KATO 2016-1


10/21/15- I'd been tempted by this steamer ever since getting Kato's Orient Express passenger train set. Despite being curious about how well Kato did steamers, I'd talked myself out of it because I have other German locomotives and the Japanese electric EF58-61 to haul the Orient Express.

Moreover, I was overwhelmed by all the different variations of Kato's D51 locomotive: A pricey "Roundhouse" version of #498 specifically for the Orient Express with fancy gold lettering on the tender, a Hokkaido version, a Tohoku version, a plain black version of #498... and what seemed to be many more based on Kato's model numbers (2016-1, 2, 3..etc.) A decision like this requires research, which isn't easy when all the websites that might be helpful are in Japanese. Whine, whine, whine.

The biggest question was DCC: Not DCC-able, not interested. Ideally, with a sound decoder. Once I'd seen that others had installed sound decoders in them and read that it could run on a R249 radius, I was lured back in. It looked like installing a sound decoder could be a stiff challenge, which is good: The deepest gratification often comes from a successful journey, not from the destination. I chose the 2016-1 with the minor gold bling because it was a safe choice with good reviews.

The Locomotive in Hand: Kato did a great job; it's a super-detailed little loco that I have to use Optivisiors to fully appreciate. And it is a tiny locomotive. Somehow the pictures seem to make it look bigger, and I expected a 2-8-2 locomotive to be bigger than my 4-6-2 German BR-01. It's not.

DCCing It: This was what I was most looking forward to (and dreading). When you get a new loco, the first thing you want to do is put it on the track for a test run to see if it lives up to expectations. With a DCC track, it's a bit different. You put it on the track to see if it works or not. Until it's DCC'd, you won't get any pleasure in running it. So deferred gratification is a given, and depending on the difficulty of the job there are the questions of how long it will take to get there and what unforseen problems you'll face. There's the possibility that what you're trying to do won't work out to your satisfaction. That's what makes it interesting.

Info, Please... The best online info that I found for how to DCC this version of the D51 (there's an older version) was at this Japanese website. It has lots of pictures of the D51 in various stages of disassembly. The commentary is in Japanese so I had to rely on the pics to make guesses about how to proceed. In this article, I refer to the pictures at that website, but I supplement them with my own to hopefully clarify some things.

Warning: If you're not comfortable with soldering, tinkering, and doing irreversible modifications, don't try this. You could conceivably end up with a broken or damaged locomotive (and it will be solely your responsibilty to fix or make peace with)!

The author starts with pictures of the engine disassembly, and eventually gets to the tender. I believe that for a sound decoder installation, the tender is the starting point because that's where the decoder and the speaker go. They're the critical unknown about the project, and if the parts won't fit, there's no point in proceeding.

Disassembling the Tender: At the end of that article, the pics show the tender at various stages of disassembly. I studied the pictures, but didn't have a clue about how to get started! This one's like a puzzle. Unlike most passenger cars, you don't pry along the bottom edge to release tabs and pull the top off. From studying the pics, I figured out that the pry-able part was at the top: The coal load, lead (white metal) weight and brass contacts are all part of the top section that fits in the "sleeve" of the body and trucks (the bottom section). It's not easy to remove the top section from the top since there's little to grasp at the top. Actually, the locking tabs are located on the underside, towards the center, hidden behind the inside axles of the bogies. To remove the top section as intended, remove the bogies first... but don't ask me how!

...because that wasn't how I did it. I managed to get a good enough grasp of the top section by prying the sides apart, and pulling; somehow that released the underside tabs without breaking anything. I didn't realize the correct way until I had removed the top and saw the underside tabs. Removing that top section was probably the hardest part of the entire project.

The top section (A & B) are separated from the bottom section (C).

Pic (B.) side view shows the tabs that lock the top section into the bottom section (C).

Remove the top section cover (A) from the top section lower half by lifting the clips at the sides. The top section (A) doesn't have anything of interest. You can remove the coal load if you want, or drill holes through it (customarily, to let the sound out), but since I ended up with a downwards-firing speaker, there didn't seem to be any point in that.

Section B top view is where the decoder and speaker go. The lead cage can be lifted out.

The side view shows the slots where the speaker fits.

The bottom view (ignore the white piece of styrene) shows where the power coupler from the engine clips on (the post at the left). Track power wires from the decoder can be soldered to the top side left edge of the brass strips. Remove the strips before soldering.

