Product selection help: Data logging for 74 volt locomotive wiring

[HeavyIron]

Hello there.

I am trying to troubleshoot an intermittent excitation issue on an old diesel locomotive. Once in a while the main generator likes to spontaneously load up much harder than is called for at a given throttle setting. Naturally the issue never wants to occur when I have meter leads attached to the control system. Thus, I find myself in need of an inexpensive data acquisition system to autonomously record multiple simultaneous voltage readings within the excitation system to try to get a better bead on what exactly is causing the problem.

I found the SparkFun power meter ( SparkFun Power Meter - ACS37800 (Qwiic) - SparkFun Electronics ) which in combination with one of the SparkFun code-free data logging boards seems like a viable DIY solution that won’t cost thousands upon thousands of dollars out of my own personal budget for a tool I will likely only need once.

The first issue I spotted right away is that the power meter appears to be rated for a maximum of 60 volts DC on the product page. My application is a 0-72 volt analog signal that is derived from a 74 volt locomotive battery system. However after digging into the schematics, reading the IC datasheet and plugging some values into the chipmaker’s formulae, it appears as though the RSENSE on that board was mistakenly sized at half the resistance SparkFun intended, allowing for measurements as high as 120 volts DC without exceeding the 250 mV input ratings on the IC. If someone could check my math and confirm this I would appreciate it.

Secondly, I am suffering from information overload when it comes to all the available options for data acquisition boards SparkFun offers. I am not sure which one would be the most appropriate choice for my needs. This project will likely be assembled into a plastic project box and powered off of a few AA batteries passed through a linear voltage regulator. Data will be reviewed at the end of the day when I am able to retrieve the unit and load the recorded data into my PC. I have had very little success reading or writing software code in my lifetime so I would like to avoid that entirely if I can. Could someone please share some insight as to the pros and cons of each datalogger on offer and offer suggestions as to which one might be best suited to my needs?

An extra feature that would be handy is some way to record four separate 74 volt digital inputs so I can correlate the throttle handle position against the analog power reference & feedback signal measurements I intend to take. (The engine governor is controlled by four separate solenoids actuated in a coded sequence by the throttle handle.)

(P.S.: This kit will only be installed and operational when I am in the cab to directly supervise it’s operation and intervene if something goes wrong. I intend to include basic safety features like DC-rated fuses, MOVs and snubber capacitors to protect both the locomotive and the kit.)

Thanks!

https://www.youtube.com/watch?v=cfFxzkNDvyI

[PaulZC]

Hi @HeavyIron ,

Welcome!

The ACS37800 is designed to measure high load currents in power systems. It may not be the best choice or may be overkill if you only want to measure voltage.

For the voltage measurement, it depends if you are using GND or LO as the reference. With GND as the reference, a RSENSE of 8.2K delivers 245mV with an input of 60V: 60 * 8200 / 2008200.

The LO pin and extra 2 x 1M resistors are only needed if you need additional isolation:

Maybe, for your application, you do need the extra isolation the LO pin would give you? I’m not sure… If you do use the LO pin as your ground reference, then, yes, you could measure voltages up to 120VDC. But the SparkFun product is not designed to used above 60VDC. Just to be clear:

It is not clear to me if you are measuring true DC, or actually rectified AC? Half wave rectified, or full wave? That said, I’m wondering if a simple 4-channel ADC could work better in your application? Would you be happy adding your own voltage divider resistors? If you are, then you could use a single ADS1015 ADC plus a DataLogger to log four voltages simultaneously. But please be careful with your grounds. It is not clear to me if it would be safe to have all four voltages plus the DataLogger sharing a common ground. Maybe you need optical isolation?

I will include a link to a similar discussion about voltage measurement below. I hope it helps.

