AC such as the voltage range and the pesky DCS inductor, I can't imagine anyone else building this.

I ordered ten of them, I'll take a look at the size it can be made and see if I should whack out some boards.  I have to see how much bulk capacitance I'll need to provide a decent amount of power from the module, I'm sure it'll be a significant amount.

To go a step at a time,I cut the capacitor down to 22ufd.  The relay is rated for a max voltage of 30 volts, so I left full-wave rectification in.  There is a hum from the relay, so it probably should to back up to 47ufd.  The attached video shows the result, using an external permanent magnet and a window alarm sensor.  But I am still fascinated by the IR approach.  I found Mouser has paired devices.  I have to find a circuit to drive the relay.  Would the circuit above do it, leaving off the vcc feed?

Meanwhile:

RJR posted:

To go a step at a time,I cut the capacitor down to 22ufd.  The relay is rated for a max voltage of 30 volts, so I left full-wave rectification in.  There is a hum from the relay, so it probably should to back up to 47ufd.

What is the part number of your 12V DC relay that is rated for 30 Volts? 

Did you ever try the flyback diode method (without the capacitor) to eliminate buzzing?

RJR posted:

But I am still fascinated by the IR approach.  I found Mouser has paired devices.  I have to find a circuit to drive the relay.  Would the circuit above do it, leaving off the vcc feed?

I'll put together a circuit for you - yes, essentially something like earlier circuit.  What is your parts "stash"?  Do you have diodes, bridge rectifiers, transistors, resistors, capacitors, etc. of various sorts just lying around?

Zettler AZ942-1CT-12DE is the relay.

I didn't try the flyback method.  Don't have appropriate diodes handy, and my local RS only has power diodes with very thick leads.

I have some resistors, rectifiers and capacitors of recent vintage, plus many that are of such age that I wouldn't use them.  It's easy enough to order from Jameco, Digikey, or Mouser.  Speaking of IR sensors, Mouser stocks the Lite-On Electronics LTR-301 and LTE-301 complementary units.   Looking at the data sheets for these, I wonder if they need a dedicated power supply????

Occurs to me that if these are placed alongside rather than under the tracks, location could be easily adjusted for more precisely spotting cars.

Using the reed switch it isn't a problem, but it's worth noting that if you use any solid-state semiconductors, such as hall effect sensors, photo-transistors, or even your regular transistors, the 'fly-back' diode across the relay coil becomes a necessary part. As the relay coil de-enegerizes as it's power source is turned off it will discharge as the magnetic field collapses.  This will send a spike backwards through the other parts and if not all at once, will damage them over time.  This is the same process a at work with the TVS diodes used to protect the electronics in your engines... you can just get away with a 1 cent diode in DC circuits.  

The recommendation on using half wave to power the relay, rather than full-wave is two fold.  There is the obvious aspect that over time applying 25+VDC to a coil designed for 12VDC may damage it, but the main reason I would use the half wave is that at 12, or so, volts the coil will draw half as much current.  You can use a smaller capacitor to hold it on, and run less current through your reed switch.  

JGL

If you want to dabble with the IR method, I just cobbled this together.

PT1 is the 2-legged IR phototransistor.  Resistor R1 is selected for sensitivity of the phototransistor detector.  It will be something in the range of, say, 470 Ohms to 4700 Ohms.  And here it is in action.  I had the 2-lamp "load" from another thread about red/green signals so used that hooked to the SPDT relay outputs to show when it switched.  The IR LED is being driven by 20mA.  R1 was chosen to give it the 1" approximate sensitivity and happened to be 1.5k Ohms as shown.  

I'm sure the Mouser part pair would work just fine; there's nothing particularly "demanding" about the components.  But  before you make a Mouser order (shipping charge will be more than the component costs) suggest making sure it's clear what you're trying to do.  If you're still experimenting with reed switches, hall-sensors, etc., that's fine but get those parts ordered too.  If you're thinking of mounting on the side for better visual alignment, that may mean getting an assortment of resistors or maybe using a different "package" style of LED emitter or phototransistor.  I realize you have a 12V DC supply that must put out several Amps to drive the UCS magnet, but you don't need that kind of power to drive a tiny IR LED.  So I figure maybe you can just take the AC coil voltage, put it thru a bridge-rectifier and capacitor and resistor and drive 20mA into the IR LED.  Etc..  Measure twice, solder once!

