Improving the LS Models
VTU coaches (6)

Original page created on 22/12/2022; last modified on 10/05/2023.

Lighting, episode 4: LED filaments

By analogy with the now banned incandescent lamps, these components try to mimic tungsten filaments. They consist of a single, thin and rigid — and fragile — substrate on which numerous light-emitting chips are implanted.

LED filaments

The selected models have the following characteristics.

Reference GWT3LWF2.DM (this is an Osram reference no longer manufactured. Apparently the Chinese have a large stock of them).

The anode (+) is marked by a hole on the electrode.

There are 18 LEDs per strip, connected in series, hence the voltage of 47 V.

There are filaments with different sizes and voltages (up to 120 V).

Here is an example with a 55 V supply:

Testing a filament

The following values were measured:

To light the room of a VTU, which is 220 mm long, I need six filaments per side. The disadvantage is immediately apparent: between two successive filaments there will be a non-luminous area of at least 12 mm, which can fall in the middle of a window.

Circuit for VTU

The circuit being placed in a VTU, the result is disappointing, as I expected, because the non-luminous spaces are very visible. We are definitely not there yet.

Episode 5: COB LED flexible strips

What are COB LEDs? Well, they are LEDs whose “chips” are implanted on a common support, hence the name Chip-on-board. See the Wikipedia article. By the way, the filaments seen above belong to the same technology.

The nOOds

I first found the nOOds. This strange name comes from the word noodles. Indeed, they are cylindrical LED strips, so flexible that they look like overcooked spaghetti! I quote:

They’re made of dozens of micro LED diodes that are bonded together on an ultra flexible metal backing, then coated in colorful silicone for protection. Since the LEDs are in parallel, you only need 3 V to light ’em up - we recommend current limiting with a resistor to let max 50 mA through.

Depending on the source, they are available in lengths of 80, 130, 260 or 300 mm in cool or warm white (unfortunately the colour temperature is not specified). The diameter is not specified, I estimate 3 or 4 mm. The strip cannot be cut. This is the main disadvantage (in addition to the price), as no length is really suitable to match the 220 mm I need. Other disadvantage: the LEDs are wired in parallel, requiring a voltage of around 3 V, with a relatively high current. This goes against my “principle”: high voltage, low current.

Here is an alternative source for these strips. Lengths of 80, 130 and 300 mm.

Other COB LED strips

Fortunately, there are other LED strips that are similar in construction to LED filaments, but available in much longer lengths. This is exactly what I need for my VTU lighting.

This type of strip is self-adhesive and is available in narrow widths (5 or 8 mm); it is powered by 24 V (also available in 12 V), which is fine with me, since the circuit will be powered by 60 V. Three colour temperatures are available: 3000, 4000 and 6000 K. The lengths offered range from 0.5 to 5 m. I chose a 5 mm wide, 5 m long strip, colour temperature (CT) 4000 K, colour rendering index (CRI) 90 (good!).

Presentation of the COB LED coil

The strip is covered with yellow-tinted silicone. It can be cut into 42 mm elements (41.67 mm precisely). The elements are wired in parallel with each other. The total length ranges from 0.5 to 5 m. The tape is made up of 0.5 m segments soldered together.

For the VTUs I need two 210 mm strips (5 mm will be missing on each side, I’ll come back to that). The calculation is that a 5 m roll will fit 11 coaches, which is less than €1.50 per coach.

Here is how the strip looks.

Appearance of the COB LED strip

The separation between elements is indicated by a line so thin that it is almost invisible! Here I have increased the visibility with a cutter.

Aspect of the connections. They are located under the tape, not tinned, and covered with adhesive, which can be annoying for soldering:

View of the connection pads

From this point of view, the 8 mm wide strips have the only difference of having their connections on the top side, which can be more practical.

Let’s turn it on… I have set a very low current so that you can see each “chip”. There are sixteen LEDs per 42 mm element.

COB LED strip lit

Test on a VTU

For this test, I used an old obsolete circuit on which I stuck two strips of five elements, the two strips being wired in series * and supplied with 55 V by a voltage quadrupler connected to my LokProgrammer (with my Lenz control unit, it reaches 60 V).

* In the end, the connection in series is not a good idea in terms of holding time in the event of a power cut: the connection in parallel will double the current, which will nevertheless remain very low, but will allow a deeper — and therefore longer — discharge of the anti-flashing capacitors. Indeed, the LED strips will practically stop conducting below 48 V in the first case, but below 24 V in the second.

COB LED strips mounted on a circuit

For my personal taste, the ideal lighting current is 0,5 mA per branch. It’s difficult to render the actual appearance with a photo.

