Remote Control Audio for Rolling Stock

Well, it's going to be considerably more involved.  I think I have a handle on the serial interface to the MP3 boards, that was a sticking point.  Once I've tested that sufficiently, I'll proceed on laying out a modification of my board.  I have to go to the 20 pin PIC to get all the functionality I desire, and my debug board only handles up to 14 pins.  I suspect I'll just try to get as close to the final design as possible and run a set of boards and develop on that. 

This is the configuration that I hope will become the basis for an actual product.  Yet to come is what it will actually cost to deliver this, that may be another sticking point.  Of course, with the price of Lionel's cars, maybe not.

I'd say the simplest is a PNP transistor switch which should get you well under 1 uA of leakage current.  I can see your uC pins are ever so precious but I'll just comment that what I generally do when there's a uC involved is to use an open-collector or open-drain output pin to shut off the battery switch.

So the uC output would drive point A in the circuit above and the uC output would replace Q2.  In doing so, the power to the uC is cut off and the uC will go into reset or brown-out and that output pin will continue to float.

OK, I realize you don't have spare outputs so how about a simple RC circuit to drive the NPN.  Whenever there's track voltage, D1 charges the cap.  So this would be driven by some point in the circuit that has + voltage relative to circuit ground.  35V cap since I'm assuming you would pick off something close to track voltage peak track voltage magnitude.  When track voltage is lost, the RC discharges keeping the PNP switch active for a couple seconds and everything turns off.  I threw in some placeholder starting values.  It depends on how much current you need to run thru the PNP.   In any event, maybe 25 cents or so in parts.  Yes, you can go with FETs instead of bipolars but I like cheap.

This basic circuit is just a starting point.  It does not address the scenario where rectified track voltage is just under the battery voltage.  In this case, the battery would be unnecessarily called into service (drained).  Alas, more complexity and additional components to inhibit the switch. 

Separately, it's your circus (and your monkeys) but I don't think you want the battery powering "all" 5V loads on the board.  A 5V servo can draw quite a bit a current...much more than the 100mA or whatever you say you're drawing on audio.

Stan, I was thinking along the lines of a simple RC circuit and a transistor or FET switch.  I don't want to use a uP pin for this function, that seems a waste.  Also, if I use a uP pin, I'd have to have track power monitoring in the uP to know when the track power dies, since the battery is continuing to power the uP, I wouldn't know the track power was gone without some sensing.

The battery is on the input side of the switching power supply module that provides the main power, so it actually has slightly less current draw than on the 5V side. 

As far as powering the servos, remember I want the battery to cut off after a few seconds, it's just to carry across conventional mode switching or large track interruptions.  The small servos I'm using actually draw very little power, they have a motor that's the diameter of a standard wooden pencil, and about 3/8" long.  They're just heavily geared, but they don't draw much power.  I tested the two sizes I have, the big one draws close to 100ma in continuous operation at 5V, the small one around 50ma in continuous operation.  I can't see this being an issue for 10 seconds or less.  Also, it's probably pretty rare that they'd be running at the brief interval that we'd be on battery power.  I don't see the servos being in continuous use in this application.  They're model airplane servos, FWIW. 

I think I'd ignore the scenario where track power was just below battery cutoff to start, that gets trickier.  Another thing is I'm trying to keep the board real estate down for circuitry, not to mention parts count.  More parts, costs more for each board for assembly.  Track power would be down in the 7 volt range to drop the rectified DC below the battery trigger point, I'll just state that you can't run that slow.

I am thinking on going to the 20 pin version of the part I'm using for the "enhanced" version.  It simplifies things if I don't have to share the debug interface pins, that always gets tricky if you want to test that part of the circuit while debugging.  That still gives me 15 pins for I/O, I'll run out of board area before I run out of functions to do with the pins.

John,

Your voice sounded completely different than I anticipated-I expected more like Santa's elves since you bring us all this cool stuff no one else can make  

You the man.

Probably just nobody else is nuts enough to do this stuff, I'm sure others here could make it if they wanted to.

