Weak DCS signals, Failed TIU Output Drivers, and Design Solutions devloped under collaboration between AGHR and MTH

WOW  Looks like I'm going to be waiting on additional and upgrades to my TIU.  New requirement to be a model railroader, have a PhD in EE!!

My layout currently in the early stages gives the "Engine not on track" message sometime, even though they will run and I get the double toot all the time.  This with 4 120ft mains with 5 blocks each loop, 2 loops per fixed channel.  Granted I have an old rev H TIU that was given to me, but I am getting by.

Great investigation and hopefully MTH will bring Rev M out before I need to by another.  MTH, please put an outside antenna jack on as well.

As an aside, anyone starting with DCS needs to by Barry's book(s), they are invaluable and saved me some hair.  This forum is great also.

Gray Lackey

 I could only wish that there was an indicator that showed when these TVS things fail. Even better, would be if they could reset themselves.

 I don't fully understand what's going on here. I'm just an operator and like to see things move forward. It's great to see you stuck with it. Most in the past, have just yanked the system out and ran away!

Adrian,

Would this problem be the same in all or most REV TIUs or just REV Ls?

Thank you for posting this!

One other question:

When the TSV blows, is anything else down the line taken out because of this?

Thanks,

Thanks.  That is great information and even better engineering.

WAY beyond my skill and/or understanding level, but I certainly do appreciate the amount of effort you put into identifying and solving the problem.

Those pesky TVS everyone likes to add to layout and devices, and low an below these short and you don't know it.  These where a fix for Rev L TIU to protect the transmitter.  Good to know.  So what I would love to know, is how long an Sans TVS Rev L or a Rev I last on your layout.  Another component to check when repairing TIUs. Thanks for posting. G

GGG posted:

Those pesky TVS everyone likes to add to layout and devices, and low an below these short and you don't know it.  These where a fix for Rev L TIU to protect the transmitter.  Good to know.  So what I would love to know, is how long an Sans TVS Rev L or a Rev I last on your layout.  Another component to check when repairing TIUs. Thanks for posting. G

Yeah they fail to short-condition because its a zener. If you were super careful you could just run without them, but the component taking the hits will be the ACT244 octal. It's beefier and has an ESD clamp on the chip die, but it's a much more painful thing to change since you need the hot air.

David Minarik posted:

One other question:

When the TSV blows, is anything else down the line taken out because of this?

Thanks,

David Minarik posted:

Adrian,

Would this problem be the same in all or most REV TIUs or just REV Ls?

Thank you for posting this!

 

I haven't been in the hobby that long so I don't know when these TSVs made their debut. It looks recent. Others here probably know better than me.

Really too deep for me to comprehend the technicalities, but I see two phrases:  TSV and TVS.  Is there a letter juxtaposition, or are these two different gizmoes?

Well, you don't really need hot air to change the ACT244, but I'll allow it's a lot harder to replace than the little TVS.  I simply cut all the legs off the old one right at the body, then sweep the remains of the legs off the pads with the iron, then position and solder the new chip on.

That's excellent work on the fix, and I'd have never imagined you could take out the 500W TVS so quickly across those parts.

Would you post a schematic of the diode PCB, I'd like to take a closer look at perhaps a more compact solution.  I use SMT diodes that would have similar characteristics as the 1N4148, it would be easy to do a more compact board.  Now that a solution is at hand, maybe there's a more compact clamping solution to be had, perhaps even other than multiple diodes.

One question arises, the diodes would only clamp in one direction, are you using two sets of them to clamp the bi-polar signal?

RJR posted:

Really too deep for me to comprehend the technicalities, but I see two phrases:  TSV and TVS.  Is there a letter juxtaposition, or are these two different gizmoes?

Yeah thats a typo. TVS (Transient Voltage Supressor)

gunrunnerjohn posted:

Well, you don't really need hot air to change the ACT244, but I'll allow it's a lot harder to replace than the little TVS.  I simply cut all the legs off the old one right at the body, then sweep the remains of the legs off the pads with the iron, then position and solder the new chip on.

That's excellent work on the fix, and I'd have never imagined you could take out the 1500W TVS so quickly across those parts.

Would you post a schematic of the diode PCB, I'd like to take a closer look at perhaps a more compact solution.  I use SMT diodes that would have similar characteristics as the 1N4148, it would be easy to do a more compact board.  Now that a solution is at hand, maybe there's a more compact clamping solution to be had, perhaps even other than multiple diodes.

One question arises, the diodes would only clamp in one direction, are you using two sets of them to clamp the bi-polar signal?

Hey,

I never did a schematic because it was too simple, I just drew it out-right. Compact 1n4148s would work but if time was spent you could find a rectifier with a higher drop so you wouldn't need so many.

Here's a doodle:

I tried both the bidirectional and unidirectional (two back to back, just like you're thinking) TVS diodes, but they're both pretty much the same IV characteristic. In the end the reason I went with the unipolar one for the complicated reason that they were in stock on digikey, while the bipolar ones were not.

The 1n4148s are two parallel strings up and down.

Thanks, Adrian.  Thought that was it, but not sure.