Best wishes,
Paul

How to monitor the health of a relay and it's contacts

Hi, I have a challenge and hope someone has and idea how to solve it. I want to monitor the health of a relay by monitoring the state of relay contacts in relation with the state of the coil. I want to see if the contacts are open or closed when the coil is energized or not energized. I want to do this with an Arduino. It must be something like you see in the drawing below. Can you tell me if there is a component or sensor that can do this. Let me know what idea you have how this can be done. Thanks [image]

[brow]

How long do you expect each test session to take? Would a multichan oscilloscope store/capture enough?

[HeavyIron]

Isolation is indeed important for this job because not every negative wire within the locomotive that I might use for reference is at the same potential. There are places in a locomotive where one encounters things like battery charging resistors and sensor shunts for reverse current relays and so forth that create odd voltage steps between some of the negative wires. Often on the order of 1-2 volts.

In addition, I would like to be able to measure the voltage across the traction motor ammeter shunt in order to record how hard the main generator is loading. The problem with that is; while the shunt itself only develops a potential on the order of 50-75 mV, both of the terminals on that shunt are part of a 1200 volt, 4000 amp DC electrical system and could very well be sitting that high above ground when the unit is making 3000 horsepower.

An ADC with opto-isolation would help a lot in both of those regards, especially if SparkFun makes an ADC board with multiple inputs to help keep costs down. The key would be making damn sure everything from the optocoupler onward is good for 1200 volt service. The plus side is that the high voltage system on this locomotive is ungrounded; It has an incidental ground reference by way of the operating coil of a ground fault detection relay, but no deliberate low impedance ground bond. So if something does fault out the worst that will happen is a small zap delivered through the relatively high impedance windings of the ground relay, immediately before said relay trips.

As for recording time, I would like to be able to obtain at least 8 continuous hours of data at a rate of roughly 4 samples per second. Sometimes the issue shows up right away in the mornings, while other times the unit plays nice well into the afternoon. Then right as you’re creeping up to make a joint with a cut of cars the thing decides to surge from 250 traction motor amps straight up to 550 and you find yourself stomping on the brakes real quick to keep it from taking off. It’s a very unpredictable issue and a very annoying one at that.

I’ve marked down some of the spots I’d like to obtain voltage readings. If I can obtain all of them simultaneously by ganging up several 4 or 8 channel ADCs inexpensively then that would be great. (Traction motor current shunt not shown on this illustration.)

One measurement I have not marked down is the current signal between the Sensor and the Sensor Bypass Panel. The Sensor is essentially a magnetic trigger amplifier that governs the main generator field SCRs by telling them when to turn on. It receives it’s control input from the Sensor Bypass Panel in the form of a current signal that is passed through it’s control winding. More current causes the magnetic core of the trigger amplifier to saturate sooner, which causes the SCRs to turn on sooner, which sends more juice to the generator field. That might be a good opportunity to leverage the current measuring capability of the Power Meter board, yeah?

As for how filtered the generator power feedback signal is, I am not entirely sure. I would imagine there is a filter cap or some such that is not shown in this simplified block diagram but I haven’t found a print for that circuit yet to confirm or debunk that suspicion.

[PaulZC]

Hi @HeavyIron ,

Thanks for the update.

To be honest, I think attempting to use Qwiic electronics to study the locomotive voltages will almost certainly end in smoke and tears… Something will go wrong and you will end up frying your electronics.

Things have moved on from when I last looked. You can now buy a low cost DMM with built-in Bluetooth (BLE). I’m in the UK and found the Brymen BM788BT mentioned in a few places. The “BT” indicates it has Bluetooth. It looks like you can pair multiple meters with your phone and log measurements using their “IoMBTC” App. It looks like the App is available for both Android and iOS. The screenshots in the App Store suggest you can have at least two meters connected simultaneously.

I honestly think this would be the best solution for you.

I don’t know if you will be able to get the BM788BT in the US. But there must be an equivalent. Perhaps the good folks here (@brow @yellowdog @ts-russell) can suggest an alternative?

Best wishes,
Paul

[brow]

I’m not comfortable in a 1200 VDC environment myself and decidedly uncomfortable giving advice for a DIY instrumentation project one.

