
This is the lightboard from the OSHI 25-901 dining car from the Tomix 92950 “Yumekukan” set. This board sits in a fitted pocket in the galley of the dining car. There are three SMD LEDs (the three white boxes on the left); three long lightpipes run from the LEDs to the rear of the car. The middle LED lights the taillights and signboard, and only lights when this car is at the end of the train—it doesn’t light when the car is at the head of the train. The outer two LEDs light two rows of table-lamps in the dining room of the car, and remain lit whichever direction the train is running. The two leads on the right connect directly to two steel strips that run along the bottom of the car and (in addition to providing much-needed ballast) contact pickups in the trucks. So, when +12V is fed across the leads (I don’t know which direction, to be honest), all three LEDs light; when -12V is provided, only the outer two LEDs light. In addition to the LEDs and resistors, there is what I’m guessing is an SMD bridge rectifier (for the table-lamps)? Although it has six pins instead of the usual four. And there’s some other tiny little resistor like thing by the middle LED.
The challenge before me: Convert this puppy to DCC. The board is too small to modify. And there’s no space to construct a replacement board. And, as many of my readers will know, controlling two independent lights with a DCC decoder requires three wires: Two “function” leads (that when activated, short to ground; when inactive they are left floating), and the +12V blue common. In my favor, there is a fair amount of room in the galley for additional circuitry, beyond just a decoder.
An aside: The last time I wrote about this lightboard, I suggested I had “a clever solution” to this problem. I was wrong, and I won’t bother detailing my many mistakes.
As it happens, Keith Norgrove and others at the MERG were thinking about the same problem (although for different reasons, I’m sure!). MERG member N C Friswell had already devised a very simple circuit for driving (some) bi-polar directional lighting boards. The problem is that this circuit presumes that the lightboard either does not have the necessary resistors to pull down the voltage for the LEDs, or that those resistors can be easily removed. My board (as noted) already has those resistors, and I can’t remove them without risking the entire thing (which is not replaceable if I mess up!). So, on a whim, I emailed Keith, and who in very kind response sent me a schematic.

Here is that schematic, neatened up a bit. Unspecified were the values of r1–3 and q1,2. Keith had only given partial specifications of those components, and so consultation with Digitrax (who manufactured my DCC booster) and TCS (who manufactured the FL4 decoder being used—A great big thanks to Jordan and John for their help!) was necessary to fill in the blanks. So here follows the component specification, and some justificatory and classificatory notes.
- r1: 56Ω ¼W
- r2, r3: 4700Ω ⅛W
- q1, q2: 2N2907 (PNP, 1A, 50V, 300mW in TO-92 package)
Q1 and q2 are transistors; since the function leads switch from floating to grounded, PNP transistors are called for (since they switch “on” when the base is grounded). Any small PNP transistor that can handle over about 120mA (preferably at least 200mA) @ 12V will do. This one is a little bit overkill, but it’s small, cheap and readily available at RadioShack.
R2 and r3 are just current-limiting resistors; if they weren’t in place, the current flowing out of the base of the transistors would likely be enough to fry them. An eighth of a Watt power dissipation is sufficient, but obviously larger ones will also do if that’s all you got. I wouldn’t go smaller.
R1 is special. You might wonder what it’s doing there, and rightly so. It’s there for one particular but very important circumstance, namely, so that you can put the car on the programming track without destroying the decoder.
On the programming track, the decoder is put into “service mode”. In service mode, your programmer will send commands to the decoder; the decoder is to respond, in most cases, with a “basic acknowledgement”. A basic acknowledgement is made by drawing a small amount of current, >60mA, for a short period, 6ms±1ms.
With the usual mobile decoders, this is accomplished by giving the loco’s motor a little nudge. This is why, when you’re programming your decoder, the locomotive occasionally jerks a little in response. But the FL4 decoder is a function-only decoder. To do the same thing, it simply activates all of its functions simultaneously, and hopes for the best. Now, if you have LEDs attached to only a couple of the function leads, at 20mA apiece, the decoder may not be able to draw enough current to register a basic acknowledgement. So, TCS recommends programming the FL4 before you install it, with a 100Ω resistor between one (and only one) function lead and the blue common. This would ensure enough current draw for a basic acknowledgment.
I want to be able to program my car post-installation. I’m prone to changing my mind about things. But, if I were to put the car with this circuit on the programming track without r1 installed, as you can see, both the function leads will short out, destroying the decoder (which is only rated to 200mA). In other words, the circuit does bad things when both functions are on, as invariably happens on the programming track (even if it can never happen in operations mode). R1 thus limits the current passing through the circuit to less than 200mA when both functions are on (in fact, at 12V, it limits the current to under 150mA). (This is on the reasonable assumption that q1 and q2 provide about 50Ω of resistance each, for a parallel total of 25Ω; 25Ω+56Ω = 81Ω, which at 16V = just under 200mA of current). When only one function is activated, in operations mode, it only removes about 1V off of the lighting circuit, meaning that the lights will be ever so slightly, but only ever so slightly, dimmer than they would otherwise be.
Here are some final thoughts on r1. First, you may notice that the usual calculations suggest that a 2.2kW resistor is called for. However, since it will ever only see a 200mA load for a fraction of a second, we don’t need such a large dissipation rating. Second, it might be possible, if you know that your train will never see voltages over 12V, to use an even smaller resistor to reduce the dimming of the lights in normal operating conditions. The 56Ω value was chosen assuming that my train might actually see 16V or more, and I wanted to keep the peak current below the FL4’s rated 200mA even at that high voltage. You might even find a decoder that has a larger current rating than that, permitting a very small resistor indeed. However, you should also be aware that while there is a standard for the minimum current draw needed for a basic acknowledgement, there is no standard for the maximum. So while, as I confirmed, the Digitrax Zephyr can handle as much as 1A during a basic acknowledgment, your programmer may not be that robust. Be sure to ask!
So there it is; By my calculations, the lightboard I’m using draws right at 15mA @ 12V; the lightboard in your hands may differ, and so you may find that the specifications above may differ for your application. But if you can’t cut up the lightboard in your shinkansen cab car (say), and you’ve got the space, this auxiliary circuit should do the trick.