gunrunnerjohn posted:Moonman posted:just a lay person following along, but I noticed a difference in your tone generator frequency and what Dale Manquen found to be good Legacy/TMCC signal. That was 455Khz with a 5 volt excursion.That would be before the addition of the TMCC Buffer.
The 5vpp is a loaded reading with the buffer. Other wise the loaded TMCC signal w/o the buffer is in the 1-2vpp range and TMCC engines runs poorly.
Yep, with a layout like NJ-HR, there is a huge capacitive load on the buffer, and it does step on signals. That's where the DCS signal has an edge, since at the NJ-HR, they have a ton of TIU's, and each channel is only driving a small part of the layout.
Adrian! posted:rad400 posted:Adrian
The Lionel/DCS module you mentioned above, is that what you left behind with the SD3R club last week or was it something else?
Would you mind sharing the drawings on the DCS telemetry train. I would like to build one for the NJ-Hi Railers. It should make it easier to diagnose DCS signal issues on our 30'X200' layout. Currently we are starting to do DCS signal testing with a digital scope the club just picked up, but to have a scope on train wheels, that would make life a lot easier.
DCS issues: At times, hard to load an engine into the hand held/tablet or to start up a DCS engine. We get one of those engine not found messages. I will do a test on Wednesday to see if engines load easier when TMCC is turned off. Last I checked the TMCC signal it was in the 5vpp range. We also lose control of engines at various parts of the layout.
Thanks for your help!
Bob D
What I left in their club is a panel with the PSX breakers, chokes, and a rev "M" TIU with my diode boards inside. The lionel/DCS switching thing isn't even finished being designed yet! (I'm thinking maybe getting into it this weekend).
I can share the telemetry train details (maybe on a separate post), but the thing is there's a lot of C++ code that drives it. I'll try to put up a thread about it this week.
Your DCS issues are typical of poor correlation (low DCS excursion or low effective SNR) similar to everyone else. A 5V carrier underneath sounds like too much based on the measurements above. I'm thinking of doing a video tutorial on some of these effects at some point...
Adrain
I have already installed 20-PSX breakers and TVS on 10 of our Z4000/ZWL transformers. As for the the TIU diode board upgrade, I am going to hold off for the time being. In the 6 months that I have been monitoring the DCS output signal, I have not seen any "0" signal outputs. I have found some TIU DCS sig readings at 10vpp vs the 17-18vpp from a new TIU out of the box. These are all no load voltage readings at the TIU output terminals.
Look forward to your post on the DCS telemetry car & Lionel/DCS switching.
Thanks again for all your efforts in trying to get to the bottom of the issues which many of us are having with large layouts.
Bob D
I made a short summary video explaining the limitations of compatibility between DCS and TMCC/Legacy in terms of signal amplitude constraints. I'm 99% sure this is the issue we we're seeing in the SD3R club a few weeks ago. I also put a brief aside about why simple frequency filtering isn't an effective signal separation strategy with the DCS signal (a spread spectrum signal)
Thanks Adrian! Not sure I completely understand it all, but I do see the problem. Sure hope you can fine a solution, and thank you for all your efforts on the projects you have been tackling around here. I love following along even though I don't understand it all, it's still interesting.
One day we will all have DCS and TMCC/Legacy systems that work even better than they were designed to work! I was always under the impression that these two systems worked well together. You have made some really interesting discoveries about all this! Please keep up you work it is much appreciated by many here, myself included!!
Again thank for the video and explanation, now a person with no electronic experience can understand what is going on.
When there are condition where a strong TMCC signal causes an issue for the DCS signal being decoded as Adrain discussed above, or any other reason where a weak DCS signal exists, can a DCS buffer be built to increase the DCS signal? Something similar to the TMCC buffer but at 3.75 Mhz range.
When a DCS system is in speed mode, you now have one way DCS communication between the TIU and engine. So now you don't have to worry about the return DCS signal to the TIU. Putting the DCS system in speed mode you loose track signal, info, and read functions, but increased DCS signal from a DCS buffer can address many poor DCS signal issues as the TMCC buffer is addressing in the TMCC world.
For those of us who play with large & extra large layouts (4,000-6,000sf range), a DCS buffer might be the answer to our DCS issues.
Thoughts, comments?
