How Many Bus Terminators is OK

[Fishmanoz]

All, just a couple of summary points from me back on topic for this and the other thread on suppressors.  I have been reading with interest but didn't get to commenting when someone suggested I might be able to add something, or until now.

 

First, the Mark Gurries reference would seem to be the definitive one for what is happening.  Note though the assumption that individual bus sections are "long", else when you add suppressors on short sections such that they are close together, they must be considered to be in parallel with each other, not isolated from each other.  This will reduce the overall cutoff frequency.

 

Someone spoke about the suppressors being high pass filters.  They would be if they were in series with the power supply, but in fact they are in parallel and so low pass filters.  They short out higher frequency components like those found in transient spikes.

 

Someone spoke about transients occurring with a short prior to overload cutout.  Transients also occur on recovery from momentary shorts that don't result in an overload cutout.

 

Someone spoke about the suppressor cutoff frequency of around 16kHz versus the DCC frequency of 8kHz.  Don't forget that DCC is not sinusoidal, it is a square wave and also the PWM means that there are short pulses produced.  Frequency domain analysis of this tells us that both harmonics of 8kHz and higher fundamental frequencies are involved and significant.  Consequently, if you put 2 suppressors close together, so reducing the overall cutoff frequency, you will get significant distortion of the overall waveform due to reduction in its higher frequency components.  The waveforms will be rounded, not square, and the rise time of the leading edge of the pulses increased, eventually putting them outside the NMRA spec.  This also explains why the suppressor in a DC power track interferes with DCC operation and must be removed.

 

Finally, inductive effects from the motor and back EMF were mentioned.  Don't forget the motor is isolated from the track by the decoder in DCC operation.  Similarly, pulses from solenoid point motor operation are isolated.

[PJR_slo]

Hi Chris, Greynut

 

Rather than say anymore on this thread, if you like I can explain the layouts/controls further in the Off Topic forum.  But not tonight...

 

Peter

[RDS]

Hi Chris, Greynut

 

... if you like I can explain the layouts/controls further in the Off Topic forum.  

Yes please!  But it would be a shame to hide it in the 'Off Topic' section that is intended to be anything other than Model Railways.

[PJR_slo]

Hi Chris, Greynut

 

... if you like I can explain the layouts/controls further in the Off Topic forum.  

Yes please!  But it would be a shame to hide it in the 'Off Topic' section that is intended to be anything other than Model Railways.

Hi RDS

OK, I will put something in General Discussion later today.

 

Thanks

 

Peter

[Greynut]

Hi Fishmanoz, many thanks for coming in on this. As you have probably noticed, I am feeling my way in all of the DCC optimisation side. With the answers in the thread, I'm getting a grasp on the theoretical side and that's new understanding for me. It's relating that to the practical side which I'm trying to get a handle on. But I do like to learn this way, not simply finding out what to do - but learning the 'why' as well.

 

So ... if the compounded effect of suppressors being too close together can be self defeating - in that the waveform is distorted due to the required higher frequencies being cut off - in practice, what would be termed as being too close together? Also, would that apply to using a suppressor too near to the DCC controller (eLink in my case)? I ask that as when I was first looking round the web, before asking here, I had seen that placing one there was an option. Not that I intend doing that but it would be interesting to know what effect that might have.

 

Peter - it would be great if you could let us know about your layouts etc in a new thread.

[Chrissaf]

I agree Peter, I think it is time to put this thread to bed (with Ray's agreement), it has been 'flogged to death' so to speak. Just one comment regarding the term 'High Pass vs Low Pass'. In principle Fishmanoz comment is correct. The orientation of the filter Series or Parallel determines the pass function when in a transmission line. I used the term High Pass to indicate that the Higher Frequencies were being passed through the filter to be shunted across the Bus wires. This still concurs with his comment.

[Chrissaf]

Addendum to my last post.

With reference to "with Ray's agreement", subject to Ray getting an answer to his question to Fishmanoz.

