Remote Control Audio for Rolling Stock

Hi John

Looking forward to the updates on this project.

Wow! Great idea Gunrunner. I can visualize a boxcar with an open door, a figure setting and holding a guitar. The sound would be Jimmy Rogers singing Train Whistle Blues!

I think I know why the transmitter doesn't have decent range.  When the MP3 module is actually playing a song, it's stomping on the RF signal to the receiver!  It works great to start the first sound, but if I want to change the sound from the transmitter, I have to be pretty close to the receiver to get it to work.

Although I wasn't initially going that way, I think I may try mating an ERR MiniCommander ACC module to this sound board and have TMCC control directly.  I was hoping for better range from the RF solution, but there's more than one way to skin this cat.

 It would be really cool if I used one of my RC servo units and opened the door under remote control.

I would look at getting a unit if I could use it to open doors.  Also, what about using a reed switch and cam/magnate on an axle to trigger sounds, kinda like the VL sound reefers and tankers?  I would so love to add those kinda of sounds to my freight cars.

Interesting project John, keep us up to date!

Now you're launching into computer control and more complex audio effects.  In order to duplicate what the VL tankers have, you have to mix audio channels.  It's certainly possible, but not nearly as simple as triggering discrete sound clips to play.  I seriously doubt I'm going to launch into that kind of development effort here.

After getting a basic rig running, I do plan on introducing the processor to expand the capabilities, that will be the next step.  I want to get some experience with the sounds and build a few cool projects first.

The opening doors is not that difficult, but it will require the addition of the processor.  The model airplane servos I use are PWM controlled with one input.  You can have the door move at any reasonable speed just by controlling the PWM output.  Arranging it to open the door is pretty simple, and the servo has plenty of torque to do the job, if the door sticks, it's liable to pull it off the slides!

I shielded the MP3 player board with a tinfoil box, didn't seem to make much difference in the range problem.  That being said, I decided to go a different way and do it with TMCC anyway.

Here's the new interface, this will go between the ERR MiniCommander ACC board and the MP3 player module.  The opto isolators are not technically correct as TinyCAD didn't have any AC opto component, but you can get the idea. 

Basically, the four channels of the MiniCommander connect to the four optocouplers.  Since two of them are negative in respect to ground, this seems to be the easy way, and it's 100% bulletproof.  The output of the optocoupler is filtered  by the 22uf cap to eliminate the AC ripple from the input side.  The resultant signal is sent to the MP3 module to trigger one of four sound selections.  The MP3 module takes a connection to ground (or at least a logic zero), so when the output from the Opto is active, that's what I get.

If the breadboard works as I expect, I will probably immortalize this in a PCB so I can install them in more than one car.  Hand wiring a bunch of these ain't in the cards!

Edit: remove old schematic.

Well, there's a clue.  More than likely the module uses a Class-D switching amplifier so when driving audio to the speaker you have 5V pulses radiating EMI courtesy of the huge antenna of the long speaker coil winding.  I notice in your prototype photo that the speaker is conveniently placed next to the RF receiver to maximize the stomping.

So first move the speaker and wiring away from the receiver.  That might do something.  If you have one of those ferrite cores/toroids to wrap you own inductors, place it near the module output and wrap the speaker wires thru it.  If that shows improvement but not enough, google "class-d emi filter" or something like that and install an engineered L-C filter right on the module.

Yes, I'm imagining how easy the job is...but IMO the greater contribution to the hobby would be if this was not dependent on using an ERR module to gain remote control...which would leave out conventional, DCS, DCC, etc. users.  

Stan, I moved the speaker way away, and had the receiver and MP3 module 6" from each other.  I also shielded the MP3 module totally with a little tinfoil box, trying grounding the shield to a variety of places, negative, earth ground, etc.  No joy with anything I tried, the range didn't change much, if at all.  I think I've expended about all the effort on getting that RF solution to work.

While the TMCC solution isn't for everyone, part of this exercise is for a solution that I can use.  Being somewhat selfish, I'd like something that can work for me.  If a different module works better (I have several more coming), I may revisit the RF solution.

I ended up rethinking a bit of the logic for the MiniCommander to solve one vexing issue I had.