Thanks,
Bob D
There's no technical reason that a booster for TIU carrier wouldn't be possible, but I do foresee some issues. First off, the signal is combined with the track voltage internally. Given that fact, the only way I see it working is in passive mode where you would simply be impressing the boosted signal onto the 60hz power that is applied to the output of the booster. It is possible to have a two-way device for the DCS signal, but I suspect the complexity would not be small. 
gunrunnerjohn posted:There's no technical reason that a booster for TIU carrier wouldn't be possible, but I do foresee some issues. First off, the signal is combined with the track voltage internally. Given that fact, the only way I see it working is in passive mode where you would simply be impressing the boosted signal onto the 60hz power that is applied to the output of the booster. It is possible to have a two-way device for the DCS signal, but I suspect the complexity would not be small.
It’s not so easy. The 244 is about the only line driver with the slew rate needed and the gain is low so it won’t really amplify a weak signal.
maybe a step up transformer is the way to go since you want to deliver voltage, not power to the track
I didn't say "easy". 
You do have to deliver power as well, you'll be driving quite a bit of capacitance.
Adrian! posted:gunrunnerjohn posted:There's no technical reason that a booster for TIU carrier wouldn't be possible, but I do foresee some issues. First off, the signal is combined with the track voltage internally. Given that fact, the only way I see it working is in passive mode where you would simply be impressing the boosted signal onto the 60hz power that is applied to the output of the booster. It is possible to have a two-way device for the DCS signal, but I suspect the complexity would not be small.
It’s not so easy. The 244 is about the only line driver with the slew rate needed and the gain is low so it won’t really amplify a weak signal.
maybe a step up transformer is the way to go since you want to deliver voltage, not power to the track
Adrain
Would a tuned step up transformer be easier to design than the" lionel/DCS switching device" you were talking about earlier in the post. The "lionel/DCS switching device" would address the issue when the Lionel signal was to strong for the DCS signal, but would it address the issue of just poor DCS signal strength on a layout w/o TMCC.
Would a step up transformer address both issues of strong Lionel signal & poor DCS signal?
Adrian! posted:I made a short summary video explaining the limitations of compatibility between DCS and TMCC/Legacy in terms of signal amplitude constraints.
Great explanation. Also, nice freehand, realtime illustration of a square wave superimposed on a sine!
Well done Adrian Thank you.
Lou N
Former circuit designer for Right of Way and Train America Studios
rad400 posted:Adrain
Would a tuned step up transformer be easier to design than the" lionel/DCS switching device" you were talking about earlier in the post. The "lionel/DCS switching device" would address the issue when the Lionel signal was to strong for the DCS signal, but would it address the issue of just poor DCS signal strength on a layout w/o TMCC.
Would a step up transformer address both issues of strong Lionel signal & poor DCS signal?
A transformer opens a whole can of worms related to frequency response to avoid the square wave getting badly distorted. It would need to be a very broadband one. Also you have to reject the 60 Hz power so that makes everything even more complicated. The spreading code has repeats of 4, so the frequency content is down to 3.75/4M (940 KHz) and you also need up to the 3rd harmonic to keep things squareish (12.2 MHz). That's a pretty broadband transformer... over a decade of fractional BW which doesn't seem too reasonable.
As for an active amp, the ACT244 is not an amplifier it's a line driver, so the gain is low. It expects a 5V rail-to-rail square wave in on a high impedance and then it drives that square wave into a low impedance. Since they expect rail-to-rail in they can't amplify a weak signal and make it big again, they can just translate a large voltage from a high impedance to a low one, meaning you can't really use them as a down-stream amplifier.
Opamps are not anywhere near fast enough in terms of slew rate to pass a 3.75 MHz line code with the harmonics needed to keep it square-ish. Microwave amps are tuned so they won't pass all the broadband content. Limiting amps don't have the output drive strength.
For a down-stream module, you'd need to use an RC filter to loose the 60 Hz, go into a limiting amp to restore the TTL levels on a high impedance, then go back into another line driver (like another ACT244 or stack of them) to translate back to low impedance. Then you'd need to come up with some way to make that bidirectional.
rad400 posted:Adrain
Would a tuned step up transformer be easier to design than the" lionel/DCS switching device" you were talking about earlier in the post. The "lionel/DCS switching device" would address the issue when the Lionel signal was to strong for the DCS signal, but would it address the issue of just poor DCS signal strength on a layout w/o TMCC.
