Technical Discussion on BDL168/BD4

Introduction

The Digitrax BD4 contains 4 occupancy detectors.  The Digitrax BDL168 contains 16 occupancy detectors on a single board and report status over Loconet.  The BDL168 front end is a quad version of the BD4 with loconet reporting capability.  In other words, 4 BD4's with each BD4 making up a zone on the BDL168.  Both boards share the same analog detection circuits on the front end.  


Background Circuit Information:  

BDL168 and BD4 

DETECTION METHOD:  The Detection method is done by monitoring the voltage drop across two back to back diodes created by the current flowing through them.   However only one polarity (One DCC phase or one diode of the two) is monitored for voltage drop.  The detection assumes the current will be the same in both directions and hence the voltage drops will be the same in both directions.  Hence only on direction or DCC phase needs to be monitored for voltage drops.  The detection diodes are implement using 8 high current standard full bridge rectifiers.  A bridge Rectifier consist of 4 diodes pre-arranged to rectify AC into DC.  However in the boards, the circuit uses them very differently.  Two of the 4 diodes are used/assigned current detection for each detection output.  Hence a full bridge Rectifier supports two detection outputs.  

BASE SENSITIVITY: The BD168  and BD4 are designed to be sensitive enough to detect a 10K resistance on the rails.   Assuming a DCC track voltage of 12V, that translates to nominal "trip current" of 1.2mA (0.0012 Amps).   

BASIC ISOLATION: The BDL168/BD4 itself do not need any isolation from any Booster or autoreversing devices that feed them.  This is possible because the board's local ground is connected to the booster common which is implemented as a common ground (0V) for the entire layout as far as DCC power is concerned.  No DCC power coming from a any booster properly installed can go negative relative to this ground.   In other words, the booster ground is the same as the board's local ground.   

LED INDICATION OUPTUT ISOLATION: There is no electrical isolation from DCC power relative to the occupany indicators outputs.  The same terminals used by the LT-5 tester.

POWER LOSS DETECTION:  IF the input loses DCC power, then power loss is reported on the Track Status LED.

BD4 only: 

DETECTION CIRCUIT BREAKDOWN: There is only one input  that feeds all 4 detection outputs.   So all 4 detection output are by definition part of the same power district or "DCC power source".

DIODE VOLTAGE SCANNING:  There is no microprocessor on the BD4.  The lack of Loconet negates the need for a Microprocessor.  Each zone detection is done with simple analog and digital circuits.

DIODE VOLTAGE DETECTION SCAN TIMING: Since there is no Loconet support, there is nothing being used to drive the detection timing in that regard.   Since I have never had a look at the BD4 first hand other than through pictures, I have not determined what the BD4 does to protect against false occupancy detection at the DCC voltage phase transitions.  I will give Digitrax's design the benefit of the doubt that it has addressed somehow in someway.

BDL168 only:   

DETECTION CIRCUIT BREAKDOWN: The 16 inputs are broken down into 4 electrically independent power sections called "zones" in which each zone has 4 detectors each.   Each zone MUST has a single input and hence all 4 detection output are by definition part of the same power district or "DCC power source".   However each zone can be powered by a different Power District.   Stated another way, only at the zones level is there full Power District independence or isolation. 

DIODE VOLTAGE SCANNING: A 16:1 analog multiplexer circuit driven by the microprocessor on the BDL168 is used to scan the voltage drops of all 16 detection diodes into the uP analog input A-to-D in sequential order.  The scanning only occurs when there is a RailSync signal is present on the Loconet input.

DIODE VOLTAGE DETECTION SCAN TIMING: Although this information has not been obtained first hand but from a reliable source, the timing of the voltage scan takes place about half way in the time duration of the current phase.  In other words, the voltage is read well after the DCC voltage phase transition has been completed and the voltage is stable until the next phase transition.  The goal is to reject noise that exist during the voltage polarity transition that occur on each DCC  signal phase boundary.