I have four outputs, so I can trigger four sound files.  One issue is that with the MP3 module, there is no easy way to stop it playing without using the serial computer interface.  Since I'm simulating push buttons, I can't stop the audio.

I figured a way, I put a 5th audio file that's just a short silent recording on the SD card and I can trigger that to stop the sound.  Of course, that brings up another problem, how do I get a 5th output from the MC?  I think I've solved that one with the following additions to my interface board.

Since the LC1 and LC2 outputs toggle on and off with discrete keys, they're normally used to turning on and then turning off the output for the sound you'd like to play as the MP3 module is edge triggered.  In order to trigger the 5th output, I turn on LC1 and LC2, this normally would just play the last one selected.  However, with the addition of a couple of gates, I generate a slightly delayed pulse to the 5th input to play the 5th song, which is the short silent one.  That results in the sound module being silenced.

I could have also just used three sound files and had the 4th be the silencing one, but I'd like to support as many different sounds as possible.

Yet to be determined is if I need a delay for the logic circuit to insure the 5th trigger really does the trick, but the logic seems right.

I edited my diagram in the previous post, I added a little delay to the trigger of the 5th sound to insure that it would be triggered after the one(s) connected to the two LC outputs when I did my "reset" by having both active.

So to be clear, you got 30' "reliably" for which functions / under what conditions?

I understand you back-burnered the direct RF remote for now, but in the context of an O-gauge layout what would you (or anyone else following along) consider to be decent range?

I like the door-opening idea as it fits with so many other rolling-stock accessories where you trigger it, the door opens, something happens, the door closes.  Is anyone else following along?  If there is really interest in mating some kind of mechanical action to the sounds, I've been messing with the idea of using a stereo MP3 module to put sound on one track and mechanical commands on the other track.  For example a series of audio bursts can be sent not to a speaker but to a 25 cent bridge-rectifier to create the DC pulses to drive a PWM hobby servos that GRJ mentions.  Maybe someone has done this already but that would be one way to synchronize sound and motion without requiring a processor chip and writing software.

Following along, yes.  Having anything to add that doesn't make things more complex then they need to be?  Not just yet, but I'm working on it.  

If the audio isn't running, when I key the first sound file, whatever it is, I can usually trigger it from 25-30 feet away in the next room.  However, once the audio is playing, 4-5 feet is about all I get.

When I get a chance, I downloaded a couple of detailed filter descriptions for Class D EMI filters, maybe that will make a difference.

That's an interesting thought with the stereo, the little module I'm testing with has stereo output.  You could really just record the PWM data stream on one channel and the audio on the other.  With a simple gate you could square up the PWM data from the tape and feed it directly into the model servo.  I don't know if you could use the on-board amp to play the audio channel, as it being mono, I assumed it was mixing the two channels.  That may be a false assumption.

Here's the pinout of the module I'm using right now.

Well, easy for me to say, but how about connecting an 8 ohm resistor (or whatever your speaker impedance) right at the module output so as to create an equivalent load.  Then when you start a song does the RF range decrease?  I really believe it's the Class-D hi-frequency modulation radiated over the speaker wires killing your RF as I'm quite certain the RF receiver uses a super-regenerative front-end which is a vacuum cleaner of EMI (as opposed to, say, a super-heterodyne design).

As for the stereo stuff, it's hard to imagine taking advantage of stereo sound in a piece of rolling stock given the size.  That said, why not take advantage of the spare audio channel.  I'd think your DACR and DACL outputs are low-current outputs and you have a mono amp on board?  Otherwise you'd think they would have provided two speaker outputs.  But the DACR and DACL may actually be PWM'd output full-scale 5V - as opposed to a true linear DAC which is expensive to fabricate on an IC.  This, to me anyway, is the key to keeping the circuitry simple.  That is, the PWM DACs or Class-D amplifiers put out 5V pulses even for small audio signals; it's just the duty-cycle that changes.  It makes is so much cheaper to "square up" full-scale pulses than to amplify analog signals up to the right amplitude.

Stan, I did try the experiment with an 8 ohm resistor, the range was still affected the same when a sound was playing.  I suspect you're right about the RF circuit, it doesn't have the components that I'd expect for a superhet receiver.

I wasn't suggesting taking advantage of the stereo in the rolling stock, that's why I figured one channel would be available for the control function. 

