24.0 Summary
This all started when I read about the details of prototype operating runs with Dynamometer Cars. They were used by many roads, and were manned, in part, by design engineers working for the railroad. I had recalled reading about HO scale cars, even some claiming to be “operating.” I found their operating characteristics to be unsatisfying (and un-usable). So this began the journey of that one wild brain cell that had bugged me before asking “I wonder if it could be done?” This exploration has ventured down many rat holes, finding a large number of methods of how not to build these. This project evolved to considering what measurements were appropriate to a model railroader on any size layout, irrespective of DC or DCC operation. The “instruments” necessary for the appropriate measurements didn’t exist in the right form factor. If they could exist, how would the modeler use/read them? A car with a display on one or two sides gives a fleeting view as its train moves past the observer. Even if you could read the information, the prototype showed that a run over the road, under normal operating conditions, is where the real value lay. That would entail gathering much data. So much that one would need a way to view and manipulate it, to make the whole endeavor worthwhile.
This article describes a model laden with instruments that measure: time, speed, distance, drawbar pull, and track voltage (both DC and DCC) all of which are accurate enough to give a layout run meaningful results. Drawbar pull and track voltage are displayed on both sides of a heavily modifies Athearn RPO heavyweight passenger car. Data gathered from the on board instruments are sent to a battery powered, handheld receiver, which displays drawbar pull, track voltage, and speed in the form of three graphic charts in real time. Raw instantaneous numeric readings are also presented at the bottom of the display. A button is provided to the operator which, like the prototype, will “mark” the recording to note items of particular import to the operator – like towns, towers, interchanges, crossings, track frogs, bad sections of track, and the like. All the data captured and displayed are simultaneously stored in a readable text file on a micro SD Memory Card (like those used in cameras) for later processing on your personal computer or laptop. An MS Excel spreadsheet template (provided) can then be loaded directly from the text file, to display the entire run as a multiline graphic set, which the modeler can analyze to their heart’s content. This might be particularly useful for maintenance in a large layout or club, where the run might indicate bad trackage or electrical connections.
The “proof is always in the pudding” so I will try to take this little laboratory on wheels to larger layouts (which is not what I have myself) to get some experience and report the results here. In any case, I do not know exactly where this will wind up. Appropriate reader ideas are very much encouraged along the way. I will number the major sections, when I remember, to aid in directing questions and comments. I am already considering three more “instruments” that might be useful: side to side angle measurement, track noise measurement for bad joint detection, and a track view video camera to measure track width. I may get to present some of the experimental trials (and failures) too, which may be too much for the weak hearted! J Consequently, I may delete some of this too! (Be forewarned!) As I said before, I intend this to be more of an exploration than a conclusion.
24.1 The Adventure Begins: What Do You Do with a Dynamometer Car? – The Prototype and the Model
Dynamometer cars were sometimes rebuilt cars (like passenger cars) and sometimes built from scratch. Their front coupler was connected to springs via a mechanical linkage, or to a hydraulic piston assembly, or in modern times, to heavy duty strain gauges, where force could be measured and readings saved on a graphic chart recorder. The rotation of the wheels was sometimes coupled to the recorders to synchronize speed and drawbar pull. Measuring pens would move from side to side across a rotating drum of paper to record measurements against speed or time. With no GPS system, a man would be stationed in a cupola towards the front of the car, and note the passing of mile markers by voice, or by pressing a button that would ink mark one of the recordings. Another man stationed at the recording desk would annotate mileage, location, and other significant data as they progressed, by writing directly on the charts and in a log book.
Photo Courtesy Of the New York Central Historical Society- Used with permission
Drawbar pull measured the force exerted by the locomotive at the coupler as it traversed its route over the line. The force times the speed is a measure of power. The power exerted over time was the measure of work done by the locomotive. The effective use of this power, coupled with fuel and water consumption gave the railroad an understanding of the efficiency of operation and the value of the new loco design. But drawbar pull, speed and distance were not the only things studied. Firebox and boiler temperatures, operating settings (valve and throttle) , boiler pressure, break line pressure and other readings and were also measured and recorded. All things the railroad Engineering Department needed to evaluate a new design.