If you're considering a 20-pin version, why not a 28?  Assuming you have these professionally assembled should be no penalty to put a 28-SS pkg vs. a 20-SO pkg.  You'd have to fuss with the hand-built prototypes of course but should free up the debug interface.  I think the conventional battery circuit is still better done with the uC (yes, requiring a track-voltage-presence input pin and a battery-enable output pin).  The circuit I showed earlier suffers from a modest bias current which draws on the battery when track voltage is present.  This would vanish if the uC only turned on the battery when it needs to.  The incremental cost of a larger uC part is probably a wash with the components and real-estate to implement the discrete circuit "logic" to implement a zero bias current battery switch when track-voltage is available. It would also handle that no-man's land situation of the battery powering the entire system at low track voltage.

OTOH if you've come up with a low-cost RC circuit to switch the battery I'd like to see it!

One problem with any circuit that directly connects to track voltage is the full wave rectifier that supplies all the power.  I have no common ground, so any sensing of track power will have to be optically isolated.  The good part is the emitter is powered from track voltage, so the power consumption isn't an issue, the output transistor would simply open when track voltage was gone.  I just have to come up with a simple delay before the battery is disconnected.

My thinking for conventional operation, or any battery backup, is to just provide a connector for the battery backup circuit.  I can't see saddling every user with the extra costs and board real estate of the battery backup circuit. 

I am making provisions for 1100uf of bulk capacitance to better ride out brief power interruptions.  I will say that I didn't have any issue on the club layout with many switches with the existing 330uf, including a number back-to-back.  I figure two 560uf caps should be good protection.  If I decide one will do, I can simply leave one off the assembly.

I've currently mapped out a 20 pin device and have most of the circuit drawn.  It allows for two PWM servo connections, three LED's, and one output FET with a diode for higher DC current, similar to what I do for the cab light control on the Super-Chuffer.  I have two pins left, and I'm thinking of optional inputs.  Certainly, one could be for sensing the car was moving with a wheel sensor.  Perhaps I could have alternate sound tracks, one is selected when moving, the other one is selected when stopped.

On the positive side, I got the serial interface working to the MP3 modules, turns out it's enabled in the default configuration.  I used my serial protocol analyzer to fire off commands to play various clips.  Other than it missing the first one after a power-up, it went great.  I think the analyzer doesn't give it time after the interface is switched on, but I figured that was a minor issue to be dealt with later.

I'll cease and desist on this battery backup thing...after one more comment.  Seems you could skip the (cost of) optical isolation and detect the presence of track voltage using a diode and driving a transistor - plus an R and a C to smooth the half-wave.  Presumably you have a bridge somewhere in your circuit so the track voltage is referenced (not floating) to your local circuit common.

More board space but suggest these inputs have 3 pins...2 to provide, say, +5V and GND to the sensor.  There are so many of these tiny $1 motion, vibration, optical, humidity(!), etc. eBay modules for the Arduino market.  Many are 3-pins, 2 for power, 1 for the sensor output.

I'm not understanding the lack of isolation, I actually do not have a frame ground reference to the internal electronics, with a full-wave bridge from the track voltage, there's no easy way to do that.  This is the same issue that the RS5 boards and the MTH DCS electronics has, no common to the frame, and it's for the same reason.  The Legacy boards also don't have a common frame ground.  Lionel uses an opto-isolated serial input to receive serial data from TMCC as well.  I don't see an easy way to frame ground reference after the bridge rectifier for the power, unless I'm missing something.  I could go with a simple diode, but that would require much more filtering and higher voltages before drop-out.

The three pins for the inputs aren't a problem, though I may have to go with finer pitch connectors.  One significant cost driver is how to get cables & connectors made for a reasonable price, the quotes I get are in the stratosphere!  In small quantities I can make them myself, but if I have several hundred connectors to crank out, I really don't want to be the one to do that!  Right now the board is 2.0" x 1.1" and there's probably space for some small connectors for the input pins. 

I did provide the standard .1" spaced three-pin connectors for the servos, power, ground, and PWM signal.  They're powered from the supply that feeds the MP3 module, and I did beef up the supply for the receiver and uP a bit, it's a 150ma SMT part that has heat-sinking to the board.  The To-92 was working, but it was running at 55C, so I figured I should add a little "heft".

