ERR Sound Converter Struggles

gunrunnerjohn posted:

Here's the 'scope pictures under discussion.  It's interesting that the Sound Converter and the transformer do the offsets in a totally different way.  I'm guessing the chopped waveform has something to do with the difficulties here.

maybe there is a little more to this sound board logic/code.  It has to work over a range of track voltages with same audio output level over varying input (slow to fast).   It needs to interrupt chuff momentarily for whistle or bell then automatically resume chuff. 

After seeing the almost textbook-like waveform outputs of the Sound Converter, I believe the issue is with your Sound Module's DC detection.  I'm assuming you don't want to modify either Sound Converter or the Sound Module.  So it seems either 1) switch to a non-chopped transformer, or 2) experiment with modifying the shape of the CW voltage so the Sound Module can make sense of it.

In regards to 2).  I realize you don't have a bipolar capacitor but I'll bet whoever loaned you the scope has a couple 22uF or 47uf or 100uF capacitors lying around.  Make a bipolar cap out of two regular caps and place it directly on the input to the Sound Module.  It won't change the waveshape significantly but if you see any change in whistle behavior (better or worse) it would point a bigger finger to the DC offset detector.

Thanks Gunrunner for posting my pics.  Stan, please pardon the rookie mistake on the 'scope probe ratio.  I do have access to all kinds of electronics hardware and can drum up some caps.  Maybe this is the end of the line for my 7 little CW-80s.  I always intended some day to step up a level but thought it would be for lack of power and not the shape of the power. 

I just did a little test when powering up the Berkshire and then holding (pulling the rope) on the remote whistle and I got 1 second whistle, 1 second off, over and over again.  It was not a continuous blow like I can do with the transformer. Maybe that supports what we are thinking already.

Trying it with a pure sine wave transformer would certainly be an option to try, that would be a good data point.

OK team, here is some more data.  I have a variac and put in 16.5 VAC.  See the scope photos but do adjust by a factor or 10 (probe ratio). 

Neither the bell or whistle worked with this scheme but we can see what the converter is doing.

 PURE SIGN WAVE:

Sign Wave Whistle:

Sign Wave Bell:

Can you post some closeups of your Sound Module?  I'm pondering how to infer the DC offset detection scheme. I'm assuming no one has a schematic.  Does it have a micro-controller on it and if so can you read the numbers on it.

For example, I'm trying to imagine some circuit or algorithm that chokes on that flat-top region where the Sound Converter clamps the voltage to 10V.

Are we having fun yet?

They had to work to come up with a circuit that would ignore the DC offset.

Here are the pics of the boards.  The sound board (tender) has a microprocessor on it.  It has a sticker labeled:  .S19

Sound Converter Board (ERR):

Sound Board Tender (Lionel):

How complicated is the back side of the sound board?

Would you feel comfortable peeling off the S19 sticker and reading off the lettering on the part itself?

My money is on a PIC processor...

Yes.  But is it a version with an A/D or just comparator inputs.  I don't see any bipolar caps on the board which is the "classic" way to detect DC offset, nor do I see an op-amps which might level-shift any DC offset to center the AC into the 0-5V range of a baseline PIC chip for A/D conversion.  Hence I'm speculating some kind of timing algorithm which looks at zero-crossings or the ratio of time above-to-below some threshold voltage to detect positive or negative offset on an AC signal.  Clearly I have a noisy mind imagining some kind of conspiracy theory or such. 

Well, it's fairly hard to buy a version without at least one A/D channel, but I don't know how they do it.  I know they get confused if there isn't AC on the tracks, so the zero crossing is probably figuring into the picture.

Here is the back of the Sound Board.  I was very comfortable to pull the .S19 sticker off the chip.  I'm not so comfortable in reading the lettering.  That is tiny and I have magnifiers on.

Sound Board Back:

Stan....your killing me.   Those letters are worse than any eye exam I have taken.  Here is the best I can do:

M908Q4CPE

ZRKT

0651

You passed the eye exam!  That's the NXP (Motorola) MC9S08QG4 microcontroller chip. 

It indeed has A/D capability but unclear if it is actually used.  That microcontroller is undoubtedly spending most of its life talking to the stickered chip on the back labeled "steam".  That stores the various steam sounds and the microcontroller must get those sound samples "serially" which can consume a lot of microcontroller effort.  So I'm speculating that the designer didn't spend a lot of time with the DC detection and possibly skipped even bothering with the A/D and the overhead to mathematically average a multitude of samples to extract the true DC offset of the track voltage.

I realize all this geek talk doesn't help you one bit. 

I'll have to sleep on it.  This may be the end of the line. 

