The connecting cable

The connecting cable has several issues. Applying the pins to flexible 6 conductor Telco style wire results in a cable that has a high failure rate. The stranded wire is simply so fragile that a small nick while stripping caused some of the strands to break off making the connection more fragile. I wound up using a fairly expensive stripping tool to avoid nicking the conductors. This fixed that problem.

The socket pins are tiny and difficult to apply to the wire with my large hands, I should probably find a 6 to 10 year old to work with them - on second thought maybe a 14 year old would be better because they could do the crimping as well. I initially was soldering them on but that makes the wire more likely to fracture when installing it in the field and one bad wire makes the board not function properly. So I bought a special crimping tool to apply the pins to the cable. The small gauge wire turns out not to have the strength to resist the handling forces when the cable is tugged on. I found that folding the conductors over the sheath and then crimping improved that situation but I still have too many fails making the cables.

I am still trying different wire to improve the durability of the cable. My latest attempt is a wire of a larger gauge but the flexibility is now somewhat of an issue. The wire has to be long enough to be able to come out of the head a few inches but needs to be short enough to not have a lot of surplus that needs to be in the base box.

The existing LB (head box) has to be modified

The existing LB (head box) has to be modified so the connecting cable can comfortably connect to the board. The LBs are from different manufacturers and are in two basic styles one is made with rounded ends and the other is square ended. I decided to remove the square ended boxes and install rounded boxes at all locations as we go along replacing the old signals when they fail. Progress has been slowed because Richard likes to repair the old boards and not update them,

First the box has to be restored to its original depth so I use the radial arm saw in the shop to slice off the extension that was glued onto the box before.

Next the feed hole needs to be reamed at an angle using a Unibit to make it easier to connect the board and slide it into the head for use. I had moved the location of the connector on the circuit board in the final design to the rear side of the bottom of the board to minimize this need but it turns out that it is still necessary.

Further changes

Since we started using the boards, I decided to remove the transistor socket pins and the LED socket pins. The reliability of the boards have made the need to be able to unplug these components unnecessary. I left the IC socketed because it turns out that it is the only component to fail from time to time so it can be field replaced easily.

Part of the reason to remove them is that the pins were difficult for me to solder onto the board. I did decide to socket the MOVs onto the board as they can be easily replaced in the field when they have been damaged by lightning from storms. I also got the club to get grounding rods and they were installed around the property so we could ground the track common rail. This made a major change in the failure rate of the signals. We also tried MOVs in the signal base boxes, at Tommy's suggestion, to short circuit any disturbances before they got to the signal head. This helps with the reliability as well but when they short they cause the signal head to do strange stuff. We decided to remove them because of the strange stuff but it turned out to be the lack of the ground that caused them to short out in the first place so we reversed ourselves and put them back in the base boxes.

Transistors

I just found a problem with the brand of transistors that have been shipped to me by the vendor. I have been using PN2222A transistors that work interchangeably in the circuit. Then my vendor apparently ran out of the transistors by the manufacturer that made the transistors I had been using. Fairchild is a large company that makes a lot of different semiconductors including the PN2222A. The latest supply of transistors is made by On Semi. This company makes a transistor that is different in construction such that they "leak" in our circuit. Most people use the transistor in a linear circuit that means that they use them in an analogue mode and they can "leak" when they are biased off but still amplify a signal OK. We have a need for the transistor to switch "off" and not pass current when biased off. I have to test a sample of the batch to find out that it is apparently a flaw in design of the transistor by On Semi.

Of course, the vendor does not agree that the transistors are defective. They argue that the transistors amplify just fine and it must be a design flaw on our part. They offer to accept the transistors back with my paying the shipping and also a 20% restocking fee. I ask if they have the transistor made by Fairchild and they say yes, send those back and they will credit my account and send the Fairchild transistors in replacement minus the shipping and restocking charges. Since the circuit will not work with the On Semi parts, I mail the transistors back to them and continue to argue they are defective. They tell me the Fairchild are twice the price and agree to make the switch.

Finally a package from the supplier and I anxiously open the package and find they have shipped me another 300 of the On Semi transistors and charged me for the Fairchild! I test a sample of the On Semi and find them to be just as "leaky" and call the company to straighten this out.