Here's the current trial layout.  Note that it doesn't include the connections for the input pins.

Maybe we're talking apples-oranges? 

Say you put a scope's common on your circuit ground (i.e., the "-" pin of the bridge).  If you probe track voltage, it should be a track-voltage magnitude half-wave signal on one half-cycle and a clamped diode drop on the other half-cycle.  If the objective is "just" to sense track voltage is present, then this 60 Hz pulsing, half-wave signal that can be 10-20V or more could drive a grounded transistor.  The current would be in the microamps for basic on/off sensing so a small cap could smooth the ripple if you need a steady on-off.

I'm not familiar with TMCC board isolation issues.  That said, I can imagine situations where you have two separate circuits each with bridge rectifiers powered by a common AC track voltage.  Depending on current and such, the two local circuit grounds will be moving around relative to each other.  In such cases, I can imagine using an optical isolator to send low-voltage, high-speed signals between the two.  The two circuits are not truly floating/isolated in the galvanic transformer isolation sense.

But in this case we're talking a slow-speed detection problem with a huge available signal level that is referenced to local ground.  Again, we may not be talking about the same thing?!

OK, that makes sense, I truthfully didn't think of it that way.  We are looking for a pretty simple detection.  I still don't like it depleting the battery as we run, perhaps a FET would give lower leakage.  I don't visualize the issue of the threshold where we are floating between to be a serious issue, that should be a pretty narrow range.  If the circuit is still sensing track voltage and holding off the battery, maybe that solves it?

I haven't really given the battery too much thought yet, though I can see it would be a nice feature to have.

Again, I'm not familiar with how Lionel does their wiring but I notice how MTH switched from multiple connectors on their PS2 boards to one big fine-pitch connector on PS3 that has all the signals, track-power, etc.  It seems to me that the level of DIY skill needed to apply this board is considerably different than the LED lighting kit.  I'd say the noble idea of no-solder, plug-and-play, can't-go-wrong, etc. is not as relevant.  That said, dealing with fine-pitch pre-crimped wires that a user has to pop in/out of the housing to configure a particular combination of lights, servos, sensors, etc. may be OK.  Just my opinion of course.

Oh, and if your regulator is that SO-89 in pink why not add some more heat-sink copper around it?  Now I also have to throw in my sucker-punch about fattening the traces of your power leads.  Fatter traces on that regulator will draw heat from the package via the pins.  So it only lowers the device a degree or two - it's free.

I hear you on this.  And the board is getting somewhat busy!  Anyway, I'm not sure you can break the circuit and do this with only 2-pins to the external keep-alive circuit/battery.  But if you use a higher-density connector you could allocate 3 or whatever pins to this external/optional circuit. 

And here I said I would stop harping on the battery circuit.  Well, I think that narrow voltage range is not as narrow as one might think.  In conventional, I think it quite common to lower track voltage to slow then stop and just sit there for a (long) while.  So your 1100uF or whatever will only be charged to 10V or whatever.  If Direction button is now pressed, is there enough capacitance?  If the engine is slowly accelerated and hits a patch of dirty track is there enough capacitance?

What's also tricky is the accuracy of detector switch.  Under reasonable load at low track voltage the rectified voltage will probably have 120 Hz ripple.  I can imagine a situation where the battery is called into service at 120 Hz on each valley if using "just" simple diode or'ing to allow whichever source (capacitor or battery) has more voltage.

There are so many dag nab scenarios with AC conventional that this is not trivial!  Yes, a FET would lower the bias current at a cost, but it may be time to re-group and demand a specification from the marketing department!  Another solution is to kick the can down the road, pony up pins on the uC, and deal with it in software later; yikes, I can't believe I just said that.

I did actually fatten the ground traces slightly, the trace widths are all .01", and the ground I upped to .012".  I was thinking of putting a copper pour around the regulator as well in the final product.  If I do that, I'll try to isolate it apart from the other parts to give it a larger heatsink area.  Of course, this layout is all tentative, I just do these to make sure all the parts fit and I can get a routable board design.  Many times it just takes moving a part a bit and the routes fall in place.  It's really to see if I have to up-size the board.  So far, there seems to be plenty of space.  Actually, if I shrink the area for the connectors a bit, I could fit it on a slightly smaller board.  However, the two rather large 560uf capacitors take up a lot of board space, and of course the MP3 module takes up close to half.  I do put stuff under the MP3 module, but I can't stick the tall stuff under there.

