2017 part 11 – Designing Next Level, Heljan MX, Current Sensors

December 3

It is about time to get all track on my layout done. I am still missing the local line SVJ – Skive Vestsalling Jernbane – as well as the front most visible piece of track at the lower level.

It is all the way through: “Timber”-work, current sensors, S88N module, servos in the new turnouts as well as leveling, test driving etc. etc.

I am attaching the front track using steel brackets. And I am using small pieces of maybe 10mm wood between the brackets and the 6mm thin plywood, so that I have something to attach screws from both sides into. Parade

SVJ is a bit more tricky to construct, since it has to “hover” 10 cm above the existing track. I.e. slopes must be calculated and implemented. It is also tricky, since landscaping and housing shall be constructed on top of it in detachable modules. I have made a few construction sketches:SVJKonstruktion_SVJ_2Konstruktion_SVJ_1Konstruktion_SVJ_4Konstruktion_SVJ_3Now it is only a matter of free time for me to get it done. It is probably not going to happen this side of Christmas, since I seem to only be able to set 10 minutes at a time aside for model trains. And that is simply not sufficient. I need at least ½ days where I can concentrate on it.

In the meantime, components for new current sensors have arrived from China. It has taken up to 59 days – probably in the Danish customs for 56 days.

In addition, I have recorded a small video with my new SQ8 mini camera (see separate page), where I attached the camera to a locomotive with a good lump of sticky tack and then just drove around the track. It is not the best video resolution in the world and colors are also far from natural. But it gives an impression on how my layout looks from the perspective of a locomotive. That is kind of funny. And it only set me back 8,55 Euro.

Another “little” thing, I have been messing with is another Heljan purchase. My excuse is that I “needed” an MX locomotive, and only Heljans can drive in the narrow R1 curves. My  previous experiences with Heljan IC3 are not exactly positive. But RC-kongen made a good offer on a Heljan MX. About 30% discount.

I had my doubts, so I investigated a little beforehand. The locomotive has no traction tires, and I know that some owners have exchanged axles and wheels with those from a Hobbytrade ME. The problem is however that those axles are no longer available. And more than one dealer convinced me that it is not a problem. But it is! Read on.

I bought the locomotive. And it was a swift delivery. To my great joy, the windshield wipers were already mounted. I have been reading about people having great difficulties mounting those themselves. And since the locomotive is of the newest generation, it is equipped with an ESU LokSound decoder as well as short coupling possibility. So far so good. The sound is good, even though rumors has it that it is from an MY and not from an MX. That might take the fun away from some, but I don’t care, since I don’t know the difference anyway.

My layout is – as mentioned a couple of times – from around 1987. That means that an MX must be equipped with snow-plows. So it was “just” a matter of mounting those. In the “manual”, it is stated as a single sentence that they shall be mounted at the back of the buffers. It is not stated that it is close to impossible to do so. But it is. However, I succeeded by using violence, a screwdriver, sanding paper and dish washer fluid. I used the sand paper to round off the edges of the back of the buffers. The dish washer fluid was used to lubricate the back of the buffers. Violence was then applied to force the buffers through the holes in the snowplows while also using the screwdriver to push on the back of the plows.

And could the locomotive run? No. It short-circuited so that the ECOS switched off when the collector shoe should enter a turnout. I solved that problem by letting the locomotive stand with rubber band holding the collector shoe tightly up against the boogie. So now the collector shoe is correctly adjusted.

Next problem was de-railing of the locomotive. After some time, I found out that the bottom of the locomotive was not correctly seated, so the the boogie in one end could not move freely from side to side.

There are two remaining problems (see the continuation February 28, 2018):

1. Every time the locomotive shall pass a turnout, it hangs a little “on top of” the turnout. It is worse on some turnouts than on others. This has the consequence that it cannot haul to wagons uphill, if it must pass a turnout uphill (which it has to on my layout). It is as if there is not room enough for the collector shoe. Bo from nettog.dk has told me that I can change the collector shoe to Märklin 7164. It should be a tiny bit more flat than the standard collector shoe (which looks like a Roco). Traction tires would supposedly be an even better solution. But as mentioned, they are not available.

2. It is impossible to make any kind of coupler work with the snowplow. Especially the Fleishmann Profi couplings. And not even the standard couplings that are delivered with the locomotive. I have to cut in either the plow or the coupling or both.

And RC-kongen has all the way been not helpful at all. He has not been answering e-mails after the purchase. He was able to answer questions before the purchase though. And latest he has been rejecting to pay for returning the locomotive for repairs. That was when I thought the de-railing problem was a thing that I could not fix myself. And that is even though the Danish e-mærket states that such a paid return is “as common as beef in a burger”. So a piece of good advice: Buy your Heljan products from your local dealer and do not try to save a few hundred DKK.