Photos Courtesy Of the New York Central Historical Society- Used with permission
Sometimes, the same kind of run would be made when new appliances like booster engines, feedwater heaters, pumps, and the like were added to an existing locomotive. The locomotive itself would have measurement sensors and gauges added, whose wiring (via cables) would run over the tender and into the chart recorders in the dynamometer car. Dynamometer cars were sometimes used to characterize the tonnage and motive power requirements over different divisions. Some cars were used to measure track curvature. Others were used to make efficiency tests regarding fuel consumption, fuel stops, station stops, and helper requirements. There are some excellent articles in the third quarter 1975 and the fourth quarter 2003 issues of the New York Central System Historical Society ( http://nycshs.org/) magazine “Central Headlight,” concerning dynamometer car use in locomotive trials.
Photo Courtesy Of the New York Central Historical Society- Used with permission
NYC Advertisement - Courtesy Of the New York Central Historical Society- Used with permission
24.1.1 Prototype versus Model
All this generated new modeling ideas, but especially the one that asked “Could a working model be built for a scale model railroad?” Ah… another challenge! Live steam railroaders have built working model dynamometer cars to evaluate their own models. You can find an example in this excellent report: “1/10 Scale Railway Dynamometer car” by Allan Wallace and John Lyas of Adelaide, Australia here -- http://tinyurl.com/j42e4dl
The September, 1937 issue of Model Railroader described a “working” dynamometer car in O Scale using a clockworks spring loaded mechanism with a limited range for measuring drawbar pull on a car mounted scale. In the September 1945 a plan was presented for a recording O scale dynamometer car using a miniature, scratch built chart recorder. No subsequent report was made that this plan was ever built.
Believe it or not there have been at least two attempts to provide a commercial “working” dynamometer car. The first one that I am aware was manufactured as a kit by the Devore Company in August, 1952. It provided a linkage from the horizontal movement of the coupler, to pull on a string wound around the axle of a circular “scale” that would show an indication of drawbar pull via the rotation of an arrow on the scale. The mechanism was spring loaded.
Devore Dynamometer Car of 1952
The second attempt was made by Walthers in 2001. This car also used a spring mechanism tied to the coupler, which moved a pointer against an unmarked, vertical scale placed at the side loading door. Walthers had initially announced the car would be equipped with a “digital readout on the car” and a “remote infrared data link” with a “data readout at your control panel.” The car was actually delivered with a note announcing that the digital readout and remote data link would not be offered.
Walthers Dynamometer Car of 2001
24.2 Construction: Recording Dynamometer Car with Drawbar Pull, Track Voltage, Speed, Distance & Time
HO Scale Operating Recording Dynamometer Car On DCC Rails
24.2.1 Measuring Drawbar Pull with a Load Cell (Force Sensor) – The First On Board Instrument
I built at least three different spring driven linkages connected to the coupler, and found all of them to do a poor job. One thing I wanted was the ability to measure a wide range of drawbar pull. (The load cell I eventually used can easily measure more than 20 ounces and is rated for 2.2 Kg.) For a spring mechanism to work over the range I wanted, required a long, weak expansion spring that would necessitate coupler movement that would be excessive. To shorten the movement, one would need a shorter, stronger spring that would lose the resolution needed for HO scale. There are 2 other force sensors I considered: resistive sensors and load cells. Resistive sensors literally change their resistance, when a force (weight) is applied. Their biggest problem is that they exhibit a hysteresis effect—that is, they don’t return to their original state when the load is removed. The last sensor I had was a “load cell” or “strain gauge.” In the form I used, this is a metal bar that has four resistive connections which form a “wheatstone bridge.” This is a sensitive arrangement of four resistors in balance. A small strain, or torsional, non-damaging, stress of the metal bar will unbalance the bridge, and this imbalance can be measured. Fortunately there is an integrated circuit (and small module) available for a modest price that performs the interface and does the conversion for us – the HX711.
