Scope readings of the DCS signal

I didn't post that thread but friends and I were texting about it last night via WhatsApp so the filter was saved.

HPF filter (1K/0.1uF -3 stage)

Hey there!

I'm the guy with the captures. If you're interested in viewing the signal there's a whole post about it (old post) you can look at. I dunno why my post was taken down, I didn't get anything from the forum on that. I know one of the replies had some concerns, but I dunno why the post itself is gone...

The track power is typically 18-20Vat 60 Hz so it's hard to catch the little 3-5V DCS signal within it on the scope. The HPF just suppresses the 60Hz while leaving the DCS so it's easier to see. If you power your TIU with DC you can just throw the scope on AC coupling and not bother with the filter. 

It's like 3.25 MHz code rate so any cutoff frequency between like 1K and 100K will work... so any mix of 1K, 10K, 100K and 0.1uF to 1uF should get close. It's not a precision filter, it just knocks the 60 Hz down a bit.

Anyways this is what a good channel looks like:

And this is what bad ones look like:

And this is what VERY bad ones look like:

Cheers!

~Adrian

Adrain

Thanks for responding and making reference to your 2016 article.  I have been working on addressing TMCC and DCS issues on my club's layout.  We were successful in addressing the TMCC signal issues but now I am trying to address the DCS issues.  The DCS Wi-Fi app has helped with some of the issues but other issues still exist. Did you come up with any DCS "magic bullets" when you were working on your clubs layout?

Thanks again,

Bob D

rad400 posted:

Adrain

Thanks for responding and making reference to your 2016 article.  I have been working on addressing TMCC and DCS issues on my club's layout.  We were successful in addressing the TMCC signal issues but now I am trying to address the DCS issues.  The DCS Wi-Fi app has helped with some of the issues but other issues still exist. Did you come up with any DCS "magic bullets" when you were working on your clubs layout?

Thanks again,

Bob D

Hi there,

Yeah I'm sort of the defacto-DCS guy at my club too. I'm actually a circuit designer so I do a lot of fun projects with DCS. I have a lot of projects on spoofing / the track signal with FPGAs, building detectors and stuff like that that are kind of interesting. Most of it is posted on here.

In terms of issues, there are kind of three categories of DCS issues:

1. Basic DCS issues: track wiring, track cleanness, keeping the right version of everything. For this stuff you can't beat Barry Broskowitz's DCS companion books. It's basically written for everyone. This usually gets you 80% of the way there unless you have a big and complicated layout with 4-5 TIUs and special situations.

2. Intermediate DCS issues: Things like adding inductors/chokes to passenger car lighting to stop them from degrading the DCS signals, clever block detection to avoid blocking DCS and so on. The ogrforum community on here is great for this stuff... some really sharp people to follow are sandiegomark, stan2004, gunrunnerjohn, and there's others (sorry if I missed you).

3. Really complicated issues: Our club can be a bit of a zoo with lots of kids, and many trains going all over at the same time so we get a lot of derailments and short circuits (like 20+ a day... way more than most users) and it seems like it's damaging the TIU outputs as captured in the above traces. We come in and change all 5 TIUs and the signals are all strong and trains are responsive, a month later there's dead spots, and the voltages seem to have dropped out on many ports. I'm a design engineer / chip designer so I'm looking not only to figure out why they are breaking but next I will do some design work to change the driver circuitry or do current limiting or something on the TIU to solve the problem, then share the design with everyone of course!

I'm not sure how involved you want to get... but if you want a really deep understanding of the DCS track signal and what's going on for real..... there's there's a few posts I did which almost no one reads:

This one describes the track signal behavior in detail (there's a pdf of the details):  DCS signal

This one describes the encode/decode process: DCS packet format

This one describes how to capture DCS packets:  DCS Capture

This one describes how to transmit: DCS_transmit

Hope this is useful!

~Adrian

Since you have special test equipment most folks and technicians do not, I would be interest in knowing how those TIUs performed with just a separate test track.

For the degrade channels, what DCS signal strength do they give on a separate test track with a PS-2 3V and a PS-3 engine.  Do you still get 10s, or does it show up as low or no signal at all.

Without your test equipment, the only real test is the DCS signal strength test on the layout and on a test track.  For a Tech, who does not go to a home or club layout, they can only test as above.

If the test track test does in fact show a degradation, then elaborate equipment is not needed, and it would seem to me that large clubs just need to get into the routine of a Monthly bench test of the TIU.

As I stated earlier, loose internal connections, channels run at high amperage were the larger inductors get over heated and start to short and loose value can also effect signal.

There is a TVS diode mod that all REV L should have, but early Rev L may not have, and there are variations in early TIU also, because of slightly different chips selections used in the design.

I would assume your club is using all REV L but????  

I would be interested in a follow up on those TIU with the bench test. G

Has anyone tried looking at this with DC on the track?  It at least solves the problem of separating the 60hz AC from the DCS signal.

gunrunnerjohn posted:

Has anyone tried looking at this with DC on the track?  It at least solves the problem of separating the 60hz AC from the DCS signal.