[rftop]

Monitoring a 2+ Megawatt floating (ungrounded) system protected by a Ground Relay makes an extremely interesting forum thread.

I spent a while talking to an AI chatbot about this system in general, and finally came to this summary: A 3,000 HP, 2-stroke V16 diesel engine driving a dual 3-phase AC alternator, which is internally rectified to provide an infinitely adjustable 0V to 1,200V DC bus to power six series-wound traction motors.

Every part of that summary is surprising & interesting to me.

[HeavyIron]

Yep, that’s a pretty good summary. This is an SD40. (Not to be confused with the later SD40-2 which has a vastly different electrical cabinet.)

The SD40 has a 16 cylinder “uniflow” type two stroke diesel engine that is forced-air scavenged by a clutch-driven turbocharger. It makes 3,000 horsepower and has an adhesion limit of around 90,000-115,000 pounds-force. It develops about 80,000-85,000 pounds-force at it’s minimum continuous drag speed of 13 miles per hour.

The clutched turbocharger in question.
The rest of the engine in question.

The prime mover drives an AR10 traction alternator with built-in rectifier banks, the mechanically contiguous D14 companion alternator, a 15 kW auxiliary DC generator and a big mechanically driven traction motor blower.

The AR10 in the SD40 as originally furnished would have supplied up to 1,000 volts DC to the traction motors, with a current limit of 4,000 amperes. This particular unit has been retrofitted with a more modern revision of the AR10 that is good for 1,200 volts DC, which eliminated the need for a complex sequence of traction motor field shunting contactors and resistors that were originally furnished to reduce back-EMF at higher speeds. Now it only needs to perform an electrical transition from 3 series, 2 parallel combination to 2 series, 3 parallel combination at 21~23 MPH.

The D14 companion alternator has fixed excitation and provides variable frequency & voltage 3 phase AC at up to 280 volts and 120 Hz for the operation of the main generator excitation system, AC feedback from various saturable-core transductor reactors and three 45 horsepower radiator fans.

The auxiliary generator provides 74 volts DC to charge the batteries and operate all of the controls on the locomotive.

Of interest is the fact that AR10 excitation is regulated by a constant-power engine governor via a hydraulically driven rheostat. The engine governor, in addition to maintaining a constant engine speed for a given throttle position, also maintains a constant fuel injector rack position by providing feedback to the excitation system via this load regulator rheostat. Ergo, as locomotive speed, traction motor RPMs and consequently back-EMF increase, the load regulator causes more field voltage to be delivered to the AR10. This in turn causes more voltage to be developed at the armature, up to the 1,200 volt limit. When a series of current and voltage sensitive transductors reach sufficient saturation, relays pick up to effect a transition of traction motor configuration from series-combination to parallel-combination, effectively reducing back-EMF and traction voltage to 2/3 of it’s previous value.

An example of transition occurring

Note that the current shown by the ammeter does not rise very much after transition occurs. This is because the traction motor ammeter is connected to a single shunt in series with the #2 traction motor armature, not the AR10 alternator. That shunt does not see much change in current because the voltage and current changes occur upstream at the AR10, not downstream at the individual traction motors.

This is what the inside of the AR10-D14 looks like:

It is a unit assembly consisting of two mechanically contiguous but electrically separate rotating-field (i.e. static armature) type alternators. Note all the small jumper wires within. The AR10 is constructed as a three phase delta-wound machine with 10 parallel armature windings per phase. Each winding terminates on it’s own individual fast-acting striker fuse and rectifier diode. Two sets of field slip rings are present: One pair for the AR10 and one pair for the D14. (Note the two different pairs of field jumpers at the end of the slip ring.)