Load Cell and HX711 Interface
The small load cell came from a Dymo digital postal scale for about $15. The HX711 module was purchased from ebay. The load cell will measure the force applied to a fraction of an ounce, and requires very little movement of the coupler shaft. The key to using the load cell in this fashion is the mechanics needed to apply as much of the force from the drawbar load to the load cell as possible, with no other flexing of the mount, flooring, trucks, or car body as possible. The load cell is held in place by a 10-32 screw, mounted to a shaped holding block, securely mounted to a 1x3/64 inch brass base extending the length of the car. In order to minimize the coupler linkage, a small brass “lip” was screwed to the bottom of the end of the load cell nearest the coupler. This allowed for a short, brass lever which the coupler shaft would pull to exert force on the lip. The lever was shaped to make contact with the lip from the top.
The Load Cell with HX711 Interface Module Dymo Digital Postal Scale
Athearn RPO Car with 1” x 3/64” Brass Base and Coupler Linkage Mounted
Shaped Brass Holding Block Load Cell Holding Block Mounted to Base Plate
Close scrutiny of the brass holding block will show that the block was sloped to allow the length (about 2 inches) of the load cell to fit within the car body, and yet present as much of an angle as possible to the drawbar lever. The drawbar lever turns on a 1/8 inch round brass rod held horizontally in place by 2 holes in the inner sides of the car frame. The car body, once mounted, will prevent any excessive side to side movement. One should also note that the front of the block has been filed down from the tapped mounting hole of the load cell forward, to expose the underside of the load cell bar, without weakening the mount. This allows for maximum torque to be applied to the bar by the drawbar lever.
Load Cell Mounted and Wired to Control Module (Yellow Color on Brass is Kapton Insulating Tape)
Here is the load cell mounted on the brass block. Unfortunately all the pics I took obscure the contact point of the coupler linkage. What needs to be done is to transfer the pulling force on the coupler as efficiently as possible down on the small brass lip on the bottom of the load cell. The load cell needs to act as a lever, solidly anchored on the brass block, and "bent" down by the tractive force of the loco at the opposite end of the load cell. Below is a slightly exaggerated link that extends into the end of a long Kadee coupler shank, and pivots about a 1/8 inch brass rod (axle) extending from side to side in the car chassis. I never glued the rod, as it is held in place by the outer car sides. As the link pivots, it presses down on the lower brass lip of the extended load cell. I think I filed a tiny notch (exaggerated in the picture below) to try to get the effective pressure on the load cell to be at a right angle to the load cell as much as possible. If you look closely I also removed material on the top of the large brass block, as close as possible to the mounting screw, to allow maximum flex of the load cell. Remember, there is a calibration process, so close is good enough.
The arrow shows the contact area where the link hits the load cell. The circles outline the brass lip on the bottom end of the load cell. What you are trying to do is transfer as much tractive force as possible to the load cell, to effect maximum sensitivity.
24.2.2 On-Board Instrumentation – Measuring DCC and DC Track Voltage – The Second Instrument
One of the specific capabilities I wanted in the model dynamometer car was its ability to measure track voltage. Track current is specific to the individual load (loco) and not relevant, but the drop off or droop in the power distribution on a layout is often debated by modelers. The RRAmpMeter is a great product often cited in DCC discussions as a great tool – and it is. Its documentation states that it measures the DCC track voltage with an RMS (Root Mean Squared) measurement technique giving an accurate measurement, and significantly better than an AC or DC meter – also true. But the statement is not always correct. RMS measurement gives the correct peak waveform voltage for a square wave. If you are using DCC “stretching” to control a DC loco on DCC controlled track it will become inaccurate. The good news is on many if not most DCC layouts, my experience is that this is a rare event. There is another way to measure the DCC top or peak voltage, and that is with a peak voltage meter/circuit. In the general case, this is not nearly as good a method as the RMS technique, but it can be calibrated to give nearly the same results in many cases at a fraction of the complexity (and cost). But this method’s accuracy is subject to noise and signal degradation. That is, there are cases where this method will give incorrect readings (read poorer, faster than the RMS technique). Nonetheless, looking at a $19 chip to do the RMS reading, I chose to build a peak reading voltmeter for the prototype dynamometer car, and see if it is worth upgrading or not. I may reconsider this later. So far, the results have been very satisfying with the lower cost solution!