Hi John,

Yeah I tried this just to check. If you run the TIU on DC and just put the scope on AC coupling you can see exactly the same thing.

~Adrian

I'm still  puzzled why the other thread disappeared , It make one think there's something MTH doesn't want us to know about the tiu. Perhaps something along the lines of Ford and their Pinto) On the other hand, threads get  deleted here all the time ,

Quote from Adrian

3. Really complicated issues: Our club can be a bit of a zoo with lots of kids, and many trains going all over at the same time so we get a lot of derailments and short circuits (like 20+ a day... way more than most users) and it seems like it's damaging the TIU outputs as captured in the above traces. We come in and change all 5 TIUs and the signals are all strong and trains are responsive, a month later there's dead spots, and the voltages seem to have dropped out on many ports. I'm a design engineer / chip designer so I'm looking not only to figure out why they are breaking but next I will do some design work to change the driver circuitry or do current limiting or something on the TIU to solve the problem, then share the design with everyone of course!

That's the difference between our layout, (4 tiu in super) . We wouldn't even come close to the number of  layout hours  you run  but our layout is protected  with 32  10 amp fast blow glass fuses. We still have an original tiu "G" from 02 still operating.

GGG posted:

Since you have special test equipment most folks and technicians do not, I would be interest in knowing how those TIUs performed with just a separate test track.

For the degrade channels, what DCS signal strength do they give on a separate test track with a PS-2 3V and a PS-3 engine.  Do you still get 10s, or does it show up as low or no signal at all.

Without your test equipment, the only real test is the DCS signal strength test on the layout and on a test track.  For a Tech, who does not go to a home or club layout, they can only test as above.

If the test track test does in fact show a degradation, then elaborate equipment is not needed, and it would seem to me that large clubs just need to get into the routine of a Monthly bench test of the TIU.

As I stated earlier, loose internal connections, channels run at high amperage were the larger inductors get over heated and start to short and loose value can also effect signal.

There is a TVS diode mod that all REV L should have, but early Rev L may not have, and there are variations in early TIU also, because of slightly different chips selections used in the design.

I would assume your club is using all REV L but????  

I would be interested in a follow up on those TIU with the bench test. G

Hey there.

Yeah everything is Rev L and all these were bought within the last month or two. 

With just the test track in isolation these weak channels "work" as long as the locomotive is not moving too fast and signal tests do come back as 10. This makes sense actually.... It sort of comes from the underlying reason that MTH uses a CDMA scheme to communicate. The point of using correlated PN codes is that it does a better job of cutting through noise (the high frequency transients generated by the train rolling along the track). On a test track there is almost no noise generated so even a weak signal will generate a good test. In a real layout with lit cars and multiple locomotives there are transients constantly being produced and then the test fails and the engines are unresponsive when the train is in areas connected to the weak TIU channels.

A second consideration is our club (like many) also runs TMCC/Legacy which is a 450 KHz carrier on the track. That signal is very weak (since it's based on RF receivers/transmitters with lots of amplification)... like 20mV (about - 34 dBv on the track), so normally if you have a good DCS channel (like 5V swing) there's no issue because the DCS packet way overpowers the tiny TMCC carrier and gets through to the engine. Once the TIU starts to fade, the 20mV becomes comparable to the DCS signal and starts to interfere with the decoding process. If you have TIUs with this weak output and pull the wire from the TMCC/Legacy setup, some of those weak sections start to respond a little better. Again the probability of successful correlation with the PN code is driven by the SNR so that makes sense. 450 KHz is within the main lobe of the DCS symbol's power spectral density (null=1 is at 3.725MHz).

For now, I have a TIUs that I plan to destroy on purpose to look at this further. I'm working on soldering monitoring taps out to the HEXFETS, mux, transformers and then I'm going to have a PC-controlled relay periodically short the outputs while the PC resets the breaker after each short and plot the I's and V's everywhere (with a 32 way analog mux) and GPIB capture on my scope.

I'm curious to see if the failure is an all at once, or slow degradation as that will help me understand how best to design around it. I'll post those results when I have them.

Cheers!

~Adrian

Gregg posted:

I'm still  puzzled why the other thread disappeared , It make one think there's something MTH doesn't want us to know about the tiu. Perhaps something along the lines of Ford and their Pinto) On the other hand, threads get  deleted here all the time ,

Quote from Adrian

3. Really complicated issues: Our club can be a bit of a zoo with lots of kids, and many trains going all over at the same time so we get a lot of derailments and short circuits (like 20+ a day... way more than most users) and it seems like it's damaging the TIU outputs as captured in the above traces. We come in and change all 5 TIUs and the signals are all strong and trains are responsive, a month later there's dead spots, and the voltages seem to have dropped out on many ports. I'm a design engineer / chip designer so I'm looking not only to figure out why they are breaking but next I will do some design work to change the driver circuitry or do current limiting or something on the TIU to solve the problem, then share the design with everyone of course!