As for the hazards present with such an electrical system, I am a licensed electrician by trade. I am no stranger to the hazards of working in, on and around big 277/480 volt, 4,160 volt and 13,800 volt motors, circuits and switchgear. This is just another slice of life for me. As it happens, the available incident energy in this sort of generator type system is likely considerably lower than that of the 1~2.5 MVA indoor substations I more often find myself downstream of. The output voltage of a generator will tend to collapse rather quickly and dramatically under fault conditions compared to a large, low-impedance transformer being backed by the full faith and credit of the entire regional power grid. Not that I have any interest in testing that theory - I’m just saying that we’re probably talking about an 8-20 calorie hazard here compared to something often calculated in the range of 100-700 calories. And the fact that it is an ungrounded system serves to provide the “first fault free” so to speak as an extra layer of defense. …As long as a guy doesn’t somehow get caught going line to line that is.

Anywhoo that’s more than enough autism for one forum post. I’ve found some more troubleshooting advice buried deep in the book for this locomotive that I’m inclined to try first before I go through the trouble of building my own data logger. We’ll see if those steps serve to shed any additional light on this intermittent problem or not. Things I was afraid to try without the book’s advice, like disconnecting the load regulator under power to see if the AR10 output current drops to zero like it should.

…Oh hell, what’s one more train video for good measure?

A cab ride in an SD40-2

[rftop]

@HeavyIron , there are soooo many things in that post that blows my mind
Thank You for taking the time to write it !

I’m much more comfortable operating the switchgear in a 25kV Substation verses even looking hard at a 480V panel. I know that sounds nuts, but I understand the transmission side, not so much when it comes to “low voltage”. Too easy to get too close.

@HeavyIron

When a series of current and voltage sensitive transductors reach sufficient saturation, relays pick up to effect a transition of traction motor configuration from series-combination to parallel-combination, effectively reducing back-EMF and traction voltage to 2/3 of it’s previous value.

Thats amazing. If you are ever in the Southeast US, I’d love to sit and talk a spell.

[HeavyIron]

Yeah, 480 and 600 are about as nasty as it gets because those particular systems are usually the ones that have both the right amount of voltage and the right amount of available fault current to really eff a guy up. The same amount of power on say, a 13.8 kV system will still make plenty of flames and sparks but it will typically lack the tremendous amount of fault current that makes 480 and 600 volt faults so explosive and violent.

This is a perfect example of that.

The video starts out with a set of 480 volt transformer secondary taps shorting out quite explosively. Then around the 3:45 mark you can hear when the 12 kV primary side of the transformer becomes involved. Still plenty of heat but nowhere near as violent.

Frankly it’s for reasons like that I am more afraid of 277 stuff than I am of 4160 circuits. Shock hazards on medium voltage circuits are easy to avoid; Just don’t touch the damn thing and you’ll be perfectly fine. It’s the arc flash from high fault currents that will reach out and ruin your day if a stupid screw or something falls between a bus bar and a cabinet.

EDIT:

Another example.

^That one really shows how different the arc flash characteristic is between primary voltages and secondary voltages. Yes it struck an arc and burned but it didn’t explode with anywhere near the violence typical of a secondary circuit like what is seen in the first video.

[TS-Russell]

How interesting…the first video is titled 12,000 volts because our intuition assumes bigger number = more violent, but the scariest parts are from the 480v system

Throwing snow on it right after in the 2nd video made me laugh

Great shares!

[rftop]

On the locomotive subject…I do have 1 train story.

My dad was a heavy equipment diesel mechanic, so I was his helper as a young fella when not in school. One day we were working on a train engine. Turns out a little bit of rainwater had entered a cylinder from the exhaust system while parked. When we pulled the head for that cylinder, there was maybe a few tablespoons of water.

That’s all it took to completely lockup that train engine. Something that can pull 30+ Million Pounds was stopped by what appeared to be a couple tablespoons of water (to a kid anyway) on top of one piston. I think of that anytime I hear “incompressible fluid”.

[HeavyIron]

Yep, that’s why we open the test valves on all of the cylinders and roll them over with the fuel off to test for hydraulic lock before we fire them up. Great way to bend a rod. On the smaller 6 cylinder switchers we go as far as barring them over by hand because it doesn’t allow as much momentum to build up as the starter motors / starter generators. On the big 16 cylinder road motors you don’t have much choice though; They’re too big to bar by hand without a special ratcheting tool.