I got so many complaints about soldering the 2.54mm stuff on the Super-Chuffer, I didn't want to have the same issue with this board.  OTOH, with the cost of connectors and the issues of getting them fabricated, that's still a viable option.  I suppose I could offer an option where all the wires were soldered and the customer cuts them to suit for extra cost.  I could probably do a jumper field of 2mm spaced holes for the external connections.

I have the can laying in the road, but if I could come up with a reasonable solution, I might pick it back up.

Not True.  MTH still has larger pitch 8 pin connector for the heavy loads including track power.  In fact I had 4 connectors to remove on a recent Tender board replacement.  G

Maybe I'm thinking of the PS3 HO?  Seems one of the new versions has all wiring including track power coming thru one connector.  The task at hand is for GRJ to come up with a connector strategy that supports a wide variety of applications and configurations - some of which have not even been imagined yet.

I'm hoping I have no really "high current" stuff to live with.   My issue is trying to come up with a connection scheme that's easy and fairly cheap, I don't want to run up the cost for connectors.

It's just that "stocking" multiple connectors on the male and female sides seems a burden.  And if going with polarized connectors it can add space, and still doesn't solve the problem of someone plugging in a 3-pin cable into the wrong 3-pin socket...unless you have the luxury of providing different connector sizes for each connector type.  I don't know if there's a simple answer, but I'd lean toward the single-connector and assume that the user has the wherewithal to insert/remove the crimped wires into the correct positions.

I don't imagine any of the wires carrying over 1 Amp continuous.

Again, I see this product as being for the more advanced DIY enthusiast.  It will surely cost several times that of the LED controller and I'd think more than your Super Chuffer.  I imagine some users will have others install the board for them...sort of like the MTH PS2 upgrade kit where, it seems anyway, lots of guys have someone else do the install.  It seems simple enough to download and program a microSD card with multiple MP3 files...but then again is it really?! 

The idea of one connector for everything but the servo connection is appealing, maybe I'll use the 2mm size that Lionel uses.  I believe that you can get pre-crimped wires from Lionel, so that would work.  The reason I leaned to separate connections for the power and speaker is that the optional connections may not be used in many applications, so you wouldn't have to deal with the large connector.  I can put the LED's, optional outputs and inputs all on one connector.

For the servos, there is a standard 3-pin 2.54mm spaced header defined by a pretty universal standard, and all the servos come with that connector.  If you plug them in backwards, they just don't work, but it doesn't damage them, so there's little risk.  And I don't want to have to change the connectors on those to use them.

The really small 1.25mm connector isn't worrying me, as that's strictly for "factory" programming, it's the one that Lionel uses on much of the Legacy stuff, as well as the ERR RS5 & CC-lite boards.

I'd like to find a connector that's universally available.  One option is dual-row ribbon connectors, you can find those and pre-crimped ribbon cables pretty cheaply, that would be an option.  Also, I happen to have the real tools to crimp ribbon cables, so I could build those pretty easily.  For installation, you just peel the ribbon cable conductors down and cut to size.

DigiKey sells pre-crimped wires too.  They have 2mm, 1.25mm, etc. for some of the connector lines they carry.  You choose wire color, size, crimped one end or both, etc..  They've been doing this for years but I suppose they could stop at any time.  Here's an example of a 10" wire crimped both ends for the 1.25mm Hirose DF13 connector.  70 cents each in small qty...not cheap.  Since for this kind of project you only need a crimp on one end, I generally buy a longer wire crimped both ends and cut it in half.

http://www.digikey.com/product...T-10110-A8-ND/425389

As you say Stan, NOT CHEAP!  Even if I bought 100 of the 12" pre-crimped wires, that's 30 cents a wire.  I was hoping for a lot cheaper solution, though that may be a pipe dream.  By the time I buy the connector shells, a 14 pin connection is around $5 cost.  If I extend that to 1,000 unit pricing, I guess it comes down some to 21 cents a wire.  That's still $3 or so cost.