As mentioned above, the fact that the strips are a bit too short to cover the whole room of the coach means that the most extreme diffusers are not illuminated. But for the others, the rendering is what I had hoped for a long time.

Some modifications to be made on the coaches

The light guides, which had already been milled in episode 1, need to be levelled and thinned down to about 1 mm high, so that the LEDs are as close as possible to the diffusers. This is easily done with a bastard file. Milling is not recommended, as even at slow speeds the plastic will melt and clog the milling cutter.

Levelled light guide

Note: the light diffusion is significantly improved by roughening the diffusers (ends of the light guides) with 500 or higher grit sandpaper, which has not yet been done in the photo.

In addition, the 0.5 mm thick Evergreen louvres already placed along the light strips to prevent light escaping to the sides have a modified height, due to the smaller gap between the light guides and the printed circuit: 3 instead of 4 mm. One of these louvres is visible on the far side of the coach.

New printed circuit board

With the test assembly validated, I drew a new circuit board. Once the circuits were received, I started to assemble a first board. As expected, soldering the LED strips was acrobatic, since the pads are underneath. I tried using hot air, but I couldn’t melt the soldering paste, even at 300° C. I gave up for fear of burning the silicone cover. Fortunately, there are only two solder pads per strip, as the segments are originally connected in parallel.

You will notice that the circuit is separated into three parts, connected by straps. The aim is to put the central part as low as possible so that the maximum amount of light passes through the light guides. In fact, the central part is 1 mm lower than the ends. I’m not sure if it’s worth the complication…

View of the powered PCB next to the coach

Circuit mounted, checking operation.

PCB mounted on the coach

The IC is held in place by 0.3 mm thick nickel silver strips slid under the arches of the body. This ensures that the LED strips are well pressed on the light guides.

Zoom on the circuit lock in the body

The ⌀ 0.6 mm telephone wire connections (straps) are not very pretty: they will be replaced by flexible wires.

View on the latching reed switch central control, for those who are not yet convinced of the simplicity of the thing…

Zoom on the latching reed switch control

Two other reed switch places are provided at the ends, for possible lighting of lanterns. This was not done on this first coach.

Here is the result once the roof is back in place. A heating jumper wire got lost. The truth is that I manipulated this guinea-pig coach a lot…

General view of the lighted coach

Close-up of the lit coach

I still have to equip my ten other coaches…

Special case: the B5rtu bar coach

With LED strips, the modification of the bar arches (see page 4) is no longer suitable. Even when reduced in width, they prevent the correct positioning of the PCB.

So, these arches are completely milled out. But beware: they contributed to the correct spacing of the walls, which must not sag when the roof is clipped on. Therefore, reinforcing crosspieces must be added. These consist of 0.5 mm thick polystyrene strips of exactly 20.8 mm in length, which are glued to the remaining stumps of the arches (Faller Expert glue).

Here you can see two of the four crosspieces being glued.

Gluing crosspieces

As these crosspieces are not very rigid, they are themselves reinforced by a plate of the same material, large enough to cover the whole bar, i.e. 115 × 15 mm, glued on top.

Gluing the reinforcement plate

On the bar side, the PCB is held to the plate with thick double-sided tape.

Here are two views of the illuminated bar.

View of the illuminated bar

View of the illuminated bar

Question: Is the best the enemy of good?

We have just seen that the extreme diffusers are not illuminated, because the LED strip is only 210 mm long for a room length of 220 mm. There is a way to avoid this: it consists of separating the strip into 42 mm elements, each centred on a group of two windows, which leaves an unlit space of 3.4 mm between each one, but which is behind a trumeau, making it invisible.

The disadvantage, considering the difficulty of soldering the strips, is that five strips per side have to be soldered here instead of one, plus they have to be cut, the adhesive must be removed from the connecting pads which have to be tinned. This makes the work much more time-consuming.

It also requires the design of a new circuit, but that’s not the hardest part. By tinkering with an existing circuit, I equipped it this way.

New circuit with separate elements

Here is the result.

General view of the coach with separate elements

Close-up of the coach with separate elements

This time, all diffusers are illuminated…

So, yes, here, the best is the enemy of good. I don’t think I’m going to make this new circuit, considering how little improvement I got.

GW T3LWF2.DM-MEMU-4950S5-C1F1-30-T01-XX
LED filament 47 V 5000 K
Price 2022: €14,05 per 100
at aliexpress.

Adafruit nOOds overview
on learn.adafruit.com.

High density flexible LED FCOB strip
Price 2022: €15,75 per 5 m
at aliexpress.

LED Edison bulb lamp (sic)
Price 2022 : €4,47 each in 300 mm
at aliexpress.