I know we are working both sides of the equation on this problem but I think the sound board designer did what he needed to do to make it work properly in transformer controlled environment.   I think it's more a case of the Sound Converter emulation of conventional is off the mark.     

Just wondering if the sound card trigger is based on DC offset range (e.g, 1.9 -2.3 vdc).  That is why we only get a short whistle with the converter because it's only there momentarily on its way to ~2.6 vdc.  

Hard to believe an additional offset would cause a problem, transformers are all different.  The CW-80 outputs a pretty steady 2.5V offset, for instance.

As before, if you have access to a pair of 22uF or 47uF capacitors, wire them up back-to-back in series to make a bipolar cap and put it across the Sound Module AC input.  I don't think it will fix the problem, but I'm interested in any behavior change for better or worse.  I figure it only takes a couple minutes to try.

I agree with GRJ that a bounded DC+ offset range is unlikely but then again the Peanut Gallery is awfully quiet with how to proceed.  That said, here's another simple experiment that should only take a couple minutes to try.  If you put a 1K resistor across pins 1 and 2 of the Sound Converter LM317 as shown, it will change the clamping action on the LM317.  This should decrease the positive DC offset when Whistle to activated.  I suggest 1K because it is a common value you might have in your pocket.  I suggest this method of altering the clamped voltage because it allows you to parallel the added resistor to presumably easily accessible points...rather than having to replace components, cut traces, etc..  You should see the scope waveform move as shown.  Again, it may not fix the problem, but I'd be interested in any behavior change for better or worse.

Well, the Peanut Gallery doesn't have useful suggestions of what to try that hasn't already been suggested.  It's not clear to me how they're sensing the DC offset and managing to exclude the Sound Converter, but the zero crossing triggering seems like it might makes sense and would result in totally missing the offset.

No peanuts here.  Just simple me and a distinguished Expert Panel.  I ran all the tests (experiments) that Stan requested.  I was unable to find a 1K ohm resistor but substituted with a 1.3 K ohm.  As for caps, I tried 22, 47 and 100 uF.  You can see what the 22 uF did (scope screen shot).  The 47 uF just started to repeatedly blow the whistle without demand, and the 100 uF made a high pitch squeal. 

The resistor made the shift that Stan predicted but didn't solve our problem.  Hopefully it helps us understand what makes it work. 

A few days ago I did send a link to the thread to Ken at ERR but have not heard back.  I am going to reach out to Lionel Tech Support the same way and maybe they can chime in sometime. 

CW-80 chopped sign wave (baseline):

Remote Control Whistle (baseline):

Remote Whistle w/ 1.3K ohm resistor:

Remote Whistle w/ 22 uF Caps:

Nice experiments...but I'm at a loss for an explanation.  

My bet is still on the Sound Module's DC detection method.  I don't think the Sound Converter is the issue.  Since I believe the DC detection is done via code within that socketed/replaceable microcontroller chip S19, I'm imagining a pie-in-the-sky where Lionel upgrades the DC offset algorithm and sends you a new chip.  You may start holding your breath.....now.

But I'd wait to hear from Lionel.

The only take-matters-into-your-own-hands idea I can think of would be to fabricate a different DC offset generator using, say, the string of diodes method.  Presently, the Sound Converter clamps the negative voltage (at about -10V DC) using a LM317 voltage-regulator chip.  So you get that flat-line at -10V.  Messing with the clamping voltage moved the DC offsets as expected but of course the flat-line is still there.  So, for example, the same signal that tells to LM317 to clamp could instead trip a relay that activates a string-of-diodes to asymmetrically lower (vs. clamp) the negative voltage relative to the positive voltage.  In effect the flat-line portion at the peaks of the AC signal would be eliminated though you would add more flat-lining at the zero-crossings.  It's a long-shot for sure...but only a few dollars in parts and the circuit could be duplicated to generate the negative DC offset for bell activation.  Again, this a one of those "it's a matter of principle" situations!

Separately, and this may fall under the "re-arranging deck chairs on the Titanic" category, but in case you haven't found the manual page, I think this is how you change the probe-attenuation factor to 1:1 so that the displayed measurements aren't off by 10x.  Probably not worth the effort since you already have so many scope shots and this would likely just confuse matters.

Thanks Stan, Gunrunner, et al. for the assistance thus far.  You pushed me outside of the electronic comfort zone and taught me some things along the way. 

Maybe Lionel already knows of this problem.  Just so happens that the Sound Board shown on the parts replacement looks quite different from my original.  Coincidence or just an update?

Steims posted:

 

...Coincidence or just an update?

As TV detective Nero Wolf once stated,

In a world that operates largely at random, coincidences are to be expected, but any one of them must always be mistrusted.