The Tender

A. Top section, upper half (cover), separated from (B.)

B. Top section, lower half, view from side, top and bottom

C. Bottom section, view from top and bottom

Decoder and Speaker Fitting: With the top off, you can start planning and testing where you'll put the decoder and speaker, and where the wiring will run. I could see that the ESU LokSound decoder would fit perfectly within the two-part lead "cage"-- such a snug fit that it seemed like the two were made for each other. Similarly, the Sugarcube speaker fit the width perfectly. However, if you put them both in, they wouldn't fit the vertical space. I toyed with the idea of modifying the coal load and doing a mutant version of the D51 tender. However, it was not to be, so I picked a round speaker and mounted it like the Japanese website showed.

I'd thought that the author had milled out the side slots for the speaker, but discovered that mine had the same slots. Not only that, but the speaker rested in an circular area within the lead (pewter?) frame, almost as if the frame had been designed for it. With the speaker placed here, the decoder fit perfectly on top, without any vertical space problems.

The extra wires should be clipped off or clipped short since they take up space. The other wires should be clipped to an appropriate length... but always leave a margin to aid in assembly. (It's not as easy to fix a wire that's too short versus one that's too long.)

Planning the Decoder Installation: At this point the project seemed very do-able. The split lead frame conducted power from the trucks to the engine via the coupler, so the frame would need to be insulated to keep the speaker from shorting the two halves. I coated the slotted areas with hot glue, which not only electrically insulated them, but provided some cushioning for the speaker. Volt-Ohmeter test showed no shorting.

The trucks conduct power to a pair of parallel brass strips that sit under the lead frame halves. The lead halves (which can't be soldered) are designed with rectangular sections that press down onto the brass strips; however, towards the front, there's a void underneath that leaves room where the brass strips can be soldered and a cutout where wire can be run to the decoder. The order of assembly/disassembly is important since you don't want to solder the brass strips while they're seated in the plastic frame.

The tender is constructed with a light pipe for a rear LED, although one isn't installed. There's probably enough room to install one though, and with the decoder right there, it's probably an easy addition. (I was more interested in getting a running locomotive though.)

Moving on to the Engine: With the speaker and power pickups taken care of, that leaves four decoder wires that need to go to the engine: Two for the motor, and two for the headlight. The split in the frames provides a convenient way to feed those wires out the tender. This is the point where attention turns to the engine. (I did the actual decoder and speaker installations after looking at the engine because I still wasn't sure it was going to work.)

The Motor: The engine turned out to be easy to disassemble. As the Japanese website shows, you just lift up from the cab at the rear. Once you've got the housing off, you can see that the motor has two leads that go... somewhere inside. I expected them to be soldered and thought I'd have to cut them, but gave them a tug and lo and behold: They're soldered to insertable brass strips just like Kato's light kits! Very convenient.

The Headlight: The headlight was another area to explore. It looks like the split metal frame needs to be separated to get the headlight circuit board out, and it looks like the job starts with two screws on the bottom plate that holds the drive wheels in place. If you're tempted to unscrew them, don't!

If you do, the wheels will likely want to come out with drive rods attached. If they do, the brass electrical pickup nuts at the sides will lift out and you'll have a heckuva time getting all of them back in place. The biggest fear is that the drive rod alignment will get messed up and if that happens, it can be a nightmare to get them aligned so that the wheels turn like they're supposed to. (In theory, based on my experience with my BR-01.) The point is, you don't need to split the frame to get the headlight circuitboard out. If you push it back from the front slot opening and bend the circuitboard, it will clear the flywheel and slide out (just don't snap it in half while bending it). It's a slide-in circuitboard with two pads that make contact with both halves of the split frame-- which apparently, conducts power from the wheels.

I was very tempted to leave the headlight assembly in place to work as designed for DC (and vary in brightness by the speed). A DCC track will light the LED, and there's no reverse light in the tender for direction change. Controlling it by the decoder would mean running two additional wires from the decoder to the front of the engine. I had a feeling that the body shell fit was very tight, with zero room for adding wires that would need to traverse its length.

This Loco Was Designed For DCC! Then I noticed some channels in the lead weight. The channels seemed to be perfectly placed to run a pair of wires from the headlight circuitboard to the back of the engine! It was almost as if Kato had designed this loco to be DCC-able, and they'd left breadcrumbs to show the way. The first clues were in the design of the lead frame in the tender: It's not a coincidence that it had been the perfect size to fit a decoder, and that there was a circular area that fit a round speaker. Although I didn't run across a sound decoder equipped D51 in Kato's lineup, chances are they've made one or will make one.