I took a look at the signals coming out of one of the DAC channels, they look like normal audio.  If they're generated with PWM, they have added filtering before the outputs.  I went pretty high frequency and didn't see evidence of full voltage excursions that I'd expect with PWM.  Here's a few samples of my test sounds at the left DAC output, I assume the right would be similar.

Coal loading into tender

Reefer Compressor

Track Running Sounds

Geez.  You're probably thinking I've sent you snipe hunting .  Well, maybe time to give the direct-RF remote a rest and focus on the TMCC version since that's what you need. If you do come back to it, there's more snipe hunts available to the hunting enthusiast - conducted susceptibility via the power or i/o lines.

I remain convinced that a low-cost remote "system" to activate rolling-stock anywhere on the layout is something O-gauge could really use.

That's pretty slick if they integrated the filtering on the chip and/or module.  As I recall, some microcontroller chips that feature "analog" output pins are BYOF (bring-your-own-filter).  Of course for this hack of using the spare audio track as a control channel, I'd rather have the digital pulses but that's tomorrow's problem.  Thanks for taking those scope shots, I can see the 1/2 scale DC offset on the DAC channels. 

As you may recall, I'm fiddling with the barebones $1 MP3 module plus external amplifier module. I'm finding each O gauge application has some quirk to work out.  For example, in the case of servo PWM, to get the pulse timing resolution to smoooothly rotate the mechanism, let's say you want 1 part in 100 resolution.  Since the pulse width varies about 1 millisec for full servo swing range, that means a pulse resolution is 10 microsec.   Back-of-envelope says that requires a sample rate of at least 100 kHz.  Well, for example, I see your MP3 module specs a max sample rate of 48 kHz...and I'm sure my MP3 module is no better.  So then you start playing games...or as the saying goes all ideas eventually degenerate into work.  I'm just learning about how the MP3 compression algorithm and these low-cost MP3 chips handle DC-coupled digital pulse streams.

This topic is way over my head but if anyone can do it, it will be GRJ.

I did more reading on class-D EMI suppression, it's obviously NOT a simple topic!  Boy, some of the stuff I've found is really deep!  Many of the discussions get into to EMI stuff I used to do for avionics, however I don't have the equipment or testing facilities that I had available back then.  When you had OPM to buy all the toys, it was a lot easier to work on some of these issues. 

Many of the references were to using ferrite beads and small capacitors across the speaker leads to minimize the problem, including some specific application examples.  With that in mind, I ordered some trial components to see if adding filtering right at the MP3 module might improve the situation.  No certainty, but it's worth a shot.  I do agree that having a generic capability without TMCC in the mix would be a good thing, but it's getting there that's the issue.  I think I may be able to stomp on much of the higher frequency EMI from the amp, we'll see when the parts arrive.

Another approach would be NOT to use the amplifier that's on the MP3 module, but rather incorporate a separate amplifier chip on the interface board.  The TI LM4861 Class AB chip would probably generate a lot less EMI.  That's what might have to happen in order to solve the EMI issue, after I try a filter I'll know.

There are also modules that play WAV files, which would solve the problem of dealing with MP3 files.  The module I'm using is one of those.

Before I lose my train of thought, one other possibility if you re-visit these modules. These low-cost RF modules appear to come in either 315 or 433 MHz versions...I think that's a SAW resonator on the TX side and I can read the frequency on the can.  It could be that one receiver is less susceptible to the MP3 harmonics/noise than the other.

I'm focusing on the what to do with the spare audio track right now, but my to-do list now has to hook up the 4-channel RF receiver to see if it plays well with the $1 MP3 player driving the 82 cent class-D audio amp.   My 4-channel TX is 315 MHz.

I have the 315mhz units as well.  Given the nature of the receiver, I can't see changing the frequency as being the primary solution.  If the filter doesn't do it on the MP3 module, maybe the alternate amp might be better.

Truthfully, I'm not as concerned about keeping the cost rock bottom, I'm more interested in functionality.  I don't want the cost to spiral out of control, but anytime you have to layout and build circuit boards, it's not going to be a $5 solution unless you're the Chinese and you're building 100,000 of them.

For either of these versions, I want to be able to play any one of the four sound files and also stop it on command remotely.