 

That's the difference between our layout, (4 tiu in super) . We wouldn't even come close to the number of  layout hours  you run  but our layout is protected  with 32  10 amp fast blow glass fuses. We still have an original tiu "G" from 02 still operating.

 

 

I'm not much a conspiracy theorist, but I do believe in the (streisand effect) effect which this post may or may not be an example of. 

Lots of kids in the club makes trains more fun, but it does come with operational challenges of more derailments, collisions and w/e else... (which I don't mind)... so we just going to have to engineer our setup to handle those things.

I've tried fast blow fuses, but we're still seeing these weak channels show up. Even a super-fast fuse is a lifetime for some of the ICs and FETs in the TIU, so I'm not tooo surprised.

Kind of the two avenues left for us are:

1. Modify the TIU to stop whatever is being damaged from being damaged. Maybe upsize the HEXFETs or something similar. This is more the discussion I was hoping to have on this forum.

2. Stop the event that does the damage (we think short-circuits) by implementing active current limiting. It's hard because most current limiting schemes only work on DC, and only the MTH trains can run on DC... other O-Gauge trains can't (as far as I understand Legacy needs AC???). I've got a few dirty plans for active ac current limiting like this one I'm building up now... but they're all awful, and not elegant circuit solutions.

 Really.... the only way I've come up with to do active limiting is to start with a DC source and current limit that ... then have something else on that DC supply trace out the 60 Hz waveform. There's a few ways to do this but basically they all end up looking like a class AB amplifier with the train as the load, which means they will be power inefficient (50% at best) and need giant heatsinks. In the one above the P/N current mirrors limit the current to about 5A (set by the resistor on the reference side of the mirrors), then the class AB stage is driven with the 60 Hz waveform. The opamp loop is just there to null the A-to-B crossover distortion. If the track capacity is set to about 100W ish (5A X ~20V) that thing is also going to dissipate about 100W... so the club will have a new heater for the winter.

~Adrian

Interesting topic as I have a Rev L TIU where channel Fixed 1 goes weak after about 20 minutes of use. I have a small 20' x 30' 2 rail indoor 2-track main that's powered via DC. That flaky Fixed one channel has already cost me $$$ as had to send in a PS3 tender board that was bricked because that channel crapped out in the middle of a soundfile download.  Luckily the Techs at MTH could reset the board 

If it's only one channel, there are several components that could be the issue.  Check the soldering for cold solder joints on the signal transformer T301.  The 74ACT244 has been a frequent offender, U341 is the component for fixed #1.  There's also an analog mux at U502 that handles the DCS signal that could be defective.

Adrian,

I measured the Watchdog signal that the TIU puts out when powered on. I don't have a storage scope just an old school Tek 2445B and I've run out of Polaroid film packs so I have to use my cell phone.

TIU on the bench - nothing connected to it except power and scope. All four outputs on 3 TIU (Rev I, Rev I3B & Rev L) looked like the above. Two power supply voltage levels were checked (12 VDC & 16.9VDC) fed in the FIXED IN1 jacks. At 5V/div, not counting the overshoot/undershoot, the signal is about 5V peak to peak (going from -2.5v to +2.5v). Scope was AC coupled in through a 10X probe.

Mark

SanDiegoMark posted:

Adrian,

I measured the Watchdog signal that the TIU puts out when powered on. I don't have a storage scope just an old school Tek 2445B and I've run out of Polaroid film packs so I have to use my cell phone.

TIU on the bench - nothing connected to it except power and scope. All four outputs on 3 TIU (Rev I, Rev I3B & Rev L) looked like the above. Two power supply voltage levels were checked (12 VDC & 16.9VDC) fed in the FIXED IN1 jacks. At 5V/div, not counting the overshoot/undershoot, the signal is about 5V peak to peak (going from -2.5v to +2.5v). Scope was AC coupled in through a 10X probe.

Mark

 

That one looks like it's working fine! I guess your layout is in much better shape than ours. 5V peak-to-peak seems like normal for a TIU straight out of the box. I love those classic CRT scopes.

Side note: if you're interested... you see how some symbols are weaker (like the 11th one in at the top)?... that's because of the spreading code statics favor a 0 there. Digital storage scopes can't see this as easily.

Gregg,  The old post had some one put out whole chapters of a what can be considered proprietary MTH data.  You don't have published Lionel technician manuals post here either.  Those individual post could have been deleted, but Adrian put links to them in almost all of his follow up post.  So I guess that got the whole post deleted.  I can also tell you some of that data is older depending on revision and not relevant for a Rev L.  G

Adrian! posted:

GGG posted:

Since you have special test equipment most folks and technicians do not, I would be interest in knowing how those TIUs performed with just a separate test track.

For the degrade channels, what DCS signal strength do they give on a separate test track with a PS-2 3V and a PS-3 engine.  Do you still get 10s, or does it show up as low or no signal at all.

Without your test equipment, the only real test is the DCS signal strength test on the layout and on a test track.  For a Tech, who does not go to a home or club layout, they can only test as above.