Sorry Ray only just spotted that one. The question that is.

 

And Thanks to Fishmanoz for clarifying the issue about having too many Suppressors in parallel too closely together. Your explanation makes perfect sense to me. Particularly, and as you say, a square wave becomes a square wave for the very fact that it contains high frequency harmonics of the base frequency. Thus, filtering out too many of these higher frequencies will cause distortion.

[Fishmanoz]

Greynut, good question about what is too close together.  In a theoretical sense, the suppressors can be considered independent of each other if there is a significant resistance and inductance existing in the bus between them.  The question is then how far before this might be the case.

 

Mark Gurries gives a clue.  He talks about needing suppressors for bus runs of 10m and more.  He then also says put one at the end and another near where shorting problems might occur.  My conclusion from this is they should stop least be metres apart.

 

I can't see a problem with putting one near the controller, then I also can't see a need to do it.

[Chrissaf]

Re: "I can't see a problem with putting one near the controller, then I also can't see a need to do it."

.

This came about because Ray's DCC Bus he is proposing is potentially a 10 legged Star configuration, with the controller at the centre. It seemed sense to put a suppressor at the star centre so that all legs had a similar (balanced) level of short circuit protection. I f he had all legs individually suppressed at the ends that would involve 10 suppressors potentially in parallel over relatively short individual legs. Hence all the discussion about the impact of multiple suppressors on the DCC signal.

[Greynut]

Cheers Fishmanoz and Chris.

I think the way to go is to start with just one suppressor in the power domain that I'm working on - but where? (If I employ a second domain, one goes there as well.)

 

As the busses all stem from the eLink and go off to power the different levels of the layout's spiral arrangement, I had thought about putting it 'central', as it were - near the eLink. I'm not certain though if it needs to go there - wouldn't the eLink itself act as a kind of suppressor with it's built in protection circuitry?

 

Also, the advice on placing it near to a shorting hotspot. I don't appear to have any of those though - no places that regularly give a problem. Thinking it through, I've had shorts, for different reasons, in various places - all on points. The ones that were problematic I've worked on until they behaved themselves and any shorts seem to be down to me messing something up. So that kind of leaves placement - to shield the layout from any particular area - as not required.

 

The idea of sticking them on the end of each bus would warrant too many suppressors and would end up draining a heap of current as well as the potential of overkill regarding high frequency suppression within the DCC signal.

 

My busses are twisted so that should be a positive in cutting down suppressor requirement anyway. My thoughts are now to put one in the centre of the power domain. This should catch any spike within the domain due to the busses being linked at the eLink output bit of the cables. To my (simple) mind, that would be no different to a short happening mid bus and relying on a suppressor at the end of the bus - it's probably doing a better job as it's being used in relation to overall bus length rather than individual bus lengths.

 

By using only one suppressor, if it works OK - great. And it does allow for an odd one to be added at a later stage if it seems necessary. 

[Chrissaf]

Ray,

.

One more rhetorical question to mull over. If the DCC controller had inbuilt circuitry that replicated the function of a suppressor. Then surely all the huge number of questions and posts the suppressor / terminator topics generate would be minimal. As you say, your Buses are twisted. Adding one suppressor for peace of mind is beneficial. Locating it centrally in your particular scenario is probably the optimum location, even if it is near the eLink output.

.

Best Wishes

Chris

.

PS - The controller inbuilt short circuit protection is more about cutting the track current to protect the controller rather than the protection of external decoders. For power supplies in general, the current rating relates to the maximum sustained output for which the maximum output is designed for. If you short a (non protected) power supply output they can produce peak currents that far exceed the rated value until they eventually burn out.

.

PPS - I've done it....hit 100 posts......time for me to have a break.

[Graskie]

Congratulations, Chris. (Good! that gives me another one!)

[Chrissaf]

Ray,

.

Haven't spotted any posts from you today. Are you having a night off??

