santa fe cajon-ish 1947 in a Kansas basement

[GaryHinshaw]

Right, I think for your application, the in situ logic makes sense.  I mainly wanted to share my experience with detection reliability.

[draskouasshat]

Thank you again Gary.
Any advice on how many cars should have resistors installed? I'm new to this portion of dcc. I was thinking 1 per car but with helpers on the rear of almost all freights, id do a couple per cabeese and one on every passenger car.

Adam Draskovich

[jagged ben]

Since you're modeling 1947 I gather you will have cabooses on every train.  I would put four resistor wheelsets on the caboose. You might use a slightly smaller resistor like 5k.  As long as your trains aren't longer than the longest block that will likely  be sufficient. 

Caveat:  with your relay logic I don't know if you'll have issues if you have a short detection section in the middle of a train - i.e. a turnout - that doesn't get detected.  You might need a lot of wheelsets on cars if that is the case.  Another thing that's nice about using JMRI is the 'debounce' feature that prevents the detector from bouncing on and off when detection is shaky.   It allows fewer wheelsets to do the job reliably.   I don't know if the NCE product has any such features or how hard it would be to augment them. That might be an argument for using the Watchman instead.  I would do some proof of concept experimenting before investing in a lot of components.  The wire length issues will seem petty if the system had functionality issues.

I think the idea of running a signal system with prototypical period relay logic is pretty awesome, btw. 

[C855B]

We discussed this at length a while back... but 1 resistor wheelset per normal car, 2 on long cars (80-86') with one at each end, none needed on passenger cars with track-powered lighting. 4 on a caboose would be OK but I think two should be sufficient; I wouldn't go lower than 10K on resistor value.

Aside from the cost, fabrication time and installation of multiple resistor wheelsets, you don't want to go overboard. With most or all wheelsets making contact in the track circuit, current draw goes up enough to decrease running power capacity. A 50-car train with one 10K wheelset per car will be 200 ohms, consuming 75 milliamps. Depending on what else you have hanging on the DCC bus, two or three trains in the same booster district might amount to something.

[draskouasshat]

Now that some great info everyone! Thank you!

I'll actually be looking into a delayed pickup on the relays as we have what's called "loss of shunt" just for this purpose. We have a 10 second delay built into some of our track circuits so I'll be looking into building that into the circuit immediately i can program it into the detector. I believe i read that the rrcirkits one has something similar to this. .

Adam Draskovich

[C855B]

...We have a 10 second delay built into some of our track circuits ...

Don't forget compression. 10 seconds is a long time on a model railroad. The 2-3 seconds that a lot of circuits build in should be good enough.

[draskouasshat]

Very true. Luckily, with direction of traffic on cajon, placing the detector in the middle of the block should suffice versus having one on each end. This should also help the reliability of the detection due to looking half the length of the block each way instead of looking for resistance at the full length of the block. At least it should work better in theory. Plus i can put a center feed case at each center feed location and know exactly where my feeders are soldered on!

Adam Draskovich

[GaryHinshaw]

The Watchman decoder definition lets you select debounce times between 0.1 sec and 8 sec, in various increments.  The BD20 almost certainly has a similar capability.  My blocks are currently set to 1 sec, but I am probably going to increase that to 2 or 4 sec.  As jb notes, the main issue is in short blocks like turnouts where there may be only one axle in the block at a time.   You'll probably need to experiment with this when you design your loss of shunt logic.

Mike is right about having too many axles, but 75 mA is still less than a typical non-sound loco draws, so it's not too bad.  So far I'm reasonably happy with one axle per car, mainly so that short blocks are covered.

[jagged ben]

The Watchman decoder definition lets you select debounce times between 0.1 sec and 8 sec, in various increments.  The BD20 almost certainly has a similar capability.  ...

I've never held one in my hand, but it does not appear to me that the BD20 has an integrated processor or any programmable features.  Am I  wrong?

[Santa Fe Guy]

   Adam.
I ran 1 resistor in every car plus one in a caboose in a 12 car train hauled by 2 locos on my SFRSD. (since sold). Every car (about 160 ) on the RR had one resistor fitted.
We tried by fitting a resistor to only a third of the fleet at first however that was a failure as some trains finished with only a few so made detection impossible.
We experimented with a delay (using JMRI) for the turnouts and went back to zero as it made it difficult for the dispatcher. All mainline turnouts were blocked separately which made a huge improvement for the the dispatcher. I ran all Digitrax Duplex Radio equipment including detection using BDL168 and SE8C for the signalling controlled by Locnet and using a JMRI dispatch panel designed by one of my good friends. It all performed extremely well.
I was wrapped to finally get operating signaling and the dispatchers panel up and running on the RR. It made for great operating sessions especially if the dispatcher (me) was occasionally in a grumpy mood. LOL.
Rod.

[C855B]

I've never held one in my hand, but it does not appear to me that the BD20 has an integrated processor or any programmable features.  Am I  wrong?

You're right, the BD20 does not have any debounce facility, or at least they do not specify anything. Output appears to be raw detection. Adam will have to supply that if he uses the BD20. The MRCS cpOD can be configured from 1.25 to 5 seconds without reliance on an external processor.

[draskouasshat]

Ok so now that bd20 doesn't seem to have a delayed pickup, cab the rrcirkits detectors be installed separately or do they have to be binned in a group of 8? It looked to me like they could be separated.

Adam Draskovich.

[jagged ben]

I guess it depends what you mean by separated.  You can place the detection coils a long distance from the Watchman, but you'll still have to bring all eight sets of wires back to it.  I believe you can configure the I/O connections to drive your relays, and once programmed leave the Watchman unconnected to any com system.  You may find that you end up grouping your relay logic near each Watchman, but I suppose that depends on choices.  Voltage drop on the outputs running the relays could be an issue, I suppose.

[GaryHinshaw]

Here is a photo of a Watchman board installed on my layout (it's the board on the right in the plastic track):



The current sensing coils are placed out on the line where needed, and the sensor signals run via CAT5 wire to the top connectors on the board.  In my installation, the outputs of the board are sent along a DCC bus called a "Simple Serial Bus (SSB)" (the grey wire along the bottom of the track) that can talk to the command station and/or JMRI.  I use JMRI to manage the logic of these board: it displays block status on a computer panel and is also used to drive signal logic.  However, as jb notes, there is an auxiliary 10-pin connector in the middle of the board that can be used to drive relays based on the block status, and I *believe* this can operate independent of the DCC bus.

If you'd like to know more about the output modes, I strongly recommend that you contact Dick Bronson at RR-CirKits with further questions. He's a great guy, but his documentation is a bit impenetrable.  Best to just ask him simple questions and he'll give you simple answers.

[draskouasshat]

Thank you Gary! I'll get a hold of him and see what these detectors can do. Does it look like there are as many relay outputs to match the 8 blocks the detectors can monitor?

Adam Draskovich