The LED Lens: At some point, this will likely pop off. It's a tiny clear piece, so don't lose it. Reinstall it before reassembly.

Modifying the Headlight Circuitboard: The circuitboard gets power from the two halves of the frame, so the circuitboard needs to be isolated to receive power only from the decoder. I ground off the two contact pads (leaving the PC board intact so it could be reinstalled) and soldered wires to where the traces lead. Conveniently, there was a large, unused pad at the back of one side. The other solder point had to be created by scraping away some of the PC board trace insulation. I tested the LED with a DC power supply to determine the lead polarity; ESU uses the positive leg (blue wire) for common, and the white wire for the cathode headlight lead (yellow for the tail light). I used a brown wire (top) for the positive lead since I didn't have a long enough blue wire and the decoder's wouldn't reach.

If installing a taillight, the blue wire would need to be tapped into before leaving the tender.).

Routing the Headlight Wires: It's easier to do this with the parts tacked down in place. I used contact cement (Goodyear's Pliobond) to tack down the circuitboard, and for insurance, put a dab of Gel Superglue in the front opening to ensure it didn't move. The fit was looser after grinding the pads, and I wanted to make sure it didn't move and allow the wires to rub against the flywheel. Likewise, the lead weight was glued in place with Pliobond.

Pliobond is great for this because it's rubbery; it's got some give, dampens vibrations, bonds dissimilar materials (like metal) and parts can be peeled up. Generally, it doesn't molecularly bond with stuff so it doesn't damage the surface... or so I thought! I used it to assemble the Sugarcube speaker during my tender planning phase and inadvertently set it down with the tip touching tender body shell. After discovering this, I scraped off the glue with my fingernail and discovered that it had eaten into the plastic, leaving a small gouge! I'd never seen this happen before. So caution: The plastic used in this locomotive is very sensitive to solvents! Lesson: Stupid mistakes happen. Keep parts that you're not working on away from your work area.

The LED wiring was tacked in place in the lead weight's channels, run along the top across the flywheels to the nylon piece above the motor (which conveniently has wire-routing nibs). I put a dab of Superglue at the end to deter the wires from being pulled out, although I'm not sure that it served a purpose: Nylon's pretty impossible to glue.

Wiring to the Tender: I'd planned to use a tiny (1 mm pin spacing) 4-pin circuitboard header plug/socket to connect the engine and tender wires. This seemed like a good idea since the power-conducting coupler rod could be disconnected, which was very useful when working on the engine and tender separately. However, soldering the tiny pins on the plug was difficult and the pins didn't want to stay seated in the plug. My thinnest heat-shrink tubing looked huge on the plug. I realized that even though it was an extremely tiny plug/socket, if I ever managed to put it together, the result would be bulky and fragile, and extremely difficult to plug/unplug, even with tweezers.

I abandoned that idea and soldered the wires together, insulated with heat-shrink tubing. The downside: If I ever needed to separate the engine and tender to work on them, I would need to cut the wires. However, four independent wires have more flexibility than four wires bundled together at a connector, which makes them less likely to bind/interfere with the coupling while navigating a turn.

I'd left a generous length of wiring that made the soldering easier; I was concerned that it would be difficult to conceal them since they were so colorful. However, the excess was easily stuffed in the spaces in the tender and engine. They're so closely coupled that the wire's aren't noticeable unless you look for them from the underside. I was going to paint them black, but I don't think it's necessary.

Test Run: The moment of truth, at last! I did the first test run before putting the body shells on because you never know... Fortunately, it worked the very first time. The motor drive worked as expected, the headlight turned on/off, and the sound worked. That was a suprise since I'd forgotten that ESU puts a stripped-down diesel engine soundset in the decoder. The body shells were attached and fit easily-- always a good sign. Success! I went to the ESU website, downloaded a USA Mikado sound file, fired up the LokProgrammer, and installed it.

Running it with the Mikado soundset, I was disappointed. It was thin, tinny, and not very loud. All my other sound-decoder equipped locos are much louder and fuller, and the BR-01 running on the track at the same time drowned it out. The Mikado soundset wasn't very loud probably because the chuffs were muffled (bassy) sounding. I replaced the chuffs, brake, and flange squeals from a BR-01 soundset which made the sounds louder and more authoritative; however, anything that had bass content (like the chuffs) still sounded thin; the brake and flange squeals had more treble and sounded unnaturally louder, so I had to reduce their volume. The problem wasn't with the the decoder: it was the speaker and the speaker enclosure, which in this case is the tender body. Setting the speaker in the frame with no sealed resonance chamber behind the speaker produces a shrill sound, without much bass. I can't see any way to easily fix this without sacrificing something else (weight/power pickup), so I'll write it off as the price paid for having the sound decoder and speaker mounted within the locomotive and tender instead of within a sound car.