LOCONET GROUND ISOLATION: The Loconet ground is only partially isolated from board ground via the use of a series 27 Ohm resistor between the Loconet Ground and BDL-168 Booster Ground which is tied to all the boosters.  The 27 Ohm resistor is used to restrict the amount of current that can flow in the loconet ground between the BDL-168 and any Booster.   27 Ohms is a 1000 times higher in resistance than the DC resistance of the wire used to connect the BDL-168 to Booster Common wire connection.  The current is NOT zero, it is just so low that in THEORY it should not cause any problems.  Great.  However as you add more and more BDL-168, that resistance isolation start to fall apart and more and more Booster current flow in the Loconet ground.  Why?  Every BDL-168 places more another 27 ohm resistor in parallel with each other on the other BDL-168s.  On large layouts with lots of BDL-168, this can become a problem in the form of ground bounce and noise on Loconet.

20K SENSITIVITY: An "option switch" sets the sensitivity which is applied globally to all 16 detectors.  10K is the default setting but the board also supports a 20K setting..   Assuming a DCC track voltage of 12V, the 20K setting translates to nominal "trip current" of 0.6mA (0.0006 Amps).
   

False Detection with Pure Capacitive Paracitic Load.

Below is a scope photo showing false track detection and why it happens.


Test Equipment:  
Oscilloscope:   Tektronix TDS-540 1 GHz 4 Chan Scope.
Current Probe: Tektronix TM-       50MHz AC/DC current probe.

Traces:
1) (Top) Differential Voltage Drop across the detection diodes @ 0.5V/Div.
2) (Middle) Booster Voltage @ 20V/Div (10X Probe)
3) Off
4) (Bottom) Current through detection diodes @ 5mA/Div. 

















What are we seeing in this scope picture?

We are seeing electrically what is happening in real time with an empty track block that is being monitored for detection by a BDL168.   The track block is about 30Ft long and there is about the 60ft (30ft or bus run) of wire feeding the track.  All of the track and wiring is part of a hidden helix so there is no track ballast involved.  Just commercial track mounted on homasote on top of plywood.

Trace #2 is the DCC voltage waveform which is transitioning from a negative phase to positive phase.  This represents the same timing as on the Loconet bus Rail Sync signal that the BDL168 actually uses to determine when to do detection.  REMEMBER the RailSync is used to DRIVE the boosters DCC signal output so from a timing perspective it is very accurate.  Trace 2 will be our reference waveform relative to what the other waveform/Traces mean.

Track #4 is the current that is flowing through the detection diodes.  (Ignore the negative DC offset for that is part of the instrumentation error at this very low current level.  It is not real.)  You will notice as soon as Trace 2 starting transitioning from a negative voltage to go positive, voltage all of a sudden there is erratic but non the less current flowing in the detector as shown in Trace 4.  Current flows positive all during the voltage transition until the voltage reaches the maximum positive voltage.  As soon as the voltage reaches that maximum point, the current quickly collapses to zero. In other words the current flow stops.

What is causing the current flow if there is nothing on the track?  1300pF or track and wire capacitance.  Unlike DC, DCC is a form of AC and as a consequence wire properties such as capacitance, which was ALWAYS THERE, now becomes a factor in what happens electrically.

Given we believe that detection is not perform until the we are about half way though this positive phase, any transitional current flow has had plenty of time to settle back down to zero and in fact it shows that has happened.  Now in theory, if the current through a diode is now zero, the voltage drop across the diode should be zero too.  Hence when the diode voltage drop snapshot is taken, it should be zero voltage indicating no detection.   The current that happens at the transition will not confuse the occupancy detector.   But lets look at the ACTUAL voltage drop across the diodes which is what the BDL-168 is looking at.