If the test track test does in fact show a degradation, then elaborate equipment is not needed, and it would seem to me that large clubs just need to get into the routine of a Monthly bench test of the TIU.

As I stated earlier, loose internal connections, channels run at high amperage were the larger inductors get over heated and start to short and loose value can also effect signal.

There is a TVS diode mod that all REV L should have, but early Rev L may not have, and there are variations in early TIU also, because of slightly different chips selections used in the design.

I would assume your club is using all REV L but????  

I would be interested in a follow up on those TIU with the bench test. G

Hey there.

 

Yeah everything is Rev L and all these were bought within the last month or two. 

With just the test track in isolation these weak channels "work" as long as the locomotive is not moving too fast and signal tests do come back as 10. This makes sense actually.... It sort of comes from the underlying reason that MTH uses a CDMA scheme to communicate. The point of using correlated PN codes is that it does a better job of cutting through noise (the high frequency transients generated by the train rolling along the track). On a test track there is almost no noise generated so even a weak signal will generate a good test. In a real layout with lit cars and multiple locomotives there are transients constantly being produced and then the test fails and the engines are unresponsive when the train is in areas connected to the weak TIU channels.

A second consideration is our club (like many) also runs TMCC/Legacy which is a 450 KHz carrier on the track. That signal is very weak (since it's based on RF receivers/transmitters with lots of amplification)... like 20mV (about - 34 dBv on the track), so normally if you have a good DCS channel (like 5V swing) there's no issue because the DCS packet way overpowers the tiny TMCC carrier and gets through to the engine. Once the TIU starts to fade, the 20mV becomes comparable to the DCS signal and starts to interfere with the decoding process. If you have TIUs with this weak output and pull the wire from the TMCC/Legacy setup, some of those weak sections start to respond a little better. Again the probability of successful correlation with the PN code is driven by the SNR so that makes sense. 450 KHz is within the main lobe of the DCS symbol's power spectral density (null=1 is at 3.725MHz).

For now, I have a TIUs that I plan to destroy on purpose to look at this further. I'm working on soldering monitoring taps out to the HEXFETS, mux, transformers and then I'm going to have a PC-controlled relay periodically short the outputs while the PC resets the breaker after each short and plot the I's and V's everywhere (with a 32 way analog mux) and GPIB capture on my scope.

I'm curious to see if the failure is an all at once, or slow degradation as that will help me understand how best to design around it. I'll post those results when I have them.

Cheers!

~Adrian

This is not what I am asking, though it can be relevant.  Your scope shows a degradation directly at output with no track section involved.  What does a engine show with nothing but a 3ft section of track show on a bench.  Sitting at idle.  Does it give 10s, does it give 7-8, does it show 5 or less.  Is it a reliable test?

As a tech repairing a TIU I use a PS-2 board my board test set and the TIU.  I read each channel.  A good channel will always read 10.   If I get something less, I work on the repair.  For bad TIUs, I usually I get the engine either starting directly in conventional mod, or it may catch a watch dog, but not have DCS control of the engine.

On the other hand it you really can get 10s on a bench test, but have this issue with reduced output, how prevalent is it.  Or is it only an issue for larger layouts with lots of noise.

I know you are thinking in terms of electronic modifications to fix your issues, and I am thinking in terms of easy tests for the consumer and hobbyist to know if they have a degraded TIU.

Lastly, it is not clear to me that you identified the failure point for each TIU  You seem to dismiss my point about loose connections, inductors and such.  Are you jumping to conclusions that the short circuits have degraded the TIU?   G

Quoting GGG:

This is not what I am asking, though it can be relevant.  Your scope shows a degradation directly at output with no track section involved.  What does a engine show with nothing but a 3ft section of track show on a bench.  Sitting at idle.  Does it give 10s, does it give 7-8, does it show 5 or less.  Is it a reliable test?

As a tech repairing a TIU I use a PS-2 board my board test set and the TIU.  I read each channel.  A good channel will always read 10.   If I get something less, I work on the repair.  For bad TIUs, I usually I get the engine either starting directly in conventional mod, or it may catch a watch dog, but not have DCS control of the engine.

On the other hand it you really can get 10s on a bench test, but have this issue with reduced output, how prevalent is it.  Or is it only an issue for larger layouts with lots of noise.

I know you are thinking in terms of electronic modifications to fix your issues, and I am thinking in terms of easy tests for the consumer and hobbyist to know if they have a degraded TIU.

Lastly, it is not clear to me that you identified the failure point for each TIU  You seem to dismiss my point about loose connections, inductors and such.  Are you jumping to conclusions that the short circuits have degraded the TIU?   G

------------------------------------------------------------------------------------------------

Hi there,

1. On a channel that offers 5V of swing the "DCS signal test" provides 10 on a section of 3 foot track with a single engine sitting idle (The DDA40x I use in all my videos). A weaker 300-400mV  swing channel also produces a "DCS signal test" of 10 on the same 3 foot track section with the same engine sitting idle.

This result is not astonishing, when there is no noise to compete with, no TMCC signal to compete with, and no losses from long wiring, both a strong or a weak signal will propagate to the engine.