They could not supply the Fairchild parts so went to another major parts supplier and found that if I ordered a reel of the transistors I could get the whole batch of the transistors for less than I was paying for 300 of them from the other guys. They are perfect and I am delighted to have them although I have thousands of them now.

Final Board Design

Now that we have a prototype board that works and we can have in operation daily, it is time to correct the problems and go for a production board. Again I correct the board design in PCB 123 and transfer it to another CAD for the final design as we can save perhaps 30% by having somone else make it. This time I add features to the design so the board can be used as an LED driver board for repeaters and dwarf signals if the board is cut at particular locations and it will fit into those LBs. I decide to specify a high-reliability IC socket to provide more reliable contact under all conditions. I look into transistor sockets and find them sorely lacking so no-go for the redesign. I found sockets for the LEDs and ordered them for the initial run of boards. After several months I finally get the board ready to place the order and Richard gives the go-ahead. The boards arrived and after extensive examination it appears that I only missed a single connection. I devise a patch for the problem and start building board one!

The board assembles well and board two is about an hour behind the first board. Now I have boards to actually use in final assemblies. I tell Tommy and Rich that they are working and try to see if I can do anything better in the manufacture and assembly of the PCBs.

As I get the parts in for the initial run, I noticed the LED sockets (pins) actually will fit the transistors so now I can socket them as well so all the active parts can be easily replaced. I need to get some LEDs from Tommy for the boards but since they are socketed I can use the three I have to test out the boards and simply move them from board to board for testing as I make them.

PCB123

Why do I keep using the PCB123 software? It is simply top-drawer for a freeware to develop the circuit boards. It works just as needed and is able to alert you if you have traces that cross and should not cross, traces too close to other things as well as unconnected components. It has design rules that make the boards high quality and PCB123's manufacturing is top-quality. They turn out to cost about 50% more than their competitors for low volume production runs. I choose to author in this software and copy my layouts to the other CAD software for obtaining lower cost PCB products. This process does have its drawbacks in that the design may be perfect in PCB123 CAD but redrawing it may cause an inadvertent short as in my first prototype board.

Prototype Needed

After discussing it with Richard and Tommy, it was determined that we should produce a prototype board run to see if the layout and execution of the board was correct. The PCB123 company did not offer any cost effective prototype boards as the setup cost was a major part of the production cost. Setting up the etching masks and drilling templates was just as costly for one or two boards as for 100. So I went looking for other manufacturers of PCBs. I found several that offered their own CAD software as well as the low cost prototyping service (less than $100). I could not use the CAD design from the software that PCB123 had so I had to re-draw it in other CAD programs then error check it and submit it for price quotes. It turned out that it cost almost the same for 10 boards as for one from the final prototype PCB manufacturer!

Now time to order the materials to make the boards. I worked with three companies purchasing surplus electronic parts to make some prototype boards. The only drawback is that you might not be able to get the same parts for the later production run as surplus parts are just whatever they can get.

When the boards arrived I set off on assembling the 10 boards and after I finished the first one I plugged it into my simulator and fired it up to find that it did not work. Now comes the hard part, troubleshooting the circuitry and finding what was wrong. Since it was shorted, it was rather easy to find the defect in the printed circuit board. It turned out to be a connection on the original design that when I transferred it to the CAD software of this company, I shorted two foil traces together! Because the foils are on two sides of the board and the flaw was under some labeling on the board I failed to detect the error before making the boards. With some help from Tommy Cebulla we developed a fix for the short by cutting the traces and making a jump over the bad spot with a component lead. The other problem occurred to be that the red LED was on the wrong end of the board. If you orient the board with the connector down the red LED was at the top so I had to fix that as well. This was done by crisscrossing the supply resistors for the green and red LEDs and then installing the LEDs on the opposite ends. The prototype board is shown in the photo at the top of the page. We debated on how to connect the LEDs to the board and this is an attempt to make the connection flexible. Later versions use a rigid connection between the LEDs and the board. Now it is time to install some boards into service at the SCRR.

New box needed

Now we needed a way to power and cycle the pcb on the bench so that we can service it and check it while building the boards. I designed a box that plugs into a wall outlet (or 12V battery) that causes the board to fully function. It has three switches and an indicator. The first switch turns on the box and this will test the default state of a green block. The next switch gives a yellow signal and the next a red that also turns on the indicator if it is sending a yellow to the earlier block correctly.