It was fun adding .wav files found on the web to make this a "Japanese-esque" sound set. I added a fragment of a Japanese station announcement, a crossing warning sound, and a couple of US train airhorns that sounded like some I'd heard on a YouTube video of Japanese trains.

IMO, having thin sound is better than no sound. A steam locomotive running silently is missing a vital part of its personality; you can get away with that with an electric locomotive (although even those really do seem to need a horn). I guess it all gets down to what you enjoy about model railroading. I get most of my enjoyment from tinkering, not from modeling railroad operations. When I run my trains, my reward is the audio-visual experience.

Performance: It's a smooth-running locomotive, and (usually) works without derailing on a small radius R249 curve.

I had some problems with the leading truck hopping the rail at the entry to some curves. My guess was that the truck, being a very light part, wasn't exerting enough downward pressure against the track. My quickie solution was to grab the wheels and pull downward, bending the plastic beam that mounted to the locomotive: It worked, but only for a while. I then stretched the spring that pressed downward on the leading truck, but that upset the weight distribution, making the drive and traction wheels slip (My guess: It shifted the weight to the trailing truck, after the traction wheels). What finally worked for me was removing the spring and adding a thin layer of Bullfrog Snot to the first pair of drive wheels (My guess: It shifted the weight forward to the first pair of drive wheels and to the leading truck). Fortunately, the electrical pickup from all the other wheels was good enough to cover the loss of a pair of wheel contacts. Solving problems like this isn't easy because it could also be the wheel spacing, or a track problem, or a combination of all. Those are usually the first things to check.

Because of my small radius curves, I'd avoided locomotives with more than 6 drive wheels, but heck-- I wanted to believe. (The D51 has 8 drive wheels, two with traction tires.) That's not to say that it doesn't struggle a bit. With load control/back EMF turned off on the decoder (I prefer to leave it off), it slows down noticeably on a small radius R249 curve pulling no load. It can pull the 8-car Orient Express set through the curve without stopping, but they're an exceptionally smooth-rolling set. However, it has problems in tight curves pulling a consist with more resistance in their trucks (like axle-tapped power pickup trucks).

With load control/back EMF turned on, the motor draws more power so the locomotive doesn't slow down at the R249 curves.

KATO 10-1230


11/03/15- The old Minitrix Orient Express set steered me towards this beautiful set of coaches, before even knowing about the D51 locomotive. By then, I'd been thoroughly impressed by the quality of Kato's stuff, and especially by the reliable power pickup from their trucks. I certainly didn't want to miss out on their production of a set of European-style coaches.

This set doesn't model the original Orient Express, but commemorates a special nostalgia set of coaches from 1988. Despite the color difference, these bear a close kinship to the creme and burgandy CIWL coaches in the Minitrix set.

Although I was perfectly happy running them with my German locomotives, I really wanted to sample Kato's steam locomotives, and the D51 was one of the two that hauled the Orient Express in Japan during its 1988 worldwide Nostalgia tour.

Kato released an add-on set of eight additional coaches. I can't run a train that long on my layout (it would look silly), so I passed and saved the cost of lighting them.

I don't have much to say about them that isn't apparent to the eye, except that the 2nd coach has red-lighted table lamps! Frankly, with the red (understated in the photo below), it's a little over-the-top and doesn't fit in very well with the other cars that I lighted with warm white LEDs. Nevertheless, it's an interesting feature.

The other thing that I thought was noteworthy was that the cars are fairly close-coupled... using Rapido couplers! While they may be oversized, the Rapidos work well and are a heckuva lot easier to couple and decouple than Kato's Scharfenberg couplers. I don't have any desire to change them.


Video Clip: The lurching at start and stop began once I turned on load control/back EMF. I turned it on to help when hauling the 8 Orient Express coaches around curves-- before turning it on, the locomotive would struggle/slow down at curves.

The load control feature is fairly complicated, with many parameters to adjust. I didn't (and still don't) totally understand them and didn't have the patience to tweak them. After uploading the video and seeing how bad it looked, I consulted the decoder manual: Decreasing CV52 and increasing CV51 eliminated the lurching and smoothed out the slow speed performance. (It took trial-and-error to get there.)