Thanks, John, for reminding me of the relay spike issue.  I would have forgotten that.  Used diodes for that purpose back in the last century to reduce arcing at relay points.

Thanks, Stan, for taking the time to work up a schematic.  Did you actually put together the breadboard I see in the video using this circuit?

The 12-volt supply I was using is not part of the layout, but for occasions when I need that kind of power to test devices.  I do have <1 amp 12VDC power readily available, to drive electronic circuits, with a bunch of Chinese AC>DC buck converters that can be set for most any voltage.  I am not wedded to using the UCS tracks.

At this time, the IR route looks much more appealing than any of the alternatives.  I'm going to give it a try.  Due to an unexpected request from a client, it may be awhile before I can complete it. (I also, in a senior moment, blew a PS2 board that I have to get to.)  I will definitely keep you posted.

And once again, Stan & John, I appreciate your help.

RJR posted:

Did you actually put together the breadboard I see in the video using this circuit?

The 12-volt supply I was using is not part of the layout, but for occasions when I need that kind of power to test devices.  I do have <1 amp 12VDC power readily available, to drive electronic circuits, with a bunch of Chinese AC>DC buck converters that can be set for most any voltage.  I am not wedded to using the UCS tracks.

Yes I assembled the circuit shown in schematic using a solder-less prototype board and shot the video. 

I assume you have a multimeter handy. While you can't see that an IR LED is "ON" it's become common knowledge that you can see them glowing purple-ish with a digital camera,  A meter comes in handy to measure the current going into the LED. You can usually tell if you're overdriving an incandescent bulb or a visible LED because it just looks too bright! 

You're probably familiar with those LED-calculator programs available everywhere to select a resistor value.  Unlike white LEDs which run ~3V, red LEDs which run ~2V, IR LEDs run closer to ~1V.  You'll also find IR LEDs are typically driven by MORE than 20mA which is commonly associated with visible light LEDs.  The datasheet will clearly show all this.

If it's convenient to supply DC from elsewhere then great.  Sometimes it's just a wiring nuisance so I figured you might just pick off track AC since obviously you'd have it there since the coal car itself must be powered to operate.  So if you run a DC output wall-wart, or eBay DC-output voltage, etc. then you just need a suitable resistor and of course the pushbutton activation switch.  Otherwise, add a bridge rectifier and capacitor to create LED DC supply from track voltage.

Stan, putting my client on the shelf for a breatherI just ordered the parts to try your "cobbled" circuit; enough parts to build 15.  Meanwhile, I did get in some magnetic switches to try.  Will keep all posted.

I'll update the schematic so we have an up-to-date drawing.  So by parts you got the IR pair from Mouser - the LTR-301 phototransistor and the LTR-302 LED?  And you plan to use a 12V DC wall-wart to drive the LED?

No as to the wall wart.  I fabricate small voltage regulators for whatever voltage I need, plus I have a dozen or so Chinese buck converters on the shelf.  Yes as to Mouser.

As long as I had the new magnetic proximity sensors, Cherry MP201901 (datasheet attached), I figured I'd try them.  They seem to work ok, on an AC-powered UCS track magnet.  They can switch 0.5 amp.  I wonder if the sensor is opening and closing 60 (120?) times per second and wondering what effect, if any, this will have on the unit's life.  Any thoughts?  There is a 45ufd cap downstream from the  rectifier.

With an "operate time" of 1 millisec, the reed switch is clicking/buzzing on/off in the presence of the 60 Hz (16 millisec) AC-powered UCS magnet.  That datasheet doesn't give operational life-cycle data but for that size switch they are in the 10's to 100's of thousands of cycles.  So if it buzzes, say, 100 times on each momentary button press you'd have years of operation...let's be realistic, how many times do you dump coal per day?!  But, sure, if convenient/practical I'd drive the UCS with DC or with rectified-AC plus a capacitor.  OTOH if the sensor is a Hall semiconductor device, it's probably easier/cheaper to filter the 60 Hz on-off cycling with a capacitor in the coal-car (at the Hall sensor output).