Well done Adrian.

Over my head but I do capture a glimpse of what you are saying.

Hello MTH.

Thanks

Dave

Oh yeah, other details:

1. In the original post I said 13 clamping diodes but it's really 10. It was 13 on the breadboard but 10 on the final PCB. I wanted the big clamp voltage a little lower than the small clamp voltage to try and get an early turn on.

2. If the original tiny TVS fails and sits that way for more than a few minutes (like a day) then the ACT244 is usually damaged too (since it's been sending commands to a short circuit instead of a train). So if you lift off the diodes and the swing doesn't come back, that's the next thing to look at.

OK, so a single diode board is used for one output, and you just have two diode strings back to back.  That's what I was wondering about.  I was thinking along the lines of a higher drop, or even some active circuit that didn't require so many components.  Also, it would probably all be put on one board with wiring to the appropriate places for all four channels.

I always start with a schematic as I'm too lazy to manually route, I just let the auto-router do that job.

Maybe three of these in each direction would do the clamping in a much more compact form?  Sanken EG01CV1, six instead of 22 diodes would be a lot easier to deal with.   I'm also not sure you need the TVS as well, I'd think these would do the trick.

gunrunnerjohn posted:

OK, so a single diode board is used for one output, and you just have two diode strings back to back.  That's what I was wondering about.  I was thinking along the lines of a higher drop, or even some active circuit that didn't require so many components.  Also, it would probably all be put on one board with wiring to the appropriate places for all four channels.

I always start with a schematic as I'm too lazy to manually route, I just let the auto-router do that job.

Maybe three of these in each direction would do the clamping in a much more compact form?  Sanken EG01CV1, six instead of 22 diodes would be a lot easier to deal with.

 

Exactly!

Yeah those might work. The best thing to do is buy a handful and put them under full conduction for an hour and see if the knee voltage changes over time.

I edited, but too slow.   I'd also consider just eliminating the TVS as the diodes have a pretty fast response.  If you position the smaller boards made possible with fewer parts, I don't think the propagation delays would be an issue.

You could also up the rating of the TVS slightly to insure the diodes are taking the heat if the response is too slow.

gunrunnerjohn posted:

You could also up the rating of the TVS slightly to insure the diodes are taking the heat if the response is too slow

also theres other speed considerations.... it’s a bit complicated but basically the TIU pcb has a ground and vdd plane so the traces on it look like microstrip lines (real part only... no jX) with low impedance terminations since the transformer and ACT244 outputs are low impedance.

Once you have a long umbical cable you start picking up inductance that can hurt your time constant. That’s the advantage of soldering the TVS right on the spot.

I only had to make 5 sets and it was a prototyping exercise so nothing was a final print .... but If I had to do 500 I’d probably do a very very low profile small area board like 0.5” x 0.5” with SMD diodes and have two vias on the back that solder directly to the points on the main pcb. That keeps the time constant up and reduces the labor per.

Again, you’re good at PCBs so I’m sure you could tighten it up a lot.

This is seems to be a Great Upgrade, MTH should incorporate this and an external antenna port, into a new Rev"M" TIU ASAP, and also develop an upgrade component for Techs, Thanks for all your work and persistance on this. I met Jason a few years ago, May have been in San Diego or Pasadena,  a good guy in the business. 

Thanks Again, RWDeano

A good idea to attach the PCB directly to the ACT244, it would probably not be too difficult to rig something that solders to the top side right on the chip to keep the leads short.  Then you could put the TVS and diodes on one board.  I'd have to do more research to see how this plays out.

I wonder if MTH is actually going to do something with this to prevent this issue.  I know the NJ-HR folks have had similar issues with their TIU's, this explains what is happening I suspect.

I've read this whole post and passing it on to our Electrical Engineer cub member. I know he will understand what you folks are talking about. We have too many witches brews on our layout trying to maintain a dependable "DCS" signal strength. This sounds like this guy figured out the fix.........Very much thank you for your perseverance, in coming to a resolve...

Great for EEs. What do the rest of us do in the meantime??!

I see a small cottage industry for someone to fix TIUs Rev Ls ; I can't imagine MTH taking this on as a warranty repair.

this fix sounds exactly what we need

AlanRail posted:

Great for EEs. What do the rest of us do in the meantime??!

I see a small cottage industry for someone to fix TIUs Rev Ls ; I can't imagine MTH taking this on as a warranty repair.

Everything here is open to use as far as I’m concerned, I just want to help people avoid the problems I’ve been having. Maybe one of the professionals like GGG or gunrunnerjohn can do the mods for folks

its not expensive, maybe $10 a TIU in quantity. The pcb is more, but I only made 25 so a larger volume would benefit there too.

(or just “run” the trains without power)

A question from the peanut gallery, can the TVS be put between the power supply and the TIU?

anything we could put in front of the TIU to handle the spikes so this wouldn't occur?

If someone wants to do the upgrades, count me in for two.  I'll GLADLY pay freight both ways plus the charges for the TVS issue and the antennae jack.

Gray Lackey

Ron_S posted:

A question from the peanut gallery, can the TVS be put between the power supply and the TIU?