2. Once the same TIU is transferred into our large layout, the track sections with weak channels create difficulty controlling trains (speed can't be changed until the train leaves that block, horn doesn't sound, ... IE DCS packets aren't getting there). So yes, the difficulties of a weak output channel only become apparent when connected to a layout with lots of transient events and other signals (TMCC mostly) that the DCS signal has to compete with for successful propagation and decoding.

3. I finished the test described above last night... where a fresh TIU is taken out of the box, and I used a PC to periodically short it with a relay, then reset power to it to simulate train derailments. The test ran with a short every 5 seconds for 300 seconds total. Here are the results:

It's going to take me a day or two to go in and see exactly which component is failing (though I have my suspicions) . It's conclusive to say this simple test demonstrates that derailments do eventually damage the TIU hardware. I'll Identify exact components shortly.

Can I ask? What do the engine's signal strength test show after the TIU is shorted? (on the layout!)

I got similar results on my outside layout. I found that my brass track clamps had lost their ability to pass track power fully. Yet before and after I would check the signal with an engine and get all tens? So I decided to load the rails with engines and the signal dropped. That's when I traced it to the connections.

 The layout is much better now with all new stainless clamps on one main. Yet there are times of strange actions by the engines (run in multiples). So I'm wondering if my TIU on that main is damaged (weaker)? I had posed a question a while back about if the MTH signal test had been modified. I felt I was getting better results with the test results than with actual performance. It could be from remotes that are failing as well? Sometimes it's a software bug that bites me.

 BTW: Yes, I did try another TIU but it was purchased used. So I really can't post any positive conclusions other than some of the brass clamps don't like going thru the winters. I have another stainless track left outside for years now that uses brass clamps and the signal stays strong??? 

 I also feel that wired connections left outside all year, need maintenance or swapping out after time. It's harder to pin things down outside. Unless you redo everything and start over. That task would require a full time crew. So I'm happy when things are stable and don't have to be perfect. Every spring requires some type of work to get things stable. My single manned MOW crew seems to handle it so far. If I can't run a half dozen engines at anytime I want, I dive in and look for answers. The few trouble free years spoiled me I guess?

Also just want to mention a bit about how the signal test works given we keep talking about it and not everyone is super familiar...

The DCS has a feature called a signal test where it rates the connection to the engine from the remote as a 1 to 10. The interesting part is it doesn't actually measure signal strength at all and instead uses packet loss statistics to infer what it thinks the signal-to-noise (SNR) should be.

From classic communications all modulations and line codes have whats called a Eb/No or "SNR-BER" relationship, a relationship that describes the expected number of errors for a given signal-to-noise level. These are usually expressed in a SNR-BER plot or what's sometimes called a waterfall plot (like this one):

DCS is a polar NRZ line code so it's going to follow the "QPSK" trace in the above very closely since the differential symbol energy is about the same.

When you push DCS signal test, the engine turns off the sound and then sends 1000s of sound commands to the engine and counts how many were successfully received (IE the point you're at on the Y axis above). From that it then infers where you must be on the X-axis (SNR) and provides you a quantity that represents this from 0 to 10. The approach is okay for basic communication tests, but the SNR-BER relationship doesn't capture the effects of contested environments.

The underlying assumption in SNR-BER is the noise is an additive white gaussian noise (AWGN), but the transients of the train pickups, TMCC signals and other things typically aren't AWGN, they're deterministic  and often periodic which leads to much higher error rates. As the energy of a second signal is increased the probability of decoding error grows exponentially.

This is why the test can give you a 10 on a bench, but under perform in the contested environment of a layout. Here's a quick review of detection and probability of error PE for line codes if anyone wanted the details: line code PE tutorial

 There was a post a long time ago by Jamie, that said to use the protocast feature playing a sound track to see how good your signal really is. I remember Barry posting a reply about that. I just don't remember his comment.

More than likely it is the transmit buffer.  You are measuring transmit output, but under a train control environment you also have to worry about receiver sensitivity issues too.  That brings everything else into play.  Track, wires, engine, noises.

Typically you can get a cascading effect and one channel transmit buffer can effect other as it drags down the 5V power supply.

Would be interesting to see when the degradation occurred.  Is it gradual, or dynamic where one of the shorts kills it. 60 shorts in 5 minutes is pretty heavy hit. 

Can you explain how heavy a voltage spike you are giving during this test? G

Adrian! posted:

 

 

When you push DCS signal test, the engine turns off the sound and then sends 1000s of sound commands to the engine and counts how many were successfully received (IE the point you're at on the Y axis above). From that it then infers where you must be on the X-axis (SNR) and provides you a quantity that represents this from 0 to 10. The approach is okay for basic communication tests, but the SNR-BER relationship doesn't capture the effects of contested environments.

 

That is interesting.  I have always use a whistle/horn pushed repeatedly to get an indication of signal before running a test through the remote or to find a problem spot on the layout.   