So here's where I think we're at for the Infrared optical approach.  The LTR-302 IR emitter has a typical voltage of 1.2V at 20mA of current.  So, for example if you have a 5V DC output from the eBay module, R2 could be a common 220 Ohm.  As mentioned previously select R1 for sensitivity - a higher value makes it more sensitive (responds to less IR energy).  A good starting point would be a 1k Ohm resistor.

Obviously use some common sense in selecting R1.  That is, once you have 10 coal cars built up, it's much easier to change one R2 than ten R1's if you have a sensitivity or ambient-light rejection issue!

Thanks, Stan.  Given that the reed is opening and closing 120 times per second and that it takes several seconds for a Lionel dump car to unload, and even longer for an MTH motorized dumper, and many pulses for a milk car,  I think your MTBF estimate is overly optimistic.

Applying DC theoretically is easier.  Given that the source of the AC to power the rectifier would be the same bank of transformers feeding the track (all are grounded together), I would disconnect coil ground wire from outer rails and lead it to outside.  Place a full-wave rectifier and cap at the screw terminals, with the coil ground connected to to one the rectifier outputs. 

I'll do one car and put it on layout, while awaiting the receipt of the IR components.

Thanks again. RJR

BTW, do you use or plan to use DCS?   If so you'll need to add the 22uH inductor/choke for each car.

If you're going to use DC, after the bridge you'll want a filter cap.  With a 1ms operate time for the reed, it'll probably react to the pulsating DC, even full wave.

I finished the prototype car, with the Cherry Magnetic Proximity sensor.  I bought the battered car, cheap, in York last trip.  Attached is a photo of the bottom, showing slide shoes gone and a pickup roller.  Also a video showing it in operation.  Took me a bit to realize I have to press the uncouple button, not the unload.  I plan to work on an IR version.

No question, GRJ.  I do have a 45ufd downstream from the rectifier.  I should probably have a ceramic disc across the reed points also. 

I think he means a capacitor after the bridge on the track-side if you drive the UCS coil with DC.  Since the current draw in the UCS coil is quite large, the cap has to be quite large.  It's a do the math situation but I'd think we're talking in the 1000 uF range to keep the reed switch on the coal car from chattering on each zero-crossing.

RJR posted:

Took me a bit to realize I have to press the uncouple button, not the unload.

That's pretty funny!

Driving the UCS with DC presents a problem, in that my power source comes from a bank of common-grounded transformers, as I noted above.  However, I have a bunch of old but unused Gargraves uncoupling tracks.  On these both sides of the coil can be accessed without any connection to layout ground.  I can just put a bridge and cap by each track, and hook up the wire from the button to that.

As an aside, I pulled these out of the box where they've been sitting and none of the magnets would work.  Some investigation disclosed that Gargraves had used blackened very short (1/8", 4-40) screws, that screw into a threaded copper tube from top and bottom to hold wires.  Replaced these with new screws and they work.  I'll have to see if the Gargraves can activate the magnet sensor. 

Update.  Tonight I hooked up a Gargraves uncoupling track with the magnet driven through a full-wave rectifier with 2200 mfd cap.  Activated the sensor ok, but the track magnet got hot very fast.  Not sure if the reed vibrated.

Almost finished a IR activation. Seems to need more space.

Using DC, you may have to reduce the voltage as the current will be greater than the same AC voltage, no inductance for the DC current to fight with.

Yes, GRJ.  The Gargraves track is easier to wire, but even with the Gargraves-furnished booster plate, it draws more amperage and has less pull that the old UCS.  I'd like to find a magnet-activated switch that is not a reed switch, so it doesn't make-and-break on AC or rectified DC.

Stan & John, I fabricated a board using IR.  Is a bit more complicated than the magnet system but should last longer than a reed switch.  Now I have to decide which route to go.