Engineer-Joe posted:

anything we could put in front of the TIU to handle the spikes so this wouldn't occur?

It's not so simple actually. You want the TVS right up against whatever it's protecting:

There's a little bit of RF design principles here (sorry it's complicated)... let me try to explain it easily:

What happens when a train derails is the motor and bouncing contacts produce transient events (fast changing voltages and currents over a short time duration... like 1-10ns range). These transient events are short in time which means they are very wide-band in frequency (fourier transform describes this).

They can be up to the GHz range,  which means they behave like RF waves on the wiring and track (because they have a comparable wavelength in the 10s of inches regime). Once you are into this wave propagation behavior it means that the whole wire length is not the same voltage at the same time. So just because you null out the voltage at one point along the wire with the TVS between the train and the TIU doesn't mean the transient voltage at the TIU or train is zero because of the waves bouncing up and down the wire.

If you want the basics it's here: Transmission lines

So the only way to be 100% sure the transient voltage at the protected device is safely held down by the TVS is to physically place it very close to the thing your protecting (like within 1/10th of a wavelength). That's why it has to go right on the board inside the TIU.

Hope that wasn't too poorly explained!

Adrian, it makes sense, I'm just one of those electrically challenged individuals who needs braille to make it beyond flipping a switch.  (Fagatrons are a necessary evil, if you don't want their environment destroyed or hunted down, go ask your mother)

Thanks for the comprehensible explanation, Adrian.

Excellent work & Nice Explanation!

A toy control system which takes an IEEE to unravel was not designed for the casual user, that's a given.

Adrian! posted:

Ron_S posted:

A question from the peanut gallery, can the TVS be put between the power supply and the TIU?

Engineer-Joe posted:

anything we could put in front of the TIU to handle the spikes so this wouldn't occur?

It's not so simple actually. You want the TVS right up against whatever it's protecting:

There's a little bit of RF design principles here (sorry it's complicated)... let me try to explain it easily:

What happens when a train derails is the motor and bouncing contacts produce transient events (fast changing voltages and currents over a short time duration... like 1-10ns range). These transient events are short in time which means they are very wide-band in frequency (fourier transform describes this).

They can be up to the GHz range,  which means they behave like RF waves on the wiring and track (because they have a comparable wavelength in the 10s of inches regime). Once you are into this wave propagation behavior it means that the whole wire length is not the same voltage at the same time. So just because you null out the voltage at one point along the wire with the TVS between the train and the TIU doesn't mean the transient voltage at the TIU or train is zero because of the waves bouncing up and down the wire.

If you want the basics it's here: Transmission lines

So the only way to be 100% sure the transient voltage at the protected device is safely held down by the TVS is to physically place it very close to the thing your protecting (like within 1/10th of a wavelength). That's why it has to go right on the board inside the TIU.

Hope that wasn't too poorly explained!

Previous advice seen here: TVS placed across the TIU output(s) does the trick, yes/no?

Kerrigan posted:

Adrian! posted:

Ron_S posted:

A question from the peanut gallery, can the TVS be put between the power supply and the TIU?

Engineer-Joe posted:

anything we could put in front of the TIU to handle the spikes so this wouldn't occur?

It's not so simple actually. You want the TVS right up against whatever it's protecting:

There's a little bit of RF design principles here (sorry it's complicated)... let me try to explain it easily:

What happens when a train derails is the motor and bouncing contacts produce transient events (fast changing voltages and currents over a short time duration... like 1-10ns range). These transient events are short in time which means they are very wide-band in frequency (fourier transform describes this).

They can be up to the GHz range,  which means they behave like RF waves on the wiring and track (because they have a comparable wavelength in the 10s of inches regime). Once you are into this wave propagation behavior it means that the whole wire length is not the same voltage at the same time. So just because you null out the voltage at one point along the wire with the TVS between the train and the TIU doesn't mean the transient voltage at the TIU or train is zero because of the waves bouncing up and down the wire.

If you want the basics it's here: Transmission lines

So the only way to be 100% sure the transient voltage at the protected device is safely held down by the TVS is to physically place it very close to the thing your protecting (like within 1/10th of a wavelength). That's why it has to go right on the board inside the TIU.

Hope that wasn't too poorly explained!

Previous advice seen here: TVS placed across the TIU output(s) does the trick, yes/no?

Maybe..... The plastic terminals are still a few inches away from the octal driver chips that are being protected. Also the PCB isn’t made of air so the wavelength is even shorter.

you can try it and see. The catch is you’ll need an oscilloscope and weeks of measurements to validate if it’s actually fixed or not.

That’s the part I can’t emphasize enough actually.... this has been more like a clinical trial than a design check. You never know what’s going to happen in the club so you have to measure over a long period to validate everything is okay. Adjusting the TIU is trivial ...but lugging the scope under the table with an extension cord and disconnecting the layout channels one by one for 5 TIUs and measuring signals is like a 1 hour start to every club visit.

Excellent write up!

Thanks for doing all the research and experimentation at your club to find this gremlin!

-Dave

Thanks for the replies!