So I built three of these test boxes.

The new board

With these ideas in mind I started over again. I measured the LBs and decided that the new board should be able to mount in the box without tilting and if it could fit straight in - all the better. It should interface with the existing LED board directly so the whole assembly could be replaced as a unit, if necessary, so it should plug in to a cable that remains with the semaphore LB. The LB measurements showed that the board had to be about 35-40% of the size of the existing board. First I lined up the resistors and divided them by input and output primarily into two groups and put the IC (integrated circuit) between them. The output transistors went up the long side of the board to drive the LEDs. Then I realized that the LED circuit board was not necessary if the LEDs mounted to the edge of the new board so I added the LED board circuitry to the Logic Control Board and voila! Now I had a design that fit the LB straight in and could be used as a module to make servicing easier. When mounted to the LB cover, it can be plugged into the semaphore connector and then the LB cover screws hold it all in place. It makes it trivial to replace a malfunctioning signal. What would take us a hour or so normally can be done in minutes now! The module can be repaired in the shop instead of the field so it speeds up repairs.

Board size

Because the existing board is so large, Tommy has been cutting the LBs and gluing them so they are deeper. This was a goal to end this process. The new board was made to fit either side to side in the LB or front to back. At first I wanted to make it side to side with the LEDs down the middle of the board such as the existing LED mounting board that I made a few years ago. After trying a number of different board layouts I realized it was not practical and they had to connect to an edge of the board. The original designs had the LED board plugging into the Signal Control Logic Board via a cable. The cost of each of the boards is bout the same for the LED board and the Signal Control board so by combining them the cost drops roughly in half (of the boards).

Modularity

Working with Tommy Cebulla on the switch machines I realized the value of his use of modularity. When he built the switch machines he made them in four parts - the enclosure, the electrical control module, the mechanical motor module, and the input/output wiring module. These assemblies can be removed and replaced with out having to replace the other parts.

Starting the Project

In 2006 I started looking at the circuit that we use to control the signals at the SCRR. I first drew the schematic of the circuit and then I got a CAD program from a circuit board company (PCB123) that allowed a person to draw a circuit board on a PC. I had a steep learning curve as I had only made circuit boards by the old manual tape and photo layout methods. The idea was to make a printed circuit board (PCB) of the wiring to make it more reliable and to add features to the existing circuit. The handmade perforated phenolic boards used by my predecessors had some inherent problems. One flaw was that the wiring was made in 3D as it had to crossover other connections and this created shorts from time to time. Another problem was the cold solder joints that occasionally cropped up. Broken wires were another plague. The usual failure was the digital comparator chip so early on this was socketed. The transistors and the LEDs were not. The LEDs were glued to the LB cover in holes and the wiring was tacked to the leads of the LEDs then the other end was soldered to the main logic board outputs.

My first attempt at designing a replacement was to duplicate the wiring on the PCB and make it sightly smaller so it would fit easier into the LB (Light Box) that we use for the semaphore heads. I wound up drawing about 34 versions of the board as I adjusted measurements and locations of various components and got newer versions of the CAD software. I added components and removed them as time went by and Richard Krueger and I talked about the needed components and features.

Then I took a break from making the PCB for the signal control circuit and made a PCB to modularize the LEDs and standardize the spacing of the LEDs in our signals. This was a quick job and it made assembling the signal heads easier and more reliable.

Not much later I lost my house and moved to Tennessee. I looked at the CAD stuff from time to time but just did not have much interest in it anymore. Then it became necessary to move back to Wisconsin so we came back to Hudson and I reconnected with the SCRR.

After the one year hiatus, I got back on the Signal Crew and I mentioned that I still had the CAD software and the circuit board layouts. Rich suggested I work on it again and gave me his suggestions for changes and improvements. I realized that the circuitry was so spread out that it made the board so large that we had trouble fitting it into the LB. I had worked in medical electronics for about 12 years and repaired and adjusted digital and analog circuitry that was assembled on circuit boards that were very closely packed as real estate was at a premium in manufactured products. I looked at the parameters of the circuitry and realized that we could reduce the size of some of the components from what was being used presently. I then did a circuit analysis and worked out what each part of the circuit was doing and I have to admit that I admire the designer (Erik Tromberg) of the original circuit. The circuit is simple and yet well suited to our needs.