If your not signal technical minded or interested, don't read this post. You've been warned. :)

Let me preface with saying that I've not yet begun to build my layout; all I've done is accumulated rolling stock, prepped the room some 90% of the way and am still working on the layout design in WinTrack. So I've got no personal practical experience, just what I remember from electrical engineering university some 30+ years ago (as I've been all software since then ).

I was watching a YouTube video of a German modeller who does very excellent and well produced videos on his build in progress, exploring the various questions he needs to resolve as he develops his n-scale layout. His trial of various sound dampening underlayment and which adhesive to use were interesting watching. He recently did one on how to distribute the digital power around the layout (DCC in his case) where he explored some topologies about how to get power from the controller (a z21 in his case) out to the various track elements.

The points that piqued my memory and sounded not quite right were why he decided to go for a bus distribution, as opposed to a loop. That is, to not close the power distribution under the layout so that it electrically fully wrapped around in a loop. He gave some reasons for this choice, and those didn't seem valid in my memory of how digital signaling and capacitance affected signals. He also explained why he was going to twist the power buss wires, to reduce the capacitance effects on the digital power bus, again to get more reliably delivery to all ends of the layout. There was something that didn't sound right either.

I'm not saying that the things he was doing were going to be a problem, but that it felt to me that those things were just unnecessary and would not provide any real benefit. Like changing ones oil in modern car engines every 3,000 miles.

But its been 30 years and I cannot rule out I've forgotten way more electrical engineering than I remember or that I just forgot about some principle of physics and he is right, but I want to get a check with the folks that have practical experience and maybe others that have some physics/electrical signaling background, to see what I'm missing or if I'm correct in my conclusions.

So here it gets a big nerdy.

The principles are similar for mfx and DCC as far as the digital are concerned except the DCC system would gain a more substantial benefit from a complete loop of power distribution that an AC based mfx would not benefit from (as it doesn't suffer from the DC based voltage drop issues over long runs from resistance, as much).


This modeler had two main points that I'm having a problem with:
1) That the capacitive effects of the digital power distribution system is meaningfully affected by the ring vs bus solution
2) That twisting the power bus would reduce the capacative effects or improve signal quality of the digital data.

Let me put my thinking out first on why the warning light in the back of my old EE memories went off...

DCC has 2 basic square waves used to encode the 0 and 1 bits, which puts DCC signalling at 5kHz and 8.6kHz, well below the 10kHz+ range that is generally considered the start of high frequency digital signalling when you start having to worry about all sorts of things you can ignore at lower frequencies, but its close enough to take a closer look.

DCC and I think mfx as well, signal decoding is all about the detection of the edges of the square wave, that is, its the spacing between the falling edges of the square wave, that time frame, that determines if its a 0 bit, a 1 bit, or no signal. So to cause problems with the decoding of the digital signal at the decoder the signal would have to be so affected that the space where the square wave should be from high to low doesn't get down far enough for the decoder to detect that falling edge (before the controller again sends the signal high).

Now capacitive effects are real, that is, two wires running parallel to each other do have a capacitance that can be measured. The 'plates' of the capacitors are the copper conductors that have an effective surface area facing the other wire, and the space between those two wires has a computable dielectric permissity: there is a formula for computing the value of a capacitor from the surface area, the space between the 'plates', and the permissity of the space between the plates. In fact there is a YouTube video where a fellow does that with two 2m long wires (that appear to be about 20cm apart) and measures a capacitance of 10pF. So if we estimate that instead of 20cm, the wires are 2mm appart, one would get an approximation of 1000pF, or 1nF.

What capacitive effects cause, is that you can visualize the power distribution as a diagram where you have two ideal (0 resistance) distrbution conductors, with a capacitor between them, and a resistor at the end that represents the resistivity loss over the line from the sender (z21) to receiver (decoder on the train). This is an RC circuit and what is known about RC circuits is that you cannot instantaneously change the voltage on the line from level A -> level B; there is an exponential delay in the actual signal change as the capacitor charges/discarges. It rounds off the sharp corners of the theoretical ideal square wave. The higher the capacitance the more rounding of the corners, the longer it takes for the voltage to actually drop.

Now for this to be a problem for the DCC system, that decay rate would have to be so slow (the capacitor large enough) that the signal cannot get past the point of the decoder detecting the falling edge, before the sender (z21) is already back to going to the high level for the next "bit" of data signalling. With DCC, the higher frequency used , 8.6kHz, that low part of the wave form gives you 58uS.