Trace #1 is the voltage drop across the diodes which is what is seen and measured by the BDL168 micro.  The initial voltage is zero.  But as soon as the DCC voltage transition from negative to positive and current flows in the diode, a 0.5V diode voltage drop shows up on the diode.  This is expected.  But as soon as the current goes to zero, the 0.5V voltage drop DOES NOT GO TO ZERO as one expects.

What is causing this voltage drop to hang around?  A property of semiconductors called "Storage Time" as in charge storage time.   Capacitors store charge.   Storage time is a simplistic way of saying this diode has a capacitance property associated with itself between it two terminals.   It remembers the voltage drop until that voltage (Charge) can be bleed off.  

THE PROBLEM: If you look at the scope photo trace 1, the diode voltage is held well into half way through the positive phase duration timing even though no current is flowing.  The BDL-168 reports the block is occupied even when there is nothing in the block.

It turns out that the amount of capacitance you have influence the duration of this storage charge.  If one reduces the amount of track and wiring capacitance, the duration the diode holds the voltage is reduced.  If the duration become less than half way through the phase duration timing, the BDL-168 will stop showing occupancy.

THE SOLUTIONS:  

1) You may not be able to do anything about the track capacitance, but you can reduce the wiring portion of the capacitance by separating the detected wire 1" apart from anything wire.  Do not twist or bundle this wire together nor with any other wires.  This is addressing the problem at its root.

2) Given the wiring cannot be modified, the second alternative is to reduce the sensitivity buy adding a resistor between the input and each detection output.  This resistor bleeds the diode's charge voltage away.  This option also allows you to tailor the sensitivity on each detected output to match the needs for each detected block.  However is selecting the final resistor, add some margin.  To learn more about what to do, go here: BDL168 & BD4 False Detection/Noise


Diode Voltage Drop versus Current Detection Mechanism:

When any current flows through a generic rectifier diode, a voltage drop will appear across it.  The voltage drop is a function of the current but not linear.  Here is a graph of a typical voltage drop of a 6 amp rectifier diodes when operating low currents (<1A).  The voltage drop for identical diodes will be very close to each other. 


Notice the voltage varies from 0.35V @ 10uA, 0.5V @ 1mA, 0.7V @ 100mA.  That 0.35V change in voltage is over a 10000 to 1 change in current!  The diodes ability to develop a vary narrow but consistently low voltage drop over a very very wide current range makes it ideal for a very sensitive current based occupancy detector while still being able to handle high load currents.   If one was to take the diode's voltage drop and run it into a voltage comparator with a trip voltage of 0.4V, then any current above 100uA will cause the comparator to trip and indicated occupancy.


Other Technical Notes about the BDL168

BDL-168 Loconet Operational Information:

Each BD168 is given it own Loconet Address and has both Loconet master and slave capability.  When there is a status change, the BD168 will become a loconet master and broadcast the detection status changes over loconet for any other Loconet device to process.  However at the same time any other device such as a PC can read status directly from a given BD168 (acting as a slave) at any time.

BDL-168 Zones

The 4 zones allow the BDl-168 to support up to 4 independent power districts including autoreversing sections and provide detection for all of them at the same time.  This allows better utilization of the 16 detectors in covering a geographical area of the layout with multiple lines going in all different directions

BDL-168 Power Loss Detection.

Each zone can also independently detect the presence or loss of DCC track power.  This is important since the loss of DCC power inside the zone will prevent occupancy detection from functioning.  Normally the detection would falsely report all blocks as unoccupied regardless if it is or worse is not.  Per the prototype, when there is any potentially unsafe and/or unknown condition, the correct response is to display the most restrictive signal aspect which is normally "red".   When the BD168 detects the loss of DCC power in a zone, it will force the 4 detected inputs belonging to the zone to indicated an occupied state along with a status showing loss of power.  This would allow the "signal logic" to force all signals that protect the dead tracks to drop to red.  The BD168 will also report a status showing loss of power in that zone allowing the same "signal logic" to take a different set of actions if so desired.