GNR and Stan,

Gentlemen, I am enjoying this thread immensely. Both of you are obviously well versed in the subject material. I am some 40 years removed from this level of electronic detail so my voice is very small in regards to this endeavor. With the knowledge that GNR has some plan "B", plural, to tinker with I would like to note that up until this point, from my outside aspect, a 20,000 foot birds eye view in you will, there seems to be a lot of Band-Aid fixes to this issue; would a foundational level review be in order? Unfortunately, I am left with leaving you with an observation and no clear solution, which goes against a post of this sort. In short, to me, there may be another angle to tackle this problem with less moving parts. IMHO  

David, we're all ears if you have a better idea.   There is obviously a more reliable way to do all of this, but it costs more money.  The object of the exercise is to leverage the stuff that we can get very cheap to do the job for a reasonable price.  I have some more involved ideas rolling around in my head, but I'm trying to come up with something with less total expended effort.

Stan, I did a new design of the RF version with the filters in place right at the connection of the MP3 board.  Once I bench test the filters, if that is a positive test, I'll consider cranking out a few of these boards to see how well it works.

Here's the trial run of the PCB topside, the MP3 module plugs in on the right, and the RF receiver plugs in on the left facing away from the board.  The board is 1 x 1.6 in size, with the receiver sticking out the end, adds about another inch.  Should fit easily into rolling stock.

Edited to remove obsolete drawing.

GNR, I am so out of my league on this issue and my hat is off to you and Stan. That said, I would go with a larger and more expensive, probably, tried and true solution and then distill it down instead of starting with the most inexpensive solution with band aid fixes to get where you want. I think you are on the right track with an FM solution using the FM carrier wave with the two sideband (read stereo) frequencies to get where you want. You are able to take it down to the IC level which gives you the advantage to distill. At the risk of repeating what you already know as a resource, Mouser.com have many industrial solutions to a duplex communications issue in a very small form factor. There is a Zigbee line of solutions that may fit the bill and are not at all expensive in the big scheme.

1. DRC would catch it but your bridge ground is floating.

2. If I were doing it, I'd change your RC delay by increasing R and decreasing C.  Resistors stay same size and same cost irrespective of value, capacitors generally decrease in size and cost for values shown.  Ceramics IMO generally more reliable than electrolytics.

3. I'd change the inverter after the RC to one of the unused HC14 gates.  Don't know about cost but it might be cheaper and smaller pkg for a single NOR than a dual NOR.  Also, for the RC time-constant you really should protect against possible oscillation on that inverter so the HC14 hysteresis will handle that.

4. Presumably you'll hold off ordering boards till you do more testing, but I'd give a bit more thought to how-to isolate the RF receiver.  This could be a simple bead filter between power-supplies should conducted susceptibility be the issue.  I also can't help think of consumer electronic devices where, when you open it, the RF module/tuner/whatever is usually in a shielded can/enclosure.  In other words, different than shielding the MP3 electronics.

In any event, I am genuinely excited that you are still considering a general purpose approach rather than a TMCC-only version requiring a somewhat spendy ERR module.  I understand your comment about cost, but if in the end you do need to go with a $1 microcontroller, a lot of SPI code talking to JGL's a $1 2.4GHz module I still think the cost will be reasonable.  Actually my concern is that the general OGR readership "gets" the significance of what we're talking about here.   I don't know what to do about that aspect though.

IMO if you can pull this off - allowing connectivity with an ASC/AIU, this will have a greater impact on the hobby than your super-chuffer and LED board combined.

Stan, the DRC didn't catch that missing trace, I was just rushing it to put it up. However I managed to screw some much bigger things up, so I'll have to fix all of them.   I managed to swap the MP3 mounting connectors around, not a good thing, the sound board would not have had the right connections!

I didn't originally have that inverter, but that's an excellent idea.  I started with the design of the MiniCommander interface, and it's using opto-isolators in place of those inverters. I didn't think to change when I got the HC14 in the mix.

I was thinking about the time constant being too short, that was one thing I was going to play with on the bench before going with the values I have.  The little 10uf cap I have is a pretty cheap one, and as far as "going smaller", I have imposed a limit of how small I'll do SMT stuff.  I know they can assemble it, but if I can't hand solder it easily, it's a big problem.  All my resistors are now 805 size, and I keep the capacitors that large as well.  I made the mistake of specifying a 201 sized resistor on a board, I couldn't even see the little things!  I tried to solder it on for 20 minutes, and three or four ended up on the floor, lost forever!  I can manage the 603 sized components, but I prefer the next size up for easy of handling.