Did you ever get the Hall sensor components?  If going the magnet route, a Hall device (plus a capacitor) addresses the mechanical on/off reliability issue.

I suppose experimenting with the alternatives is part of the fun - but the overheating electromagnet also seems like a long-term reliability issue.  I realize your system is a one-off and it appears it's down to how well you can dodge the peanuts being thrown from the gallery!  The reality is if you want an engineered design you should measure the available magnetic field strength 1" off the track (or whatever the gap from track magnet to sensor), then select a magnetic sensor (reed, Hall, whatever) with suitable sensitivity.  You can buy magnetic field strength probes, but for hobbyists they can be built for a few bucks and hook up to a voltmeter to measure the field strength (Gauss, mTesla, Amp-turns, whatever)...or an oscilloscope to see the changing field from an electromagnet driven by AC or rectified DC.  I'm not suggesting you do this and don't want to distract you from making progress, but here's one I use.  Basically you get a so-called analog Hall sensor which puts out a voltage proportional to field strength (rather that a digital on/off sensor which trips at a threshold):

The point is an engineered design lets you approach the problem from a system level to, perhaps, lower the drive power into your Gargraves magnet (to solve the heating issue) in conjunction with choosing a suitably sensitive magnetic sensor.  In the case of the Cherry switch and magnet, they "hide" the technical parameters and since you'd presumably buy them as a pair, you just get an end-user specification of operating distance.

Yet another alternative is to have your coal-cars sense BOTH IR and magnetic!  One relay, one AC-to-DC conversion (albeit that's just a diode/capacitor), etc.  The two sensors can be wired up so either can trip the relay.  I don't know why you'd do this but just saying...

stan2004 posted:

Did you ever get the Hall sensor components?  If going the magnet route, a Hall device (plus a capacitor) addresses the mechanical on/off reliability issue.

I suppose experimenting with the alternatives is part of the fun - but the overheating electromagnet also seems like a long-term reliability issue. 

 

I have started to use "undersized" resettable fuses in all my uncoupler and similar feeds. I pick one that trips in 5 seconds or so at the actual load, after picking out the correct supply voltage. Normally, they never get to the trip point. Waiting to see if I get any failure of the fuses after repeated trips, but so far so good.

As far as the chatter problem goes, I personally would bite the bullet and use DC on the uncoupling track. Yes, it is one more power supply, but that is probably easier to deal with because you don't have the space constraints of car- mounted equipment.

In the past, I used the expedient of an old motorcycle battery or gel-cell with a wall wart arranged as a trickle charger to run some of the "rattley" PW accessories and uncouplers. Not sure I would do it that way today, but it gave me a whole lot of current and very pure DC for almost no money. And no ground references to worry about.

Just my 2 cents worth. 

Prof, I have a "breadboard" (actually a plank) circuit I designed that shuts off power after it has been running a predetermined number of seconds.  Intent was to use it if I ever decided to use DCS accessory circuits to operate uncoupling tracks.  Haven't proceeded with that revision.

Your use of resettable fuses---obviously you size them to take advantage of the fact that a fuse doesn't blow instantaneously on overloads---is intriguing.  You ought to post the brand, rating, and time delay to blow of what you use, and what types of uncoupling tracks.

I have proceeded to the point where I'm applying DC to the uncoupling track.  However, at this point I have not ascertained whether the pulsating DC derived from a rectifier with a large capacitor is causing the reed to make-&-break.  I plan some experimenting later today if I can find the time.

Stan, some operating cars, like the milk car, require longer magnet-on durations.  Unloading a string of milk cars requires considerable on-time.  Overheating could become an issue.

I'm going through the windshield wiper syndrome:  first lean towards magnets, then to IR. 

I can't find any non-reed mechanical switches for the former. If I can eliminate reed vibration possibility , it is a simpler-to-fabricate system. 