[Greynut]

Chris, well done on your 100 posts - I'm a 12 further posts of your's late in saying that.

 

Many thanks for your reply. I see that any protection in something like an eLink would only be there to protect that unit. Might be an idea for the tech lads to think about adding a suppressor in as well though - it would only cost a few pence more to do - just a thought.

[PJR_slo]

Chris, well done on your 100 posts - I'm a 12 further posts of your's late in saying that.

 

Many thanks for your reply. I see that any protection in something like an eLink would only be there to protect that unit. Might be an idea for the tech lads to think about adding a suppressor in as well though - it would only cost a few pence more to do - just a thought.

Hi Greynut.

 

You forgot about the many thousands of pounds it would cost to redesign the circuit board and re-tool for the extra components...

 

Peter

[Greynut]

Ahh! Good point ... ignore my thought tech lads (lol)

[Gary_Leeds]

Just an idea, and sorry if its already been suggested, but rather than having all the hassle of putting bus terminators at the end of every siding how about using one of those LEDs that light the buffer up that just clip between the track? Wouldnt that do the same thing or would it not work because the signal is AC and the LED only lets one half of the wave through?

[Chrissaf]

Sorry to say but, LEDs are a completely different technology to a suppressor (a suppressor is a tuned frequency filter). Whereas an LED is primarily a DC technology device. So an LED wouldn't IMHO prevent voltage spikes. You may be thinking of a solution (that many people used some time ago) where filament bulbs were wired in series with the track power supply, so when a track short occurred the bulb lit up and acted as a current limiter. But that too is completely different to the way an LED works.

[Fishmanoz]

I would never suggest that a LED is primarily DC, it clearly isn't, nor would  I talk about a series RC filter as being tuned.  What is actually need here is a low pass filter which allows the DCC signal to pass but stops the higher frequency transients causing the problem.  A LED or any other diode can't do this but a low pass series RC filter across the track can.

[Fishmanoz]

Just on that star bus with 10 legs - I'm not sure why you would need 10 legs in the first place but, having done so, I suspect all legs are pretty short, although some will be longer than others and some will go to areas of the layout where the generation of shorts is more likely.

 

For a start, Mark Gurries tells us that short (less than 10m) under 5 Amp circuits may not need suppressors.  So I wouldn't assume that every leg needs a suppressor.  I would just look at fitting suppressors to the longer legs going to problem areas.

 

Just on whether controllers contain low pass filtration or not, it is largely irrelevant.  What we are trying to suppress is transients generated under fault conditions by the stray impedances in long DCC buses, and to suppress them where they occur at the end of the buses, not at the start near the controller.

 

And a word on the theory of long and short may be relevant here too.  Circuits that are short are those where along the length of the circuit, there is little or no phase change in the frequencies being transmitted.  Long is where there is a significant phase change, in fact even a number of wavelengths at signal frequency.  Put this way, simple circuit theory applies to short while more complicated transmission line theory applies to long.  It's going to be quite an enormous layout before it is wavelengths long at 8kHz (I'll let you do the maths on a Sunday morning thanks - we are talking speed of light divided by 2 x pi x frequency), but clearly phase change becomes significant over 10m given the Gurries advice.

[Chrissaf]

I would never suggest that a LED is primarily DC, it clearly isn't, nor would  I talk about a series RC filter as being tuned.  What is actually need here is a low pass filter which allows the DCC signal to pass but stops the higher frequency transients causing the problem.  A LED or any other diode can't do this but a low pass series RC filter across the track can.

 

A standard basic LED has a positive leg and a negative leg. It has a maximum reverse bias voltage that if you exceed, the LED will fail. Of course if you exceed the maximum forward bias voltage it will fail too, but the maximum reverse bias voltage is typically much lower. Because of the mismatch between the forward & reverse bias voltages, the LED does not lend itself to being used easily in an AC environment. In my view, those facts alone makes it primarily a DC device. Yes, there are circumstances where you can operate it in an AC environment, but for general usage you design your circuit to operate the LED in a DC one. Hence why I purposely and deliberately included the word 'primarily'. If I had used the word 'typically' instead, then would you have been so picky.