And then there is the red color of Heljan versus Mck products. I have been reading here that the Heljan color (RAL 3003) is the one DSB actually used, whereas the Mck color (RAL 3002) is supposed to be how the rolling stock looked like after a few years of rain, wind and sunshine:RAL3002vs3002The thing is that that they look different. But so did the actual trains:GM-litraAB

December 17

I will leave the rolling stock be for a while and will start the building process instead.

Current sensors first, however. I have been making another S88N module. And I have made a board with 16 current sensors of the new type. I have still not programmed the PICs and I have not interconnected the two modules:266E8B49-1348-4E09-9A02-52F050A8AFD5The new sensor type is more compact than the old one. And the green connectors instead of the home-made ones looks quite a bit nicer. Now I only hope that it works.

The grey wire on the picture is hopefully the explanation why many of the old sensors have stopped working:

My S88N-P module is supposed to do galvanic separation between the ECOS and the feedback system as well as being a power supply for the feedback system.

Anyway, I have so far been mixing ground connections between the two parts. This way there are several meters of ground wire between the S88N module and the current sensor. So who says that doesn’t make up a voltage difference big enough to hinder the input on the S88N module to become low enough? Hence the grey wire, that is real S88-ground. I am going to use this ground connection for the optocouplers in the current sensors. I should have thought of this from the beginning, since my Uhlenbrock LocoNet feedback module is equipped with such a (LocoNet-)ground connector.

But all this ground-talk is still only a theory. Next step will be to actually start testing and measuring.

December 18

So much for the ground-theory: That is not the problem. I have been measuring on both the old and the new sensor type.

The new type works great. Output from the optocoupler is 5V when no current is drawn and 0,2V when current is flowing.

But the old type never comes below 2,5V. And that voltage is undefined in the sense that it means neither 0 nor 1 for the PIC inputs.

I have spent a couple of hours to solve the mystery: The old optocouplers only detects current in one direction. The new optocouplers detects in both directions. This means that the low signal from the new type is a fairly decent DC on about 0,2V, whereas the old type output is a square switching between 0,2V and 4,8V. The DC mean that I can measure with my voltmeter is about 2,5V.

I could solve the problem by introducing the capacitor that I had on my original diagram for current sensors:

Processed with Snapseed.
Processed with Snapseed.

Haakon from litra.dk has helped me by finding an easier solution: I shall turn the IR-diode in the optocoupler the other way, i.e. switch legs 1 and 2 on the optocoupler. That will work because the DCC signal is not a fully balanced square. It is more of a series of pulses with varying spaces between them. I have tried doing so, which doesn’t bring the low voltahe down to 0,2V, but only to 0,4V. But that is sufficient.

The explanation must be seen along with a picture of how a DCC signal “looks”. This illustration from NMRAs DCC standard shows just that:DCC-signalThe horizontal line is 0V.

I.e. looking at the two wire comming out of the ECOS, they are alternately + and -. That also means that current is flowing alternately in both directions in 2 out of the 4 diodes in the diagram, and a voltage drop will arise over the two diodes that current is flowing through. This voltage drop across the 2 x 2 diodes will thereby be a replica of the DCC signal – just with only +/- a couple of volts as opposed to the maybe +/- 16V coming out of the ECOS.

The optocoupler is only activated with current flowing in one of the two directions. And in the DCC standard there are long periods “above 0”, i.e with current flowing through the diodes from right to left in the diagram and thereby with the optocoupler inactive. Unless that is I am turning the IR-diode in the optocoupler the other way around, in which case it is going to become active during those long periods.

Note that if there is no locomotive in the block, no current at all will flow and therefore the optocoupler will stay inactive no matter which way it is directed.

December 28

The old current sensors have been fixed and JMRI warrants works again, since JMRI now again knows which blocks are occupied and which are not.

I have moved the layout away from the walls and I have sawed the first pieces of wood, so that I can build the last ½ floor on the layout. Some of it has been attached using clamps:E433328A-9ECC-470C-AFB8-E096DAC6C6A1

December 29

The outer support is nearly finished. And the inner one is attached using clamps. It is a good question if I am going to be able to putting de-railed trains back on track on the lower level:

December 31

That was the end of 2017. I was at Silvan yesterday to buy 6mm plywood. But it was sold out. I have plenty to do anyway by constructing the skeleton that the plywood shall be mounted to. Today, I have been attaching Skive H the way it should be attached and I have removed the threaded rods, that were placed right where the new track is going:A0C6292C-7C92-48EA-BAC8-4B487DF06BE3 18B9145D-7B8C-43D8-ACC4-227683DFC87D

And here with the new track hovering in the air since I could not get hold of any plywood yesterday:CC551690-64B4-437B-9E16-1D1BADE133A1

Happy newyear.EE02B25A-BBC2-4E61-98A8-7598DA88C21E