GGG posted:

More than likely it is the transmit buffer.  You are measuring transmit output, but under a train control environment you also have to worry about receiver sensitivity issues too.  That brings everything else into play.  Track, wires, engine, noises.

Typically you can get a cascading effect and one channel transmit buffer can effect other as it drags down the 5V power supply.

Would be interesting to see when the degradation occurred.  Is it gradual, or dynamic where one of the shorts kills it. 60 shorts in 5 minutes is pretty heavy hit. 

Can you explain how heavy a voltage spike you are giving during this test? G

Yeah I'm thinking the transmit buffer also, I'm still waiting for a tape of replacements to prove that out.

60 shorts in 5 minutes is actually pretty representative of our club, you can have 2 users setting up 2 trains with 10 cars each with shorts on maybe 50% of the cars. The senior members can't see the wheels well and the younger ones aren't tall enough to see over the edge of the table.... we have all kinds.

The test setup is pretty simple:

In the test I'm just using this relay module and have the C and NO terminals directly across the TIU output. I use a signal from the PC via the old K8055 boards to close the relay, which simulates the short. The TIU is powered from the Lionel 180W brick and I just sit there and push it back in each time it pops over the 5 min. I was thinking about an electronically reset-table breaker at first to more automate the test but I want to match the breakers reaction time of the ones in the club so I just went manual.

I could use the scope with the GPIB port and take data after every relay fire to plot the signal voltage over time, but I'm not sure I want to connect my scope to all that excitement... plus I'd need to initiate a DCS packet after each short which makes it more complicated... let me think if there's an easy way to do that...

~Adrian

Hi there,

1. On a channel that offers 5V of swing the "DCS signal test" provides 10 on a section of 3 foot track with a single engine sitting idle (The DDA40x I use in all my videos). A weaker 300-400mV  swing channel also produces a "DCS signal test" of 10 on the same 3 foot track section with the same engine sitting idle.

This result is not astonishing, when there is no noise to compete with, no TMCC signal to compete with, and no losses from long wiring, both a strong or a weak signal will propagate to the engine.

2. Once the same TIU is transferred into our large layout, the track sections with weak channels create difficulty controlling trains (speed can't be changed until the train leaves that block, horn doesn't sound, ... IE DCS packets aren't getting there). So yes, the difficulties of a weak output channel only become apparent when connected to a layout with lots of transient events and other signals (TMCC mostly) that the DCS signal has to compete with for successful propagation and decoding.

3. I finished the test described above last night... where a fresh TIU is taken out of the box, and I used a PC to periodically short it with a relay, then reset power to it to simulate train derailments. The test ran with a short every 5 seconds for 300 seconds total. Here are the results:

It's going to take me a day or two to go in and see exactly which component is failing (though I have my suspicions) . It's conclusive to say this simple test demonstrates that derailments do eventually damage the TIU hardware. I'll Identify exact components shortly.

~Adrian

------------------------------------------------------------------------------------------------------------------------

Hi Adrain

Your description above can explain some of the erratic performance we are getting from our clubs layout which is close to 6,000SF.  On the test track we get readings of 10, but on the layout especially during an open house where we have a dozen trains (DCS & TMCC) running, the DCS performance is flakey.  I could have a bad TIU and still get readings of 10 under ideal conditions.

I should be receiving the components today for the filter you used to take the TIU scope readings.  With the TIUs where you saw a weak DCS signal output, did you ever investigate to find the bad component in the TIU?

Thanks for your detail explanation of the DCS operation.  It helps to diagnose an issue when you better understand how the system works.

Bob D

rad400 posted:

------------------------------------------------------------------------------------------------------------------------

Hi Adrain

Your description above can explain some of the erratic performance we are getting from our clubs layout which is close to 6,000SF.  On the test track we get readings of 10, but on the layout especially during an open house where we have a dozen trains (DCS & TMCC) running, the DCS performance is flakey.  I could have a bad TIU and still get readings of 10 under ideal conditions.

I should be receiving the components today for the filter you used to take the TIU scope readings.  With the TIUs where you saw a weak DCS signal output, did you ever investigate to find the bad component in the TIU?

Thanks for your detail explanation of the DCS operation.  It helps to diagnose an issue when you better understand how the system works.

Bob D

Hey there,

I'm actually in the middle of doing this... it's just a bad week for free time. One of the chips I designed is sitting in a spacecraft in the high-bay and spitting out error codes I don't even remember writing into it.

Looking at the TIU schematic I figure it's either the mux, the line driver, or the transformer itself (I'll give you 10:1 odds it's the line driver) .

->I doubt it's the transformer because the circuit breaker in the Lionel brick should act fast enough to protect it from getting too hot.

->I have the replacement parts for the other two (ACT244 line driver and MAX4639) but just need to find a moment to swap them. The Mux is okay but the ACT244s are sort of hard to find. They are super common in the more modern TTSOP and SOIP packages, but not in the much older SOIC package you need to fit the footprint on the TIU board. Hopefully by Monday I'll have had a chance to replace them.... It's all SMD so you need hot air... it's too difficult to get 20 pins off with a soldering iron.