24.2.2.1 Measuring the Peak and Adjusting for Height!
The microcontroller has a 10 bit analog to digital converter on board. The Moteino is a 3.3 Volt device so the DCC voltage needs to be lowered to a safe level for conversion. A full wave rectifier converts the DCC signal to a DC noisy signal that a 5 to 1 voltage divider reduces to a safe level. A 1.0 uf non polarized capacitor not only filters the high frequency noise but will retain the peak voltage so converted, since there is very little current draw by the analog input of the processor chip. There is a voltage drop through the bridge rectifier, but the processor will “add back” the drop in converting and scaling the voltage reading. In fact, by reading the track voltage with a “true RMS” meter or RRAmpMeter at the same point on the rails as the dynamometer car voltage pickup, you can change the offset calculation to “calibrate” the car reading to be exactly the same! I have been a bit surprised at just how well it works.
24.2.3 On-Board Instrumentation – Tachometer for Speed and Distance – The Third Instrument
The tachometer was built last. My thinking was that it was the simplest to design and build. Wrong again! The Hall effect sensor I had used in my wire guided tracked crane (see MRH January 2016 http://mrhpub.com/2016-01-jan/online/files/252.html ) was a special micro power latched sensor that was way too slow for a tachometer! That only took 3 days to figure out it would never work! A new, faster sensor was found.
The last axle on the rear truck has no power pickups. It has four tiny rare earth magnets super glued at 90 degree increments around the inside of one wheel. As it free spins in transit, a Hall effect sensor (Allegro A1126LUA-T Omnipolar Hall Effect Switch Digikey part # 620-1423-ND) picks up the movement as each magnet switches the device on in turn. Direction of movement is not observed. The sensor is soldered to a short section of PC board (or PC tie) material screwed to the truck bolster. Thin flexible wires are soldered onto the PC material and routed through a hole in the car chassis to the microcontroller.
24.2.4 On-Board Control and Data Transmitter
On board measurement, sequencing, control and communications is what this is all about. The heart of the prototype car is a “Moteino R4” – almost exactly an Arduino Pro Mini with an RFM69HW transceiver mounted on the bottom the processor board (available here: http://lowpowerlab.com/).
This is the first time I have used nearly all available memory space on the little microcontroller. This sketch in the car controller uses 6 Arduino libraries, and making them “cooperate” was …a challenge. Data packets (telemetry) consisting of the time in milliseconds from the start of the run, a tachometer reading from the last packet sent, the drawbar pull in fractional ounces, and a 5 point average of track voltage with resolution to better than 0.1 volts are sent out about every second of operation.
The sketches for the on-board Moteino and the receiving/recording Moteino are here: Dynamometer_codeV3.zip
Basis of the Instrument Control Module (Keep Alive Caps, DC-DC Converter, and Arduino/Moteino)
Straightforward Point to Point Wiring
Dynamometer Car Sensors and Transmitter
24.2.5 Special Track Pick-Ups
To make all this work, I depended on every wheel in both three axle trucks. The six wheels in the front truck are all used for power pick up for the on board electronics. They feed a full wave bridge rectifier with 8800uf filter and stay-alive storage with a high efficiency DC-DC switching converter/regulator providing the 5 Volt power for the car. This is the same configuration that I described here: SMA17 – Cheap Flicker Free Car Lighting for DCC, DC, and AC – a New Kind of KAOS
https://forum.mrhmag.com/post/sma17-%E2%80%93-cheap-flicker-free-car-lighting-for-dcc-dc-and-ac-%E2%80%93-a-new-12200310
Two axles on the rear truck feed the separate full wave rectifier used to measure track voltage.
The last axle on the rear truck has no power pickups. It has four tiny rare earth magnets super glued at 90 degree increments around the inside of one wheel. As it free spins in transit, a hall effect sensor picks up the movement as each magnet switches the device on in turn. Direction of movement is not noted
Modifications for the 6 Wheel Trucks
Mods for the Rear Voltmeter & Tach Pickups PC Tie Slotted for Hall Effect Sensor Mounting
24.2.6 The On-Board Displays
Two 0,96 inch OLED display are mounted in the baggage door windows on both sides of the car. These are fixed color 128 by 64 pixel displays. The top of the display is yellow and the lower, larger portion is blue. The instantaneous Drawbar pull measured in tenths of an ounce, and the peak track voltage at the car’s location is displayed. The car’s on board, 5 volt supply powers the displays.