The emitter for the latter could be mounted between the rails of Gargraves track, facing upwards, and fed from normal uncoupling power through a small full-wave rectifier with a small capacitor and a resistor to hold current to 20ma.  The sensor could be on the bottom of the car, facing down and therefore minimizing the effect of ambient light.  But the on-board control, with more components (e.g., transistor & resistor) to stuff in a car than the magnet approach, presents space issues (prewar log cars have zero underdeck space).

My first attempt at producing an IR control board was interesting.  On first try, a soldering error caused the input diode to short and the transistor failed.  Found that with a large magnifier, cleared the joint, and replaced the transistor.  Then I found the relay would mull in as soon as I applied power.  Took a few minutes to realize the lighting in my workshop was activating the system.  I did use the 1k ohm resistor you recommended, and find that I get about a 2-3" range.  Any thoughts from your experience on whether a well-lit train room would affect a IR sensor facing downward under a car?

I really don't understand the Hall effect sensors.

Thought occurred to me that when reed switch is vibrating, effective voltage is reduced, so I ran a test.  I used a 4 D-cell battery pack.  Meter read 6.3 volts across leads.  I connected the pack to the test meter it through another magnetic reed switch, and held a permanent magnet near it.*    Voltage went from 0 to 6.33.  Then I held the switch over a 60 Hz UCS track, and got a reading about 3.5.  Then I fed the UCS through a full-wave rectifier with a 2200 mfd capacitor and got 6.33 volts.  To me, this mean that the reed was staying closed.  Would you, Stan & JGL agree on the conclusion?

*It appeared that after encountering the permanent magnet, a strong one, the switch would stick.  Don't know if the switch is defective or a strong permanent magnet jams it.

Given the price of these sensors, $3.95, that I've 11 cars to convert, and that each also needs the coupler and plate assemblies changed in each truck no matter which way I go, tempts me further to the IR approach as more economical.

RJR posted:

... But the on-board control, with more components (e.g., transistor & resistor) to stuff in a car than the magnet approach, presents space issues (prewar log cars have zero underdeck space).

...Took a few minutes to realize the lighting in my workshop was activating the system.  I did use the 1k ohm resistor you recommended, and find that I get about a 2-3" range.  Any thoughts from your experience on whether a well-lit train room would affect a IR sensor facing downward under a car?

I really don't understand the Hall effect sensors.

Not to be argumentative, but I find it puzzling you say that the few additional components to support the IR phototransistor is a space issue.  That is, the incremental components circled in red above what you "need" for the reed-switch plus relay method seems small in comparison (the large relay being the elephant in the corner!).  I may be mis-understanding the space situation wrt where you are trying to stuff the components on the variants of your operating accessory cars though.

As for the IR sensitivity issue, 2-3" is probably too much sensitivity.  If you lower 1k to, say, 470 the operating distance will drop and simultaneously make it less sensitive to ambient/stray IR energy from your workshop overhead lighting.  However, if you simply aim the phototransistor lens downward and put an opaque backing (to simulate the bottom), I'll bet it won't trip at all even with 1k ohm.  When the phototransistor is installed, you can also put a small skirt/shroud around it so it only sees straight down rather than gather IR noise reflecting off the track bed.  Note that if you decrease the 1k, you can always increase the IR emitter power using a stronger IR LED to compensate which would still consume a tiny fraction of the energy you need to drive that electromagnet.  I am 100% confident the IR method will work even in the brightest train room.

As for the Hall sensors, think of them like the phototransistor.  The phototransistor turns "on" in the presence of IR energy.  The Hall sensor turns "on" in the presence of a magnetic field.  They are both solid-state or no-moving parts, and about the same size.  As shown earlier via schematics, they both require a few extra components (resistor, transistor, diode) to mate them to the existing DC relay circuit.  If you want to pursue the Hall sensor approach I will cobble together a proof-of-concept circuit and shoot a video as I did with the phototransistor method; as with the IR method I am 100% confident it will work.

One final observation if size matters.  I know you already have the relays in hand, but if starting from scratch, I'd look at whether you can drive the solenoid in the car(s) with rectified DC.  If so, then you can use a solid-state power transistor to switch the DC current.  A 50 cent power transistor can switch Amps of DC current and so much smaller than a relay...and removes another "moving part" though I don't think you'll have a reliability issue with the relay.