Now using some data, such as 100m of copper wire of 2.5mm^2 has a resistance end to end of 688mO, and assuming that the decoder cannot detect the falling edge if the signal doesn't get past the 1/2 way point down, and the principles of how an RC circuit works, and that we have 58uS of time, I estimate that there would be a problem for the DCC decoder if the capacitive value on the power bus was 120uF or more. The approximate capacitance of the two parallel bus wires is on the order of 10nF maybe. No where close to being any concern. (this is why at such low frequencies like 5kHz these effects are generally ignored).

So in my thinking and rough analysis, the capacitive effects are no concern to DCC signaling frequencies to begin with.


Now beyond that, the bus vs loop and twisted pair assertions still bothered me.

1) There are a 'lot' of loops on a digital power distribution on a layout to the decoder. Every pair of droppers from the rail forms a loop. Each wheel pair on a loco/wagon that acts as a conductive power pickup forms a loop. So to rule out the main power feed not being a loop having an effect on signal quality seems wrong. In fact, for DCC having a loop would reduce the resistive losses of the power part of the signal as it cuts the worst case resistance to a remote point on the track down (in half) because you have 2 parallel wires to get to the power feeder to the track. So my assertion is the loop on the power delivery doesn't matter to signal quality, choosing to not have one doesn't help ( and in DCC gets you more resistive loss along the run).

2) The 'twisted pair' reducing capacitive effects or improving signal argument - that doesn't seem right on two fronts:

First the capacitive effecte between two parallel wires still exists in a twisted pair, the same facing each other surfaces fo the wire exist in both cases, they are just twisted, but I'd bet the capacitance is roughly the same.

Second, the benefit of twisted pair is it allows for reducing the effect of electromagnetically picked up interference along the run, from the signal and has nothing to do with capacitance. Its use in network cables, like Cat5 ethernet, works in a very specific way that I don't know if DCC (or mfx) necessarily use. The twisting ensures that the two twisted wires are equally exposed to whatever interference they may encounter along the run of the cable. This allows the elecctronics at the ends of the twisted pair connectors to extract the 'signal' by computing the voltage difference between the two wires. The interference being equally present in both wires just cancels itself out. But that has nothing to do with capacitance and that isn't how a DCC/mfx power distribution system works. The rail distribution system power lines connecting to many places on the rails, and all those rails then run, often multiple paths, to where a locco's wheels sit on the track then up the wheels and into the decoder. That isn't at all a twisted pair system from z21 to DCC decoder. And I doubt there is differential signal handling in use on the DCC or mfx system (correct me if I'm wrong). Its just a simple digital bus system with one master talker (the z21/cs) and a lot of listeners (decoders) some of which get a designed window to answer back when the master asks them a question.



So for anyone that has managed to read all the way to here and/or understands electronics and/or has experiences in building layouts and worrying about the digital signal quality:
1) Did I miss something substantial and as a result my analysis has a flaw?
2) Does it matter to have a digital power distribution system loop or breaking the loop to produce a bus/tree structure of feeders to the rails?

Hi!
I do not understand what you mean with "AC-based" and "DC-based" and why one would gain more than the other. DCC and mfx are both "digital" and both very similar.

DCC has no fixed frequencies, it has minimum and maximum timings for 0 bits and 1 bits and 0 bits allow for large variations and large asymmetry. The encoder side must comply with stricter timings than the decoder side and AIUI this allows for parasitic capacitance and other wire imperfections.
With DCC I think the polarity changes matter. So the edges can be detected by looking at the polarity.
mfx sometimes emits dummy bits to prevent mfx messages from containing valid DCC start sequences. mfx and DCC are very similar on the transmission level.

You do not mention feedback (unless I missed something). mfx feedback is much more fragile than RailCom feedback.

Long story short: I don't know if ring vs. bus makes a differences with the wire lengths and frequencies we deal with on MRR layouts - it should work either way IMHO.

Hello,

My layout is quite long, measures almost 7 meters (22 ft) long, but is narrow, at the ends 80 centimeters (31 inches) and 60 (23 inches) centimeters in the center, the command position is almost in the center, let's say that to the left there are about 3 meters of layout and to the right 4 meters. The digital and analog power supply (lighting and accessories) I do by cable buses, straight without twisting, separated from each other a few centimeters, (see photo). Besides, the model has 2 floors, and each floor has its own buses. It works. Once I was tempted to twist cables, especially those of turnouts and signals to the corresponding decoders.

This is a picture of a part of my layout


Now, I have an official publication of märklin, this is it:


And making the story short, in the part dedicated to large layouts, says that a loop is not a good solution because it can make an antenna effect, even be confused with a clandestine emitter. And that for big layouts the best way to do the cabling in arborescence, that is to say of the digital power station they leave a series of cables that branch in other thinner ones etc.