I think he's turning blue Stan, probably time to stop holding his breath!

The odd part about all this is this project (the Sound Converter) was originally started to trigger a Polar Express sound car, I think the announcement car.  It's pretty amazing that another Polar Express sound board, presumably using the same technology, doesn't work with it.

gunrunnerjohn posted:

The odd part about all this is this project (the Sound Converter) was originally started to trigger a Polar Express sound car, I think the announcement car.  It's pretty amazing that another Polar Express sound board, presumably using the same technology, doesn't work with it.

 

You don't say...

I have that car also and wasn't even thinking about remote control of it.  This may be the end of the line for the Sound Converter portion of this project.  I set out to remotely control our Polar Express around the Christmas tree without me having to sit on the cold tile floor and run the transformer.  We have accomplished that with the Cruise Lite.  The icing on this cake was to blow the whistle and ring the bell that same way.  Maybe it was not meant to be.  But still, my cup is more than half full.   

I put one of my prototype sound modules in the baggage car of the PE that I upgraded to TMCC, and then I could control the sound tracks with the little 4-button remote.   No crawling on the floor for me!

I looked all over the Henning's Trains site and didn't see that little doodad anywhere.  Please let us know when they go into production.  

That's why it's a prototype.   I am planning bigger things, but I think I may have to scale back a bit so I can actually get something out.

stan2004 posted:

Yes.  But is it a version with an A/D or just comparator inputs.  I don't see any bipolar caps on the board which is the "classic" way to detect DC offset, nor do I see an op-amps which might level-shift any DC offset to center the AC into the 0-5V range of a baseline PIC chip for A/D conversion.  Hence I'm speculating some kind of timing algorithm which looks at zero-crossings or the ratio of time above-to-below some threshold voltage to detect positive or negative offset on an AC signal.  Clearly I have a noisy mind imagining some kind of conspiracy theory or such. 

Stan, You are correct!  The ISND design uses timing, not DC offsets.   

STEIMS, sorry for the unsolvable problem.  I salute you for your energies!  Ken mentioned we are making it right for you.  I appreciate your patience and efforts to try to make the Sound Converter work.

Well I tried to get assistance from Lionel Technical Support but I only got as far as the Customer Support Director.  He told me to take it up with ERR.

On a positive note, ERR is willing to take the item back for a refund. 

Not how I wanted things to turn out but I learned a lot and maybe others learned a little along the way.  

Thanks Stan and Gunrunner for all your advice.

So you're giving up? 

Well, I think that Jon kinda' put the nail in the coffin with his verification of how the the sound board works.

Stan, You are correct!  The ISND design uses timing, not DC offsets.  

Well, as long as we are in the world of science projects...

If you were to take the output of the Sound Converter and pass it through an appropriate capacitor, it would eliminate any DC offset, creating equal areas under both halves of the curve. This would put the zero crossing some distance up into the "unclipped" side, causing the zero crossings to be staggered. I don't know if it would cause enough shift to trigger the sound card but it is food for thought.

How much current do the sound cards draw when triggered? Anyone know what their input circuit looks like?

By it's design, the Sound Converter has to power the entire sound card, so it's not just the input circuit, it's the total power draw of the sound card.

gunrunnerjohn posted:

By it's design, the Sound Converter has to power the entire sound card, so it's not just the input circuit, it's the total power draw of the sound card.

Understood, how much is that? 100 mA? Does the sound card use a standard FW rectifier on the input?

A couple 500 uF back to back would be about 10 ohms at 60 Hz.

I've measured various sound cards at 100-150ma maximum when I was doing the TMCC battery design.  The 150 was with background sounds and the horn/whistle blowing, that seemed to be the peak power usage.

I don't know what the sound board he's using has, but the newer ERR RS Commander has a bridge rectifier.  The older Railsounds power board has a simple diode as it has a common AC and DC ground.

A follow up on the use of the Sound Converter Module with an LCRU. Thanks to GUNRUNNERJOHN for remembering an oddity with the serial data on the LCRU. The Sound Converter was loading the serial signal down to a low level. Adding a buffer also thanks to GUNRUNNERJOHN on another thread solved this loading problem and the horn now works with the TMCC whistle button. Having the schematic available of the Sound Converter also helped. A schematic diagram showing the buffer is included below.

Cool, glad it all worked out.

I used the Sound Converter along with a Mini Commander 2 in my MTH Galloping Goose. It will whistle, squeal the brakes and increase rpms but nothing else.  Too bad as it has some whimsical things like dogs barking, ducks honking and so forth. Oh well.

Any idea what happened to the Sound Converter?