For the receiver, I was going to make a tinfoil box for it and see if that helps.  However, I figure if I can whack most of the EMI with the filter, that will make the receiver problem easier to deal with, and any shielding more effective.

The key to "pulling this off" is getting an MP3 solution that plays nice with the remote.  So far, this one has had the other attributes that I was looking for.

Edit:

I corrected the graphics in the previous post, still a work in progress, but each step we get a little closer.

I'm not ordering boards yet, the boards are just to see how everything fits as I go.

Good .luck John sounds like your getting close to figuring it all out. how about using a Bluetooth speaker then you can install that anywhere on your layout!

What do you think!

Alan

I don't know if this is "good" news or just news but I tested the range of my MP3 player and class-D amp controlled by the LED remote system I showed on the other rolling stock thread.

I got 15 ft. reliably around a corner down the hallway.  But the key is the performance was the same whether turning ON or OFF.  The audio amp is class-D using the PAM8403 chip running at 260 kHz; there is no separate LC filter the board.  The datasheet claims the chip is a low-EMI design but who knows what that means relative to your module's amp.  But I can't imagine and class-D design having lower EMI than a class-AB style squeezing out battery-life is not an issue with rolling stock.

The LED controller remote uses a 3V coin battery with no external antenna so 15 ft. is pretty good.  The 4-channel remote uses a 12V battery and has the telescoping antenna which together almost seems like cheating compared to the LED controller remote!  I opened up the LED controller receiver and it appears to be a superhet design.  From various listings for the remote, it is 433.92 MHz and that crystal on the right (6.7458 MHz) matches what various single-chip superhet IC receiver datasheets say as the crystal to use for 433.92 MHz.  I could not correlate the "PP833L" chip marking on the so-8 receiver IC to anything.

So while I agree that you don't want to get in the business of designing RF receivers, this LED controller shows it's getting simpler with an 8-pin SO, a crystal, and maybe half a dozen 0603-sized passive components and no tuning-coil to mess with.

Stan, the thought is always there, but I'd have to build both ends in order to have the complete package.  I certainly wouldn't want tuning coils, that raises the complexity and reliability I would expect.  It would make for a more compact package.  I'm eager to see what the EMI filter on the speaker does for the existing setup.  If that works well, I'll probably order the basic $30 worth of boards (minimum order) and build up one to see how it functions.

This board appears to use a BY8001-24SS chip to do the basic functions, there's an Atmel serial EEPROM, presumably for parameters, and then a 8 pin chip labeled MIX2002CN?4S0N, the ? is a number I can't read.  I find nothing on that chip in a search.  The mystery chip seems to most likely be the amplifier, it's next to the speaker leads.  Since two of it's pins are directly connected to the speaker, that's a good indication.

I've done a "trial run" of the RF version final assembly and PCB layout.  If anything jumps out at you, let me know.  As soon as I get to test the EMI filtering, I may pull the trigger on one of these if that's a success.  Of course, any comments welcome before we go to copper.

Edit: remove obsolete diagrams.

John, what I would really like to have is a very inexpensive microphone and sound card run off a single AA that would pick up the wheel noise coming through the car body and convert the sound to prototype steel wheels sound with clank on a live basis. One in each frt. and passenger car would be awesome.

At least, I'de like to try it. Maybe one for locos first that rumbled. 

Ron H

I'd like to win the lottery, but that ain't happening either.

Everyone want's "very inexpensive", but if there's no potential for any profit, who's going to build it?  While it sounds simple, one only has to refer to my signature saying...

You want to do live digital sound processing on the wheel noise, and that's going to be "very inexpensive"?  I'll buy them if you get them to $10-15, but I doubt even the Chinese are going to do that!

John, digital amplifier bds available on Ebay from china at 99 cents to $2.48. Micro microphones available also. Hard part is turning the actual sound to steel wheels on rail sound.

Ron

And no, if it's so easy, I am not an electric guy and cannot do it myself. 