The nice thing about the EMDs is that the scavenging ports at the bottoms of the cylinders serve to help drain any accumulated water out. At least for whichever half of the cylinders are on their down stroke anyhow. Anything accumulated just dribbles out into the air box, which in turn gets emptied by periodically opening the air box drain valves.

[HeavyIron]

Just for giggles, what do you make of the Texas Instruments AMC3330 isolation amplifier as a candidate for paring up with the SparkFun ADS1219? Looks like it’s more or less a drop-in solution for differential isolation at common mode voltages of up to 1.2 kV and has a transient isolation rating of 6 kV. Could pair it up with a couple small 2 kV cartridge fuses, some medium-volt input resistors sized to protect against “oops, 1.2 kV differential” and some kind of differential clamp for schmafety to help harden it for a CAT IV application like this.

[PaulZC]

@HeavyIron & @rftop ,

Thank you both for sharing. I am really enjoying this thread.

@HeavyIron : Sure, with my mad scientist hat on, four AMC3330 plus a ADS1219 on a suitably designed PCB with the appropriate safety features and correctly rated connectors would make an interesting addition to our product range. It would need a suitable enclosure. We would sell maybe one a year, one a month if we’re lucky. But mainly it’s the liability thing. If we say something is designed to withstand common mode voltages of 1.2kV and then someone hurts themselves using it or sets fire to something, it ends up in court. So, sorry, no, we can’t help you there. Again I would steer you towards an off-the-shelf tested certified safe isolated meter with a Bluetooth interface so you can log the data on your phone.

All best wishes,
Paul

[HeavyIron]

I was thinking more along the lines of having a handful of boards featuring that IC printed by some rapid prototyping fab. And only two of those ICs would be able to pair with a single ADS1219 board because TI’s design docs say that their IC requires a differential ADC input to avoid instability.

The problem with using a Bluetooth interface - or anything else involving a cell phone for that matter - is that phones are a big no-no on the railroad. (For good reason.) Any kind of personal electronics that could potentially serve as a distraction are expressly forbidden. That is why I’ve been pursuing an autonomous data logger; It is very similar to the “black box” event recorders that locomotives are already required to be fitted with and would feature no user interface that could provide any sort of distraction. It is my intent to align the nature of this project with the spirit of that personal electronics rule as best I can; To keep it firmly rooted in the definitional territory of a diagnostic tool, not a personal electronic device.

[rftop]

@HeavyIron , I have a few questions that are more about my curiosity… and not in the least a suggestion :

Wouldn’t we consider the likely source of the intermittent issue to be one of these (3) moving parts : Throttle Position, Load Regulator, or Governor ?

And Is there a single failure mode that would cause more than 72V (+ battery charging, so 82V) on the control system? I’m trying to think of a way (besides Wiring Arc Flashes, etc) that would involve the Amps or Voltage from the Traction Motor System to reach these points in your diagram. Again, I’m just naturally curious.

Measuring 82VDC across the Throttle Position System’s potential “seems” safe for a datalogger…but this is obviously outside my wheelhouse.

[HeavyIron]

Those are valid points. Let’s start with a functional description of all the parts involved.

The excitation system on the SD40 begins with the 72 volt stable reference from the static voltage regulator. This reference signal is passed through the throttle response panel. The TRP is a rheostat of the tap-changer type. Four relays within - corresponding to the four governor solenoids controlled by the throttle - operate to select a resistance value that is inversely proportional to the throttle setting. Higher throttle setting, lower rheostat resistance.

The current signal from the throttle response panel then gets passed down to the rate control panel. The rate control panel is in essence a glorified damper built around an R-C series circuit. It’s job is to smooth out changes in throttle position to avoid harsh step changes in tractive effort that would otherwise result.