 

With the filter, the choice of component values determine the frequency characteristics. This is, in principle, a simple form of tuning.

 

I might be using different terms to the ones you would prefer me to use, but at a broad high level there is nothing wrong with them.

[Chrissaf]

A CORRECTION TO MY LAST POST DIRECTLY ABOVE:

I wrote: "It has a maximum reverse bias voltage that if you exceed, the LED will fail. Of course if you exceed the maximum forward bias voltage it will fail too but the maximum reverse bias voltage is typically much lower"

 

This should have read "It has a maximum forward bias voltage that if you exceed, the LED will fail. Of course if you exceed the maximum reverse bias voltage it will fail too, but the maximum forward bias voltage is typically much lower."

 

After "LED does not lend itself to being used easily in an AC environment."

Add "Particularly as the bias voltages used by an LED are significantly lower than those of normal signal and rectifier type diodes that are perfectly at home in an AC environment."

 

I wrote: "but for general usage you design your circuit to operate the LED in a DC one"

 

This should read to provide greater clarity of meaning "but for general usage you design your circuit to operate the LED in a DC one, because with DC you can more easily control when and if the LED lights up"

[Fishmanoz]

I agree with you largley (is that the same as primarily) on the LED.  However, I would expect on DCC to find a number of applications that just use a suitable resistor and let the LED look after its own reverse bias.  An example might be red LEDs on buffer stops, although I haven't checked. 

 

On tuned circuits, these are conventionally circuits containing both L and C components such that there is a relatively narrow passband around the resonant frequency of the filter, with frequencies above and below being rejected.  Think the tuning of your radio or TV set as examples where all except the desired band of frequencies must be rejected to receive only one station or channel at a time.

[Chrissaf]

I forgot in Australia it is the middle of the day, not the middle of the night (here in UK). I have the occasional bout of insomnia, so it is not unusual to find me sitting at my PC through most of the night. I'm sure you didn't mean it to be, but at first reading I found your comment abrasive, hence my long winded reply of defense and follow up fact correction. The first reply being written in too much haste.

.

I'm glad to have cleared the air, so to speak. No hard feelings I hope....

.

Yes and I believe you are right, the commercial buffer stop lamps (track clip variety) are simple LEDs with hidden suitably large resistors, working quite happily on DCC AC.

.

R - Resistance

C - Capacitance

L - Inductance

There is a fourth term that I have forgotten the letter for, for measuring Leakance......I think it might be G

And, of course, Z for Impedance.

Plus if my memory serves me (40 years ago now) there is some quite complicated logarithmic and calculus math that go with filter calculations.

[Chrissaf]

Follow up to last message:

.

Fishmanoz, I appreciate I may have inadvertently used filter terminology that has not been totally accurate in your purest view, but that is due to not having needed to do any filter theory for 40+ years. But I am familiar with the principles of high pass, low pass, band pass filters. Between 1971 and 1981 my work involved commissioning long distance (long distance within the UK that is, not the distances involved in OZ) FDM (Frequency Division Multiplexing) line transmission systems. The FDM equipment operating in the 64Khz to 12Mhz range. FDM by its very nature uses extremely tight tolerance filters to separate the individual analogue modulated carrier bands. Additionally, Low, High & Band Pass filters were deployed on lower bandwidth systems carrying mainly audio speech and music circuits. Of course, in those scenario's my work colleagues and I only had to commission and maintain them, not design them. Thus, the theory training given was more of a basic overview than in depth design. We were however expected to do some design theory and calculations in classroom training environments. After 81, the FDM systems were largely replaced with TDM (Time Division Multiplexing) and PCM (Pulse Code Modulation), the digital age was dawning and the use of filters diminished.