Thinking outside the lines here-

Not sure what kind of car you are trying to activate, but here goes. If you are resigned to putting some electronics on the car, maybe a magnetic pickup, transformer style, is the way to go. Maybe 100 turns or so of #34 or 36 positioned squarely above the UCS magnet. Not sure if you could disguise that in your cars, but throwing it out there for consideration. I was thinking of a coil about the size of a sewing bobbin. Might even be able to use the coil out of an old O-27 uncoupler.

I will fiddle around and try to get a feel for what can be done with a coil suspended above a UCS magnet, just for the heck of it.

Prof, what I am trying to do is eliminate the sliding shoes on operating cars: log, coal dump, milk, etc.  There have varying levels of available space.  Some electronic means of transmitting a signal that will connect power from third rail seems to be the way to go.  Some of these cars have little undercar clearance; many have steel or metal bodies that would block magnetism.  When there's a lot of operating to be done, like unloading milk cars, the track magnet does get warm. 

I am intrigued by your use of resettable fuses on uncoupling tracks.  I was speaking to a well-known guru tonight, and when I mentioned it, his ears perked up too.

Stan, I don't think you're argumentative at all.  You've been giving me good guidance.  I do very much appreciate the schematics you've worked up (even labeling the transistors so I don't mess up the symbols).

You have apparently been designing circuits, both schematically and geographically, for some time.  Laying out the components is not something I'm skilled at, so I may consume more volume than if I had more experience.  50 years ago, when I ran RC boats, I had plenty of space in which to squeeze capacitors, relays, etc.

I appreciate your offer to design a hall effect circuit.  However, since that involves the track magnet, I'm going to tackle another coal dumper and install an IR to see what happens.

Stan, using your latest full schematic nearer the top of this page, I have built 2 IR receiving units, and they both work perfectly, but I need your advice on sizing a component in the IR emitter.  For initial testing, I powered it with 2 1.5 volt batteries in series, feeding through a 100 ohm resistor.  But for the final version, I will have to power it vis 16-18-VAC.  I initially fashioned a power supply using a full-wave rectified, a .47uf capacitor, and 670-ohm resistor.  I found this yielded 20VDC.  So I tried a different approach.  A diode (half-wave) and a 470-ohm resistor.  However, the LED must flash because the relay chatters.  Any suggestions on the ratings of a cap to put on the diode output?

47uF/35V like the one used in the car should work fine.  Note that whether half-wave (single diode) or full-wave (bridge-diode), the capacitor in this application charges up to 20VDC.  Do you really have a 670-ohm resistor (vs. 680-ohm)?  Note that the resistor dissipates 1/2 Watt or so.  And while it should only be "on" for a few seconds at a time, any good design should not stress out the resistor if the stuck on.  So while I wouldn't expect you to have a 1 Watt resistor lying around, how about two 330-ohm in series making a 660-ohm resistor with twice the power capability.

Also, it's a bit frustrating but as you know the IR LED is operating at about 1.3V DC and when starting from 20V DC, less than 10% of the power is going into the IR LED...over 90% is wasted as heat in the resistor!  Well, it still draws a fraction of the power as the electromagnet.  In the amusement category, it occurred to me that if you simply use your 16-18V AC source and drive a small incandescent bulb (like the ones in a track lock-on), those bulbs generate enough IR energy that I'm sure it will trip the IR detector in the cars!  And of course the filament time-constant means no worries about relay chatter even with the bulb driven by AC.  No diodes, no capacitors, no resistors...just a tiny bulb!  And you get a "free" indicator light whenever you push the button.  Yeah, a bit tongue in cheek but then again...

Why not use a simple regulator circuit to power it?  Being the lazy sort, I just use the standard 78L05 with a diode and 100uf 35V cap.  If you're driving with 20-30ma, this will work fine over a wide voltage range and supply the LED with a constant voltage source.  Keep the resistor, only make it a 200 ohm, and you're all set.  I've used this in a number of applications, works great for a low current 5V source.