From the RCP, the signal next enters the high side of the load regulator rheostat. This rheostat - despite it’s name - is in fact connected as a potentiometer. It provides engine load feedback from the governor by modulating the strength of the reference signal we’ve thus far obtained. When the throttle handle is placed in idle, the load regulator is held in the the maximum field position by the governor. It remains there until enough back-EMF is generated by the traction motors to raise the AR10 voltage - and consequently the AR10’s power output and engine load. (No traction motor RPMs means very little traction motor voltage which means very little engine load despite the motors passing several hundred amps. It’s pure torque without any meaningful horsepower except for I^2R losses until they start to spin.)

The voltage signal from the load regulator then arrives at the Sensor bypass panel. The SBP is essentially a terminal board fitted with an NPN transistor. It is here that the reference signal that we have been conditioning so far is compared against a power feedback signal from the AR10 main generator. The reference signal lands on the base of the NPN transistor, the feedback lands on the emitter and the control winding in the Sensor (fed from the 72 volt regulator) lands on the collector to create a low-side error amplifier.

So yes, the question is what components or connections could cause an intermittent surge in excitation.

So far I’ve already tried swapping out the rate control panel to no avail. Some other things I want to try include ohming out the load regulator rheostat, checking for bad negatives and disconnecting the load regulator from the sensor bypass panel. One of the symptoms I’ve noticed that is consistent is a high voltage value between the throttle response panel and the rate control panel in notches 1-3. Above notch 3 it falls back within published specs.

One of my suspicions is that the NPN transistor on the sensor bypass panel might be of the old germanium type, which are notorious for becoming leaky with age. The metal-cased ones in particular are known for growing internal tin whiskers that can cause short circuits.

As for the 1200 volt common-mode hazard, that stems from the #2 traction motor shunt that I’d like to measure, which is part of the high voltage system downstream of the AR10. It is not shown on any of the prints I have shared thus far. Alternatively I could do what EMD did and measure AR10 load vicariously by way of it’s field current. The book notes that AR10 field current is nominally proportional to AR10 armature current. The problem is there is a large distortion at the very bottom end of the field-armature current curve - which inconveniently is exactly where I need to take my measurements. The shunt I am trying to measure is visible at the bottom of this photograph:

(Also visible above it are some of the wheel slip detection transductors in case you’re curious what a transductor reactor looks like.)

And here’s what the front of the No. 1 cabinet looks like if you were wondering. (Some of the power contactors are hidden beneath the raised cab floor - you can just see the tops of some of the lugs in this photo.)

[just_Tony]

I recently instrumented the starter system on my motorcycle to learn as much as possible about the voltage and current dynamics about the power flow during cranking and recovery. Obviously the physical magnitudes here are at a micro scale compared to yours!

I started with a do-it-all concept for measurement and logging via the Sparkfun Openlog Artemis. My current sensing resistor was a 1 milliOhm shunt at the negative terminal on the battery, and a simple voltage divider was sufficient for scaling the battery positive terminal down to what the Artemis could accept even including my guesses for possible inductive spikes.

Note that the Artemis can safely accept voltages only at or above system ground. In the current measuring channel the direction reversal during battery recharge zapped the Artemis on my first try. This convinced me to select components that were very good at just one thing. My next phase was to pair Qwiic ADC modules to an Arduino R3, with better front end protections added.

I cribbed an application circuit example out of the data sheet for the ADS1015. It uses an OpAmp in a voltage summing arrangement in order to add a voltage shift to an input signal which I then feed into differential inputs of a second OpAmp for gain scaling. I could now dial in any offset (with a pot) and amplification (with precision fixed resistors) that was needed in order to fit the differential input range of the following ADC. Instability in the pot causes no harm because it’s common to both sides of the differential. My final measuring ranges were -900 to +100 A and -2 to +30 V. The OpAmps in the example are available from DigiKey only in surface mount but adapter cards for through hole are easy to find in Amazon and eBay.