I was looking at stuff like these 14 pin IDC connectors for $30 for 200 of them, and matching 14 pin rainbow ribbon cable, 15ft for $10.24.

I could make 120 six inch connecting cables for $70, and have all 14 pins supplied.  This would allow including the connector with wire in the package, which I believe would be a big bonus.

Stan, how about something like this for the battery circuit.  I realize that it doesn't address the low track voltage question, I'm not sure what it would take to deal with that.

No joy. Even after the RC discharges, the body diode in the N-FET keeps the battery on via 2 diode drops.

I thought that looked too easy.

OH, well, back to the drawing board.

Now we are starting to see why the manufactures sell these car for $120 or go the cheaper route of a sound chip blob, like a Christmas car for $80.  G

Yep, and the latest Lionel sound car has no battery backup. 

Does anyone know what quality sound is in these $120 sound cars?  I'd be surprised if they are using MP3 technology.  While it's arguable whether you "need" MP3 quality sounds in a "toy" train, the fact is it really sounds good and you can create long, imaginative, and quite interesting sound effects.  I think earlier in this thread someone mentioned the Polar Express.  With the huge memory in any MP3 player module, it's no problem to put the entire narration of the Polar Express story.  Obviously no manufacturer could do this because of licensing or whatever, but with GRJ's module it would be easy for an end-user to do so for "personal use".

GRJ, so what's your thinking on the battery-backup issue at this point?

I understand your thinking on not wanting to burden the main uC with the battery-backup code and pin count.  I understand the idea of making this an option though I think that introduces logistical nuisances (not just a battery but a battery with a bunch of circuitry and a PCB board).  I still believe if you want to support conventional there should NOT be a restriction or an "oh by the way" that you don't want to operate at low track voltage in case the battery is unnecessarily drained.

But it is that narrow low-voltage range which is challenging to design a practical, low-parts-count, 25 cent solution.  For example, if you don't want to burden the main uC, I was considering one of those 5 or 6 pin uC chip for 50 cents or so with the sole purpose of managing the battery for conventional; this uC chip could reside on that optional board.

Your switching 5V regulator module operates down to 7V (according to datasheet) and presumably it "behaves" well below 7V - in the same way as a linear 7805 regulator that is.  A discrete circuit has to deal with a partially depleted 9V battery at 8V, 7V or whatever.  It's not rocket science, but not trivial without breaking the bank...

I'm leaning to maybe considering supercaps again.  I get away from any issues of depleting batteries or having to replace them. 

I did a little experiment.  I set the input to the module voltage to give me exactly 10 volts out of the bridge and series connected two 2.5F 5V caps across that output.  When I pull the power, I get 19 seconds of normal operation before it dies.

Obviously, that's not in itself a truly workable solution, but it demonstrates that the supercaps certainly have the energy to solve the problem.

I could probably do something kinda' absurd, but maybe not.  How about a 10V regulator charging the supercaps directly from track power  They are connected to the output of the bridge rectifier with the blocking diode, probably a Schottky diode to minimize the voltage drop.  When the voltage out of the bridge falls below the supercap charge, they will supply current.

A diode, cap, and the supercaps.  I'd probably have a small value resistor to  limit the inrush current since there's no fire to get them fully charged.  The trick is to keep the below their rated maximum voltage.

Well, if you're leaning toward supercap, I'd want a way to get by with just 1 5V part using the existing 5V power supply to charge it.  Then I'd take the $3-4 I just "saved" by using only 1 supercap and apply that to whatever power circuit to either step it up to 7+ volts.  Or see if the supercap can run the 5V supply itself down to 4V without resetting the system.  The volume will fade a bit and maybe the lights will dim a bit but slowly.

Given the price of a supercap, the "option" approach makes sense.  So if the 1100uF works OK for command, then (more than 2 wires) could go to this optional conventional board. I'm thinking even a stepup module ($1.50 qty 1 on eBay) could be soldered to this optional board to supply a full 10V (or whatever) into the 5V regulator.  Seems kludgy but combined volume (supercap+charging components+stepup module) ought to be about the same as a 9V battery.