The text of this publication is in French


Best regards

Hi Minok,

It’s an interesting topic and questions I have pondered on as well.

I recall we have previously discussed the topic here and I never understood why the recommendation of not having a bus as a loop while the track itself many times will function as a loop...

I don’t have the answers, but there is some more technical reading on this website (scroll down a little bit): http://www.wiringfordcc.com/track_2.htm#a40

Maybe some users with large exhibition experience can answer the questions?

PS. For my small to medium sized layout I do have buses underneath the layout I have chosen not to loop it, but they are running out in a star configuration.

Best Regards
Lasse

Thanks.

Indeed, and it all comes down to what is 'high frequency'. I'm contend ding that the 8.6kHz high end frequency used in DCC (and similarly whatever the mfx frequency is), does not constitute a high frequency signal where any of the concerns about inductance, capacitance, etc actually matter. The higher the frequency, the smaller that half-wave of the ideal square wave is and the more of it can be consumed with capacitive, inductive, etc effects.. but you have to get up in those 100kHz, or MHz and higher frequencies before it maters. At least thats my contention.

I think there is some lost in translation happening here that is confusing you.

I'm talking only about the distribution of the digital power on the layout - the track signal if you will. It delivers the electrical power and has the data signal modulated on top of it. One electrical signal. (what the digital encoding scheme is on the layout, DCC or mfx, doesn't matter, as they are in the same general frequency range as far as I know, and the actual messaging protocols don't matter for the purposes of this discussion. I'm just looking at can you encode a 0 and 1 and decode it, not what a sequence of 0's and 1's mean in terms of the messaging).

A ring, as I'm describing, doesn't mean a short. I'm not connecting the power/signal line to the ground/return line; that would be a short.

What I'm asking about is if the capacitive effects matter if one chooses between these two power distribution styles:

Of course as soon as I post, I get the response from my electrical engineering contact that knows a thing or two about electromagnetic interference and signal processing. So before I forget I want to summarize here what he had to say on the topic, and I'll start with his summary statement (which is somewhat funny, but true from the perspective of a signals engineering perspective):

8kHz is practically DC.

OK, so it isn't, but as I'd mentioned in my own post, the various electromagnetic and signal processing effects that one has to worry about in high data rate transmissions happen only at far higher frequencies than 8kHz. Think MHz, or GHz. At the 8kHz range, these things just don't mater.

So on the two main points he had the following to say:





So I'd say: use the layout that makes sense for your layout geometry - that is don't use one that geometrically complicates things. If your layout naturally favors a tree or buss, use that, but if you do get a loop or can easily make one, do that. The resistance of the distribution run will be reduced to any point along the run, so less voltage drop at the feeders.

Maybe when I finally start making my layout building videos I'll make one on this topic and let the YouTube community throw darts at me for a change, but at least I'll die knowing that the material has the backing of two electrical engineers, one of which worked in signals and interference suppression for a career.

You can find the standard here:
https://nmra.org/sites/d...rical_standards_2006.pdf

Quotes:
"Digital Command Station components shall transmit "1" bits with the first and last parts
each having a duration of between 55 and 61 microseconds. A Digital Decoder must accept bits whose first and last parts have a
duration of between 52 and 64 microseconds, as a valid bit with the value of "1". [...]
Digital Command Station components shall transmit "0" bits with each part of the bit having a duration of between 95 and
9900 microseconds with the total bit duration of the "0" bit not exceeding 12000 microseconds. A Digital Decoder must accept bits,
whose first or last parts have a duration between 90 and 10,000 microseconds, as a valid bit with the value of "0"."

So we have factor 100 for the 0 bits. I don't call this a fixed frequency.



With MM protocol the centre rail will sometimes be positive and sometimes be negative.
With mfx protocol the centre rail will sometimes be positive and sometimes be negative.
With DCC protocol the centre rail will sometimes be positive and sometimes be negative.
Some people call this "bipolar pulsed DC". It behaves much like AC.
Only positive voltages - sometimes with plus on the outer rails, sometimes with plus on the centre rail.

Another quote:
"The NMRA baseline digital command control signal consists of a stream of transitions between two equal voltage levels that have
opposite polarity.
Note that since a locomotive or piece of rolling stock can be placed upon a given section of track facing in either direction, it is
impossible to define, from the point of view of a Digital Decoder, whether the first or last part of a bit will have the "positive" voltage
polarity."