HOT TIP: Adafruit sells bidirectional I2C isolators which are rated at 2500V that include Qwiic connectors. Cascade them if you need more isolation, which I didn’t need to do. One side receives power from the logger and on the other I have a 4 cell AA pack feeding a low dropout 3.3V regulator. One HUGE and unexpected benefit of the isolator is that it also blocked digital hash from the microcontroller that I had previously assumed was noise inherent to the vehicle. I use 12 and 24 bit ADCs from Sparkfun in Qwiic form factors, which before the isolation had been limited to only about 8 bits of effective precision. Each ADC module has it’s own isolator, since I initially had crosstalk between my voltage and current channels via the voltage offset I was applying only to the current measuring path.

Qwiic modules provide a ton of simplification on the coding side but the libraries push that complexity behind a curtain that unfortunately still needs to fit inside the microcontroller. My Arduino R3 ran out of memory space early during development. I switched to the Teensy 4.0 which I soldered onto Sparkfun’s Teensy Arduino Shield Adapter. They say “This adapter is not compatible with the Teensy 4.0 due to the differences in the pin layout” but I think there is >= 99% compatibility in reality depending on the application. The potential for incompatibility comes from niche applications that use pin assignments specific to the 4 that you won’t need. In my application I have perfect compatibility. I’ve reported that back to SparkFun but that statement is still there in the product overview.

The Teensy has about 10X more code space than I need and it’s high triple digit MHz clocking allowed me to use the highest available sampling rates on my ADCs.

I use an RTC + SD card Arduino shield for the recording task.

HOT TIP: SD card speed specifications are not directly relevant to what I will call low rate, small data block logging reality. Their memory controllers are often rated at huge blocks of data that stream in nearly continuously. But my logging occurs in tiny transactions of 25 to 45 bytes at a time. Probably the file open and close overhead was the issue. The bottom line is that slow cards of 32 GB or less ended up having fewer dropouts than the fancier cards. Go figure. I still get logging dropouts of up to 180 milliseconds that would occur at intervals of around 10 seconds, which I decided I would tolerate. I was never successful at employing the provided buffering example sketch, my bad.

Speaking of data rates, I had originally interpreted some signal hash as being analog noise in the motorcycle starter subsystem, back when I was logging at 20 samples per second. I also saw what I thought were instabilities or ringing in the current which I thought could have been inductive effects from the started motor. At 100 Hz I started to see aliased structure in the hash. My imagined “ringing instabilities” turned out to be the changing torque loads from compression strokes. But still the max current draw peaks contained only a single sampling point, therefore undersampling.

At 1000 Hz I was collecting redundantly sampled peaks and valleys on the voltage regulator output, and I could then reliably integrate battery drain and recovery curves whose S/N ratio was about 1:1 during idling. A common refrain from MSEE’s I’ve worked with is “More data is better”.

A question for you: How simultaneous does your logging truly need to be? Maybe you don’t really know until you’ve learned some nuances of what your system does (refer to my prior 2 paragraphs). My system was sequentially sampled l because one of my ADCs does not have a trigger control. I made my system pseudo-simultaneous via high rate sampling in which my voltage and current measurements were interleaved. You might want to consider using broadcasted sampling triggers that you then read out from each of the ADCs “at your leisure” between samplings. That can drive your ADC shopping.

[HeavyIron]

Yep, choo-choos demand big cranking power. On the order of 40-80 horsepower. The older units used starter-generators. The newer ones (on the EMD side of things anyhow) use more familiar starter motors and ring gears. (Two of them.) A lot of them now use air starters. GEs still use starter-generators. Always have.

The inrush is on the order of kiloamps. Steady-state is somewhere around 600~1200 A depending on battery health and engine displacement. Big 32 cell, 450 AH diesel starting batteries. About 3500 pounds of lead and acid. Sometimes a bunch of 8 volt modules, sometimes a pair of big 32 volt monoblock chunguses.

As for sample resolution, I only need about 4 Hz to be able to see what the excitation system is doing. I’m looking for trends and relationships, not transients. Plus I need about 10 hours of recording time.