Would a step up transformer address both issues of strong Lionel signal & poor DCS signal?
This is correct. Making the DCS signal voltage excursion larger solves the a weak DCS signal and overcomes a tmcc signal inflating the correlator floor. The switching device I'm working on just stops the tmcc signal from being involved. However making the DCS signal larger would be really complicated.
Adrian,
We have a DCS on/off switch for each of our four separate tracks at theSD3R layout. I am not sure of the effectiveness of this switch since we installed the "rev M" board. But I was thinking: what if I used this switch to block the TMCC signal for a particular track. Granted that the TMCC signal "wafts" itself over multiple tracks, but if I block the direct feed to a particular track it may be enough to allow the DCS signal to prevail for that particular track. Granted that I may have to make some changes in the "common ground" in our layout, but I think this is problematic in several ways anyway.
Roger L. posted:Adrian,
We have a DCS on/off switch for each of our four separate tracks at theSD3R layout. I am not sure of the effectiveness of this switch since we installed the "rev M" board. But I was thinking: what if I used this switch to block the TMCC signal for a particular track. Granted that the TMCC signal "wafts" itself over multiple tracks, but if I block the direct feed to a particular track it may be enough to allow the DCS signal to prevail for that particular track. Granted that I may have to make some changes in the "common ground" in our layout, but I think this is problematic in several ways anyway.
This could work!
If you wanted to make it even more effective to stop the track-to-track coupling, you can change to a SPDT switch and have the track on the common post toggle between the TMCC feed and earth ground on the two switch poles. If the outer rails are earth grounded the TMCC shouldn't be able to couple in a meaningful way.
Adrian! posted:rad400 posted:Adrain
Would a tuned step up transformer be easier to design than the" lionel/DCS switching device" you were talking about earlier in the post. The "lionel/DCS switching device" would address the issue when the Lionel signal was to strong for the DCS signal, but would it address the issue of just poor DCS signal strength on a layout w/o TMCC.
Would a step up transformer address both issues of strong Lionel signal & poor DCS signal?
This is correct. Making the DCS signal voltage excursion larger solves the a weak DCS signal and overcomes a tmcc signal inflating the correlator floor. The switching device I'm working on just stops the tmcc signal from being involved. However making the DCS signal larger would be really complicated.
Tell me more about this switching device? One of the principal benefits of the current structure is the ability to run TMCC/Legacy and DCS at the same time on the same tracks. Are you maintaining this capability with your switching device?
Adrian! posted:rad400 posted:Adrain
Would a tuned step up transformer be easier to design than the" lionel/DCS switching device" you were talking about earlier in the post. The "lionel/DCS switching device" would address the issue when the Lionel signal was to strong for the DCS signal, but would it address the issue of just poor DCS signal strength on a layout w/o TMCC.
Would a step up transformer address both issues of strong Lionel signal & poor DCS signal?
This is correct. Making the DCS signal voltage excursion larger solves the a weak DCS signal and overcomes a tmcc signal inflating the correlator floor. The switching device I'm working on just stops the tmcc signal from being involved. However making the DCS signal larger would be really complicated.
Adrain
Thanks for all your efforts in trying to address these DCS issues. The four threads you are currently running on DCS are great and helps many of us to better understand the technical portion of DCS. I am very interested to learn more about the DCS/Lionel switching device and try it out at NJ-HI Railers.
rad400 posted:Adrain
Thanks for all your efforts in trying to address these DCS issues. The four threads you are currently running on DCS are great and helps many of us to better understand the technical portion of DCS. I am very interested to learn more about the DCS/Lionel switching device and try it out at NJ-HI Railers.
Hey,
Here's the idea of the switching device I'm working on...
The basic idea is to detect the very first instant when a DCS packet appears and turns on a switch that blocks the TMCC/Legacy signal, but only for the very short (millisecond time) while the DCS packet is going through. The TMCC/Legacy system won't notice the short delay since it's like 0.001 seconds per second or so, and the DCS system won't feel the effects of the legacy signal since it isn't present during the packet.