0.96 Inch OLED Displays Temporarily Mounted in the Door Windows for Alignment
0.96 Inch Display with Four Wire Cable and Connector
Bill of Materials (May be incomplete, but it's what is available now):
24.2.7 The “Chart Recorder”
Well now, if one has a working dynamometer car, one must have some chart recorders! Enter the “chart recorder.” Pictured is the chart recorder display: time is linear point by point in approximately 1 second intervals on the horizontal axis. The top chart in white is instantaneous speed. The next lower chart, in red, is a 5 point average of track voltage for the same time. Next down in blue is plotted the instantaneous drawbar pull in ounces (and yes, for all those modelers in Metric land, one could change the calculation to grams!). When the charts fill up to the screen, they continue with a new “page” from the left.
A 4.5 Volt (3 AA 1.5 Volt) battery pack is attached on the left. The red button mimics the “man in the cupola” on the prototype dynamometer car, and allows the modeler to “mark” locations of note (towns, crossings, frogs, bad track, or whatever) – a special marker is recorded at the time and mileage point that the button is first depressed.
The lower portion of the display presents the numerical, instantaneous data. These are updated on receipt of new telemetry.
All data is recorded in a readable, MS Excel compatible test file on a Mico SD Memory Card. I’ve tried successfully using a 32GB card inserted into the top of the chart recorder case. Based on the data collected already, a 1 GB memory card will hold about 33 minutes of data from a continuous run. So a 32GB card should hold enough data for more than a 17 hour run. I don’t know if the batteries (or I) will last 17 hours. I haven’t tried 128 GB cards yet.
The Chart Recorder – Battery Pack, Graphic Display, Telemetry Receiver, Mile Marker Push Button
24.3 Using The Data
I mentioned before that the text file could be loaded into a spreadsheet.
A comma delimited, text file (you can load a formatted, empty card) will be created on the card and named DYNALOG1.TXT
If you open Microsoft Excel with this spreadsheet (it’s a preset template):
http://www.scalemodelanimation.com/Dyna/Dyna_template3.xlsx
and on the top Menu select Data then From Text (in the External Connections section) you can load the generated DYNALOG1.TXT file (comma separated) into the default cell of the template, and it will automatically generate the charts for the first 2000 sample sets of the run. (The 2000 can be extended.)
The significant advantage of dropping the file into the spreadsheet is that you can take your mouse and hover over any point on any graph and it will identify the time (horizontal axis) and value (vertical axis). Knowing the time to can go to the distance graph and find the exact location (distance from start) at that time step. So you can locate any voltage drops, disturbances, or whatever you find along your route. The “Mile Markers” from the button presses are seen as small vertical “blips” on the bottom of the chart. I don’t have a way (yet) of annotating the mile marker blips automatically. BC&SJ’s Horace Fithers suggested that the chart recorder be equipped with a microphone for adding verbal annotations! (Not likely) I think Horace had been spending too much time in the lounge car at the time.
I have yet to get to filtering, averaging, scaling, adjusting for planetary shifts, gravitational waves, and making things look pretty. I am fully aware that conversions can be done to scale MPH, scale distance, fast time, slow time, no time, the metric system, and language translation of labels.
24.3.1 The Test Loop Run
My test loop is exactly that – a double loop of track that I have hacked up a couple of hundred times that still works, and can be taken down and put up at will. On this expansive pike the first trials were run! Much to my very great surprise the car actually performed well! (You have no idea how high my frustration level was running that day.) I started learning things that I had not paid attention to before.
First Trial Run (I Hoped It Stayed on the Track!)