6000SF sounds awesome.... It would almost make me move to NJ... except...  I don't think I could handle having seasons.

I'll keep you posted!

~Adrian (in sunny southern California)

The 74ACT244 is a common failure item when a channel loses the DCS, so that would be a possibility.  OTOH, the MAX4639 analog mux would be a component that could degrade without being totally dead, so it's a distinct possibility as well.

We have been running our TIU's in passive mode on the club layout to avoid issues of power cooking things, so far, so good.

gunrunnerjohn posted:

The 74ACT244 is a common failure item when a channel loses the DCS, so that would be a possibility.  OTOH, the MAX4639 analog mux would be a component that could degrade without being totally dead, so it's a distinct possibility as well.

We have been running our TIU's in passive mode on the club layout to avoid issues of power cooking things, so far, so good.

Honestly the way the TIU design uses the ACT244 is not great. They gang together 4 of the channels in order to lower the output impedance before going to the transformer. It may be smarter to slip something else between the 244 output and the transformer ... the question is what... driving 3-4 ohms at 12 MHz is not easy (3.7 + 3rd harmonic for squareness).

The passive connection is a good idea too. How does that work compared with the active connection?

The fact that the ACT244 is a fairly common failure item buttresses your contention that it wasn't the best component for the job.  

I hadn't looked into the actual use of the part in the circuit, I just know that I've had a couple fail, and GGG here who probably does a lot more TIU repairs than I do suggests that he replaces those on a fairly regular basis.  Given what they're doing with it, I can see why it might fail.  The interesting thing with that arrangement is the degradation of the DCS signal suddenly makes sense.  If one or two of the gates in that part fail, the signal gets weaker!  That probably would bring on a cascade effect where the remaining overloaded channels would fail soon as well.

Passive mode works fine, but you do lose the ability to shutdown all power through the TIU with the remote as the TIU no longer controls the track power.

gunrunnerjohn posted:

The fact that the ACT244 is a fairly common failure item buttresses your contention that it wasn't the best component for the job.  

I hadn't looked into the actual use of the part in the circuit, I just know that I've had a couple fail, and GGG here who probably does a lot more TIU repairs than I do suggests that he replaces those on a fairly regular basis.  Given what they're doing with it, I can see why it might fail.  The interesting thing with that arrangement is the degradation of the DCS signal suddenly makes sense.  If one or two of the gates in that part fail, the signal gets weaker!  That probably would bring on a cascade effect where the remaining overloaded channels would fail soon as well.

Passive mode works fine, but you do lose the ability to shutdown all power through the TIU with the remote as the TIU no longer controls the track power.

Yeah! I like your way of thinking about that... as gates drop out from all 4 to 0 the signal will get weaker, and you're right, the remaining buffers would have to deliver all the current and also possibly drive an unfavorable output impedance of the dead channels, further stressing the remaining gates.

I'm thinking through the advantages of the passive mode, and have some thoughts. I think it really depends on the failure mechanism of the chips. I see 2 failure cases (there may be more):

1. If the failure mechanism is the large current change during the short makes a big voltage at the transformer primary (V= L di/dt... coil voltage is proportional to change in current) and that voltage damages the buffer then being in passive mode should make the failure go away.

2. The other possibility is the ACT244 themselves don't like delivering their output signal directly into a short and fail from the stress of trying to drive 0 ohms with the 3.7 MHz squarewave  (independent of what the track power is doing). In this case I would imagine a passive connection wouldn't improve the situation.

You've convinced me that I should find time to stress test another TIU in passive mode....

-Adrian

Another point on the ACT244 usage here, I doubt it was designed with parallel outputs in mind.  One wonders if that's a problem all by itself, regardless of outside influence?  Hard to believe that the chip design was done with that kind of connection in mind.

gunrunnerjohn posted:

Another point on the ACT244 usage here, I doubt it was designed with parallel outputs in mind.  One wonders if that's a problem all by itself, regardless of outside influence?  Hard to believe that the chip design was done with that kind of connection in mind.

Exactly.... when you connect drivers in parallel there's always mismatch (drive voltage, output resistance and delay) so usually some current flows out of the best gate and into the others which stresses the device. I'm sure the guy who did the IC schematic design and layout never though about supporting that connection.

To be fair to whoever drew it up at. MTH... in the historical context of 1997... there weren't a lot of options to drive <10 ohms at multi MHz speeds.... this may have just been a lack of options given cost and time to market constraints

Is that like saying...  It seemed like a good idea at the time!

gunrunnerjohn posted:

The 74ACT244 is a common failure item when a channel loses the DCS, so that would be a possibility.  OTOH, the MAX4639 analog mux would be a component that could degrade without being totally dead, so it's a distinct possibility as well.

We have been running our TIU's in passive mode on the club layout to avoid issues of power cooking things, so far, so good.