12/28/15- You may be wondering, why make a sound car when the D51 steamer has a sound decoder? The answer is: Consisting.

From what I've read, the 1988 Orient Express often ran in Japan with an additional locomotive, the electric EF58-61 (aka, the Imperial locomotive). I'd made a sound car for the Showa-era Imperial train with the EF58-61. Since the locomotive could do double duty with this Orient Express set, it seemed like a good idea give it a sound car as well.

This concept widened once I found other diesel and electric Imperial locomotives. Instead of giving them their own set of coaches, I decided to "bundle" them all together in a system of two sets of coaches via Advanced Consisting. Both sound car decoders would need diesel and electric drive sounds. (I tried to add a third steam drive sound, but failed.)

I used the baggage coach as the sound car for the Orient Express because it's the first coach in the set, so it would be in close proximity to the locomotives. It has fewer windows than most coaches in the set, which is desirable when the innards will be filled with stuff that shouldn't be seen from the outside. The main downside is that it's the shortest coach in the set.

As a rule of thumb, shorter cars mean less reliable electrical power pickup because the trucks don't sample as large an area of track. While the design of Kato power pickup is more reliable than most, it's not always perfect, and nothing shows that like a sound car. In fact, that's what I discovered. There weren't many sound glitches, but all it takes is one to get your attention.

You can try to fix the causes (usually, the track), but that's a huge target: The track has so many minute variations that can cause problems for some trains but not others, and it's always changing due to the ongoing accumulation of grunge. (Track grunge and wheel grunge are inevitable, so periodic cleaning goes without saying.)

Flywheels: Otherwise, it's the trains. Many locomotives have flywheels, which use inertia and momentum to re-establish good electrical contact with the track. That's effective only when trains are traveling fast enough to coast through short sections of track with poor/no electrical contact (like turnouts). It doesn't do anything for the power loss that causes sound glitches while the train is coasting.

Shared Power: Improving the power pickup through the trucks is another solution. The more wheels picking up power and the wider their pickup area, the better. This will get the train through sections of track with poor/no electrical contact if it's getting power from both rails anywhere along its pickup area. To increase the pickup area, cars can be electrically coupled with plugs/sockets or special couplers (rare-- Tomix uses them in some of their Shinkansens) to share the duty of collecting power for sharing with the whole train. This eliminates power glitches because the train never loses power; it's always picking up power from somewhere (unlike the motor flywheel). The downside is that everything that shares power has to be fitted with either ugly plugs/sockets or special couplers.

Capacitors: Another solution is the electrical equivalent of a flywheel: The capacitor. The capacitor is like a short-term rechargeable battery; it charges when it receives DC power, and discharges when power drops. Depending on its capacity, that may be enough to keep a decoder and/or motor running long enough to get through the bad section of track... hence the name, "keep alive circuit" for a DIY or a commercial product based on the capacitor.

The ESU LokPilot and LokSound decoder manuals have diagrams showing how to add this feature using off-the-shelf components or their Power Pack product. The DIY solution requires a 2200uF 25v capacitor, a rectifier diode, and a 100-ohm resistor. (The diode and resistor are to handle inrush current, and aren't directly involved with the keep-alive function.) These are soldered to two pads on the decoder. (It's a little hard to see where to solder the LokSound ground lead in their diagram because of the light contrast, but it's the pre-tinned pad at the edge of the backside).

I'd tried this before but couldn't get it to work and gave up: It's all about the capacitors. A 2200uF 25v electrolytic capacitor is huge, especially for an n-scale train. Laid on its side, it looks like it might fit in an n-scale coach, but you might not be able to get the body shell back on. That's what I learned.

When I first tried this, I had some compact, high-capacity Gold Caps, so I thought I'd be clever and use them instead. The problem was that they're rated for about 5 volts: You can't use a capacitor rated for 5 volts in a circuit that runs at 12+ volts. It's likely to pop, or worse, catch on fire. (Also, don't use them directly with DCC track power, which is alternating current.) The 25 volt-rated capacitor that ESU recommends has a safety margin, which is wise since DCC system voltages generally range from 12 to 16 volts. So how do you get around this with lower voltage rated capacitors? Wire the capacitors in series.

The voltage rating of capacitors in series is additive. Therefore, to increase the voltage rating, you connect them in series. So to get the Gold Caps up to the 25 volt rating, I'd need five connected in series. Except... in series, capacitance decreases, just like resistance does for resistors in parallel. So two 1000uF capacitors in series acts like a single 500uF capacitor. At the time, I didn't realize this and wondered why the circuit didn't work. I'd created a large roll of high uF capacitors rated for 25v that had very little capacitance! Stupid me.