You're exactly right!  Doing the sound processing to come up with realistic sounds is not a trivial task.  Truthfully, I don't have the background to even guess how that might work!  The electronics of the picture are the easy part.

Well, the Class-D EMI filter was a huge disappointment!  I soldered the little surface mount components directly to the speaker pins, that's better than they could be on a circuit board.  It didn't change anything that I can see with the remote control, it still starts the first sound great, but if that sound is running, you aren't going to trigger any others from more than about 5 feet away.

If this is going to work, I'm going to have to find a better RF remote control package I suspect.

If you still haven't given up on that 315 MHz 4-channel receiver, I'd look at conducted susceptibility over the power supply.  I take it you're still using the white prototype board albeit with receiver and MP3 modules separated by (at least) 6 inches.  One of those DCS 22uH inductors has 40K ohm impedance at 315 MHz.  I figure you only need maybe 10 mA max into the receiver so those cheap 1/4-1/2W inductors can handle the current.  I'd just put a 22uH on on ALL the lines to start including the 4 output lines.  Or just disconnect all but the 2 control lines to turn MP3 on and off.

Separately, since you apparently have multiple receiver modules all with the 0.1" square-pin headers, how about hooking up two receivers where one is completely electrically isolated (runs off a separate 5V supply).   So the 2nd receiver would maybe just drive some local LEDs or whatever's enough to show that it is decoding the commands.

If you have just the RF rcvr 4-pin module that comes paired with the bare tx module (99 cents), even that could be used to evaluate if/how the super-regen rcvr behaves when the MP3 player is playing if you look at the data stream output with a scope.

So the idea is to determine whether the MP3 module is just so darned noisy that it blows away anything in the same county that isn't even hooked up to it.  Yeah, all these high-school science projects take time...

-----------

Completely on a different issue, regarding the PCB.  I can't imagine anyone else trying to solder-up this circuit given the complexity.  I say this based on what guys will do/did for various LED lighting circuits.  That said, if going to the trouble/cost of making boards, I figure you'd just put in a 14-pin (or whatever) microcontroller.  Initially it could eliminate all the logic gates since it would be a fairly trivial program to do the inversion and the minor timing to pulse the 5th channel to turn off audio.  But obviously the next step would be decoding the RF packets using the cheaper receiver that is 1/2 the size of the 4-channel receiver.  Even if the board is just for yourself, I figure you probably want to be able to set the address of the board or else ALL you sound cars will respond to same RF controller.

OK, if you're a "glass half-full" person:

I used the basic $1 MP3 module + $1 PAM8403 class-D audio amp module powered by 5V. When powering a 4-ch 315 MHz receiver module with the same 5V supply, I get data loss whenever audio is blasting away. During MP3 silence gaps, data goes thru fine.

Then I powered the MP3/amp with a separate 5V power supply. No problem getting data thru even with audio blasting away.

Then I connected a wire between the grounds of the MP3/amp and RF receiver. No problem!

Even draping the un-filtered speaker wires 1-inch from the RF receiver...no problem!

My working theory: it's coupling thru the power supply.

Stan, I'll try some of the 22uh chokes in the power and ground.  I see the receiver already has a cap across the power/ground, so the addition of the chokes may be a winner.  FWIW, in order for this to work, I'd like the receiver to be plugged in to my board, so I do want it to work closer.  I'm thinking you may be on to something with the power, I was thinking about that my self.  I have a whole raft of the little 22uh chokes, so it's an easy test.

I also put in an order for a superhet 4-channel receiver to see if that has better luck.  It has the same pinouts as this one, so it would be a drop-in if it works better.

We'll see how the chokes in the power feed work...

I added a 78L05 right at what you call J1 to power the RF receiver separately. Let's just say the results were dramatic. 

The 22uH did NOT work in my situation but we do have a slightly different configuration.  In my case I was looking at the noise on the 5V supply to the receiver on the scope; it's low-freq audio noise from the current variations that I believe is messing up the receiver sensitivity.  The 22uH does not have enough of a low-freq effect.

This is one of those cases where I hope Your Mileage does NOT vary!

BTW, not today's problem, but are you really using 10 or 12 mil traces on your power supply lines on the PCB.  Given the average current when pumping out a couple watts of audio, I'm thinking 25 mil minimum and the more the merrier...you have the room.