Have you noticed any decline in DCS performance when you put your club's layout in passive mode?  Some books indicate that you may loose DCS signal strength when in passive mode.  I have tried it both ways and can't see a difference.  I have my clubs TIU"s in passive mode for a year now without a major TIU failure, but when you have a very large layout like we do, every little bit of signal helps.   I guess when I get the parts to take scope readings from Adrain's post, I will find out if there is a difference.  

Would there a logical reason why the DCS signal be lower when track power is introduced on the output side of the TIU?  

Bob D

Bob, I can't think of a reason you'd lose signal in passive mode, but that doesn't mean there isn't one.  I haven't noticed any difference on our club layout, clearly much smaller than the NJ-HR spread.  After several channels croaked in active mode, I decided to go with passive mode.

John, What else if any causes your channels to croak in active mode, outside of dead shorts. The RVHR group have been down the passive road and the signals were too weak for our 40 X 50 ft. layout. And this all took place before I changed the molex connecters to the Powerwerx connecters, which by the way are GREAT, thanks to you. We're back to active mode and the loco's are running better than ever. Our next Greenberg's show in Edison is the Thanksgiving weekend. I've been busy soldering copper jumpers across the joints where steel pins are in the modules hoping to get every crumb of signal out there. Our DCS signal seems to be getting in areas that  weren't there before and hope with the additional jumpers, things will be that much better. Also using 10 amp fuses with TVS' at the Xformer connections. Plan on getting the PSX's to give even more protection for the short circuits. Let me know when your next show is and maybe a few of us will take a ride over to see the layout and compare notes!

Steam Forever

      John

rad400 posted:

gunrunnerjohn posted:

The 74ACT244 is a common failure item when a channel loses the DCS, so that would be a possibility.  OTOH, the MAX4639 analog mux would be a component that could degrade without being totally dead, so it's a distinct possibility as well.

We have been running our TIU's in passive mode on the club layout to avoid issues of power cooking things, so far, so good.

Have you noticed any decline in DCS performance when you put your club's layout in passive mode?  Some books indicate that you may loose DCS signal strength when in passive mode.  I have tried it both ways and can't see a difference.  I have my clubs TIU"s in passive mode for a year now without a major TIU failure, but when you have a very large layout like we do, every little bit of signal helps.   I guess when I get the parts to take scope readings from Adrain's post, I will find out if there is a difference.  

Would there a logical reason why the DCS signal be lower when track power is introduced on the output side of the TIU?  

Bob D

Hey there,

I think there's a sqrt(2) factor penalty in voltage for passive mode because your signal driver is in parallel with the power supply output impedance, unlike active mode where your modulator is in series.

There could be, I probably have not used it on a large enough layout to see the difference.

Gunrunnerjohn's mention of passive mode has really got me thinking now.

Do you guys agree with this thinking, or am I missing something?

 Anecdotally people in my club were telling me that DCS started responding worse once we installed WIUs throughout the layout. At the time I dismissed this since it didn't make sense. Now I'm revisiting it with a thought experiment:

Our layout used to have all 5 TIUs setup with no aux power, and all 4 inputs of each TIU (FIXED1,FIXED2,VAR1,VAR2) were jumpered together, so the same Z4000 output drives all 4 channels. We didn't change that wiring but have added 5 WIUs to service the 5 TIUs.

In our club we have two or 3 members that run what I call the derail-a-tron 4014 which tends to short out on curves in tunnels that are not easily accessible and so the train may unfavorably sit across the rails for a good 2-3 minutes before one of our younger and more flexible members can crawl in there and fish it out. So the key is to think about what goes within the TIU PCB during the derail-a-tron extraction operation...

1. With the track powering the TIU like described above and no WIUs, when the derail-a-tron does its thing (shorts out) the breaker trips in the Z4000 and the track power is off. The TIU is off. Logic off. Power off. Everything off.  Sure damage may have been done, but it was an extremely brief transient event and now it's over (except for breaking all the handrails off the train).

2. Now with the TIU and WIU connection (with the USB port on the TIU) the logic on the TIU board is powered by the +5V coming on Pin 1 instead of track power. So the train derails, the breaker trips and track power is off while the 2-3 minute train extraction begins. So during this window we have track off, but TIU logic on. The logic doesn't know or keep track of whether it's in an active or passive connection, so it isn't alarmed that power suddenly disappeared from the TIU inputs... it just knows it has a job to do... send DCS packets to trains. So "the little ACT244 that could" (its a reference a children's book about trains) is going to try to keep swinging its load (now a derailed train) to +5 and -5V for that entire 2-3 minutes, (especially if other people who don't know the derailment happened are still trying to send commands).

It seems to me 2-3 minutes of that means "the little ACT244 that could" is going to need a new output stage.

Adrian! posted:

...Now with the TIU and WIU connection (with the USB port on the TIU) the logic on the TIU board is powered by the +5V coming on Pin 1 instead of track power...

If you have an elegant proposition to get rid of this power coming from the WIU (via USB), that would solve this problem and at least one more that I know: The more complicated reset of the TIU after the use of the Emergency Button when a WIU is present.

Would the USB work properly without Pin1 ?

Thanks for your post !