Capacitors in parallel act differently: The capacitance is additive, whereas the voltage rating is that of the lowest-rated capacitor. Therefore, you can use two smaller 1100uF 25v capacitors connected in parallel instead of the single large 2200uF 25v capacitor.

No free lunches: While the total volume of two 1100uF aluminum electrolytic capacitors may be more than a single 2200uF capacitor, you may be able to spread them out in the car and fit the body shell back on. That's definitely a good thing, even if you have to cover a few more windows to hide them.

Expensive lunches: There are more energy-dense capacitors available, like SMD Tantalum capacitors... but they're considerably more expensive than aluminum electrolytic capacitors. I've seen prices in excess of $20 apiece and wondered what special mojo they contained. I suspect that they're manufactured to very tight tolerances for very demanding applications. I wouldn't put model railroading in that category.

Fortunately, the factories in China produce much cheaper versions that probably have inferior specifications, but are affordable and can be bought on eBay in quantities of less than 5000. (The low price does make me wonder what you're giving up, according to the law of "no free lunches".) Steamlined Backshop also has them for a higher price, but with a much swifter delivery in the USA.

CAUTION: Tantalum capacitors can burst into flames and pop! That's not just hearsay, I witnessed it in my second attempt to make a Keep-Alive circuit: 10 bargain-priced eBay Tantalum capacitors from China in parallel, correct polarity and wiring double-checked. Tested on a programming track. Within a second or so of turning on power, one of the 10 blew up: A sharp pop sound, with a long flame shooting out of the side; caught on fire. Although startled, I put out the flames quickly.

Collateral Damage: Charred, melted plastic compartment walls and floor board in an area about 1.5 cm in diameter. The debris on the right side is melted into the plastic, on the other side of the two adjacent Tantalum capacitors that didn't explode. These things get seriously hot when they fail! Fortunately, I didn't have the body shell on. The decoder was on the other side of the melted plastic wall and worked afterwards. The body shell fit back on, and the charred interior area isn't visible behind the frosted windows. Just lucky, I suppose.

Causes: I can only speculate. Defective/marginal capacitor? Too much heat during soldering? 16-volt rated capacitor, without a sufficient safety margin for DCC? Tantalum capacitors have very low tolerance for voltages exceeding their rating, which quickly starts them down the path to thermal runaway.

Suggestions: If possible, use aluminum electrolytic capacitors instead; use capacitors with a generous voltage safety margin; measure individual capacitors with VOM for correct polarity markings and proper capacitor behavior; solder quickly without dwelling too long; stress test assembled circuit outside of train frame with adjustable DC power supply before installation; cross your fingers! Or buy a commercially-produced Keep Alive circuit and trust that it's safe.

Maybe there's a good reason why the Tantalum capacitors that Digikey chooses to sell are so expensive?

Getting back on track... The Orient Express Keep-Alive Circuit: Sharing track power wasn't a practical solution for this sound car's minor glitching problem since all the locomotives that ran with it would have to be fitted with power plugs or sockets. This motivated me to explore the Keep-Alive circuit again.

This time, I used Tantalum capacitors because of their compact size, and relaxed the specifications a bit. Streamlined Backshop uses (and sells) 220uF 16v Tantalum capacitors in their passenger light kits. The Digitrax Zephyr Command Station that I use is a 13.8 volt DCC system, so those capacitors are within the rated margin-- granted, not by much. 220uF Tantalum capacitors with a higher voltage rating are hard (if not impossible) to find.

I'd need 10 of the 220uF capacitors in parallel to get the 2200uF that ESU recommends. What if I used 6 for a total of 1320uF? I didn't need the train to run for 2 seconds without power, and since the sound car didn't have a motor, I expected that the current drain would be lower. I tested it, and it worked: The sound continued for a few milliseconds after killing the power, and the LEDs (for cabin lighting) gradually dimmed-- more than sufficient to deal with minor power glitches. Cool!

Six Tantalum capacitors fitted in the space grinded out (in vain) for a 2200uF 25v Electrolytic capacitor that was sooooo close to fitting.

Bear in mind that the energy-dense Tantalum capacitors aren't a magical panacea (They can also be risky-- see "Caution" above). A block of 6 or 10 still takes up a considerable amount of space. However, their small size and rectangular block shape gives you the flexibility to place them inconspicuously, wherever you can find the space.