Photo recycled from another but showing the tiny size of the stepup module.  Could almost soldering a supercap to the back of this module, plus a resistor for charge current limiting from 5V, plus a diode or two...and then shrink wrapping it.  Just thinking out loud here!

Perhaps apples-oranges but I make the observation how MTH went from 9V battery to 2.4V battery.  I'm pretty sure 2 x 1.2V batteries is cheaper than the 9V style.  So while there are obviously differences in the applications, space available, etc. in the 2.4V case a stepup regulator is used to generate the 5V needed for conventional operation.  And of course with the supercap-based battery replacements, the cost of the 9V style is substantially more than the 2.4V style as there are more of those spendy caps inside.

I actually ordered a couple of those modules to take a look at them, they should be here soon.  I think the connector for connecting the "battery" board makes sense, that way you have more mounting flexibility.

I figure if I have the unregulated positive, the regulated positive, and the DC ground, I can come up with something to hold up the power for a few seconds with a supercap. I'll have to draw this out and make sure it makes sense as far as current flow and current paths.

I wonder if I can get perpetual motion going here?  I have the cap charging from the supply that I've feeding from the cap?

I wonder if I can get perpetual motion going here?  I have the cap charging from the supply that I've feeding from the cap?

 

As we all know, the typical perpetual motion machine fails to account for an external power source feeding the beast...in this case the bridge.

Edit: diagram changed to show charging cutoff switch.

That looks good, pretty much what I envisioned.  I was thinking a slightly higher voltage out, say 10 volts, so I could take advantage of a bit more of the capacitor charge when the power goes out.  It only has to handle a few seconds, but with one of the caps, that might be close at 8 volts.  I wonder how low a voltage that switching board cuts out at, that could be a factor...

I looked at the module specs, it claims to work down to two volts.  That's pretty good, I retract my higher voltage comment.

The Lionel tended to follow the RS tech so 2.5, 4.0 and I would think the Legacy could use Legacy Lite if they want.  So as good as your engine.

Some of the MTH accessories with sounds have the same memory chip on a PS-2 board.  Sounds were very good.

Other just used a simple chip so lesser quality.

I understand where this is going, but at some point your quality is limited by your audio amp and speaker/enclosure regardless of MP-3 tech.

What is funny is most MTH PCB for accessories only run about $30.  They are larger but also drive motors, lights and sounds.

You would think they could create this too and allow you to load different sound files. similar to the engine.  Especially with the Flash 4Meg memory chips they are using.

But they are not doing it so it is up to the entrepreneur to fill the gap!  G

G, thank you for filling in some blanks and your insightful observations.  Please keep your comments coming! 

The challenge is where to draw the line on features, cost, etc.  It's one thing for GRJ to make something for his own needs but what's exciting is prospect of making this available to the community.

Well, I'm back to the drawing board.  Upon reflection, the now edited diagram needs a cutoff switch to turn off the perpetual motion machine when the supercap is discharging.  Otherwise it's like stepping on the accelerator and brake pedals at the same time.

Question.  Can you run the 5V circuitry at, say, 4.7V?  In other words a Schottky drop from the 5V regulator.  I'm pretty sure hobby servos run fine at 4.5V. 

I'm thinking of another way where voltage is or'd at the 5V level rather than the higher rectified voltage level.  The above diagram still has a quirky zone at low track voltages where the supercap is supplying power when it does not have to.

Stan, I know the MP3 board will run at lower voltages, and I'm pretty sure the servos will too.  The uP and receiver run off a separate regulator and so wouldn't be involved in the feedback loop.  Funny thing, I was looking at the battery circuit last night and thinking about the same issue.  The issue is trickier than it appears.

Doing the battery at the 5V level is complicated by the fact that I have two regulators to isolate the high current stuff from the receiver and uP.

I have a feeling that I'm going to have to live with the issue of the supercap switching in at low voltages, I don't know how I can make that go away.

I keep looking at this issue, and a relay is looking better and better.

Dear gentlemen

I have been following this post and some of this stuff is over my head. I do have one question for both of you. Why can't you reuse the MP3 player battery to solve the problem of voltage drops? A question from novice but learning a great deal from both of you gentlemen.