Here is the plot from that run: http://www.scalemodelanimation.com/Dyna/TestLoop1.xlsx
Time versus Distance on the Test Loop
Measurements from the Test Loop Run
24.3.2 The Bear Creek and South Jackson Railroad Run
Having secured proper authorization from Mr. Charlie Comstock, General Manager of the Bear Creek & South Jackson Railroad, the new dynamometer car was placed behind a lash up of two BC&SJ RS-3 road diesels with an additional coach and some 21 boxcars of dubious origin. Starting out of the Mill Bend yard, we passed through Deschutes Jct, Baynes Valley, Canyon Creek, Tunnels 3 & 2, Oak Hill, and around the big turn! Three runs were made in the same direction. The return trips were not recorded. Pick up wiper problems were found and were repaired. The raw data follows. Of particular note is the BC&SJ’s track operating voltage is on the average lower than my test track. Minor drop outs did not seem to have much effect on the motive power. One of the diesels lost power temporarily, and its side effects can be seen in the drop out in speed and drawbar pull in the middle of the third run. Additional data demonstrates changes in drawbar pull as the train goes up and passes a 2.3% grade, later level, and then downhill. Drop outs occur over dead switch frogs too. Some minor electrical glitches are indicated too. Speed jitter is presented in the raw data due to the way the tach interacts with the packet transmission. The slope of the distance curve indicates relatively smooth speeds, which likely means the speed plots and/or data transmissions will need to be averaged over the raw data. (More to come) The loaded, full Excel spreadsheet can be downloaded here:
http://www.scalemodelanimation.com/Dyna/BC&SJ-1.xlsx
The raw data can be found here:
http://www.scalemodelanimation.com/Dyna/BC&SJ-1.TXT
The best part of all …IT WORKED! YEAH!!!
2518,0,55,0.00,11.93,0
3417,0,55,0.00,11.95,0
4317,1,56,0.02,11.93,0
5216,1,57,0.00,11.95,0
6116,1,58,0.00,11.93,0
7015,0,58,0.00,12.00,0
7914,0,58,0.00,12.08,0
8814,0,58,0.00,11.95,0
9713,0,58,0.00,12.00,0
10613,2,60,0.03,11.88,0
11602,3,63,0.06,12.00,0
12592,1,64,0.00,11.95,0
13581,0,64,0.44,12.01,0
14570,0,64,0.63,12.01,0
15560,0,64,0.68,12.06,0
16549,0,64,0.74,11.98,0
17629,1,65,5.68,12.09,0
18618,0,65,0.86,12.03,0
19607,0,65,0.88,12.08,0
…
1482549,4,796,4.81,11.79,0
1483629,3,799,4.99,11.75,0
1484618,2,801,4.72,11.80,0
1485607,8,809,4.17,11.82,0
1486598,2,811,4.17,11.92,0
1487587,7,818,4.49,11.85,0
1488577,1,819,3.74,11.98,0
1489566,0,819,3.22,12.09,0
A Small Portion of the Raw Data from the Bear Creek Runs
Excel Spread Sheet with Graphs of the 6210 Captured Data Points of Three Short Runs on the Bear Creek
Distance Tracks for the 3 Bear Creek Runs
Detailed Data Plotted for the 3 Bear Creek Runs
24.4 Comments and What’s Next!
There is much more to say, describe, and show. More pictures, descriptions, schematics, code (sketches), trials, data, and analysis are coming. It may not seem like it, but this is one of the most complex models I’ve ever built. This all started at the beginning of 2016, and it’s not over yet. I once thought that everything worthwhile I could think of putting into this was covered. Since that point, I made the mistake of reading another article about Southern Pacific’s dynamometer cars (and the additional measurements they made), and another about track geometry cars (now entering information overload)!
Charlie Comstock pointed out, after learning that the “chart recorder” (the telemetry receiver) could easily be interfaced directly to a PC or laptop, that one could directly measure the actual speed of a locomotive (DCC decoder equipped), and with a bit of coding and an interface to JMRI Decoder Pro, with speed measured at every speed step, one could “speed match” a decoder’s speed table to a modeler’s ”standard” speed table, in one run with a single loco! I think I’ll put that on the stack of future enhancements for now.
I hope you find this of interest. As always, appropriate comments and suggestions are always welcome!
Have fun!
Best regards,
Geoff Bunza