The hard part is detecting the packet quickly enough to turn on the switch. The DCS bit period is about 260ns (3.75 MHz), so you'd want to do the detection and turn on in like a 10th of that time... 26 ns. The only thing that can run that fast is an FPGA which is what I'm using.
Here's a doodle of what I've been testing so far:
How are you detecting the DCS signal in 26ns? Are you just looking for a fast edge from the track? Given the fact that you could have multiple TIU channels and multiple TIU's, would you have to do this detection for each output?
BTW, you keep saying 3.75mhz for the DCS signal, but all the stuff I've read on DCS says 3.27mhz. Where exactly are you getting the 3.75mhz figure? Just curious what the base carrier frequency actually is.
gunrunnerjohn posted:How are you detecting the DCS signal in 26ns? Are you just looking for a fast edge from the track? Given the fact that you could have multiple TIU channels and multiple TIU's, would you have to do this detection for each output?
BTW, you keep saying 3.75mhz for the DCS signal, but all the stuff I've read on DCS says 3.27mhz. Where exactly are you getting the 3.75mhz figure? Just curious what the base carrier frequency actually is.
I'm luckly the SD3R club I'm working on this for only has 1 TIU so I only need a 4 channel ADC. You're right, I'm triggering off the first rising edge. If I was doing this for my own club it would be epically complicated as I would have to coordinate 20 triggers!
I dunno, in the old post with my FPGA decoder I found 7.50 MHz works best which seems to imply a symbol rate of 3.75 MS/s. I'm sure I didn't read anything, I probably just measured it and copied it.
It looks like you're shorting the TMCC signal directly to earth ground. That may not be good for the drivers in the TMCC command base, and I'm pretty sure my upcoming TMCC buffer won't be happy that's happening either. Might I suggest you open the signal at the command base (or buffer output) and short the outside track to earth ground, it'll be less traumatic for the TMCC/Legacy base or buffer.
gunrunnerjohn posted:It looks like you're shorting the TMCC signal directly to earth ground. That may not be good for the drivers in the TMCC command base, and I'm pretty sure my upcoming TMCC buffer won't be happy that's happening either. Might I suggest you open the signal at the command base (or buffer output) and short the outside track to earth ground, it'll be less traumatic for the TMCC/Legacy base or buffer.
You're probably right. The left pull down device can go away I guess. The FETs come in packs of 2 so I was trying to be sophisticated
I think abusive is the better term, at least in terms of the poor TMCC signal. 
The key with this unit is probably putting it really close to the command base so that you don't get any radiated track signal as that actually can affect things. However, with the outside track returned to earth ground, that should deal with most of that kind of issue.
gunrunnerjohn posted:How are you detecting the DCS signal in 26ns? Are you just looking for a fast edge from the track? Given the fact that you could have multiple TIU channels and multiple TIU's, would you have to do this detection for each output?
BTW, you keep saying 3.75mhz for the DCS signal, but all the stuff I've read on DCS says 3.27mhz. Where exactly are you getting the 3.75mhz figure? Just curious what the base carrier frequency actually is.
Okay folks.... While conceptually cute this idea is not going to work because of basic referencing. I got 1/3 of the way done before catching this.
The digital waveform on the TIU is referenced to the Layout ground, but the TMCC signal is referenced to the earth ground. So if you build a digital circuit (microcontroller/FPGA) referenced to the layout ground to grab the TIU waveform, then you won't be able to control a transistor because your DCS bottom is the legacy's top and you can't establish cutoff potential on the switching device.
If you sense the DCS with reference to earth ground that has to consider the 3.75 MHz signal flowing through the single legacy base wire, the base itself and the building wiring, so it really doesn't look like anything.
It's time to start looking for fast opto-couplers but 26ns propagation delay seems ridiculous.
I don't find any opto couplers with that speed, that's pretty fast. Here's one that looks possible...
gunrunnerjohn posted:I don't find any opto couplers with that speed, that's pretty fast. Here's one that looks possible...
Okay a new plan has been formulated:
It's going to be the 3 pole RC filter to trigger a resettable monostable multivibrator using a 74HCT123. to trigger the DCS waveform off each TIU output. Then a galvanic isolator to leave the detector part that is referenced to the layout ground and enter the control side referenced to earth ground. So it'll need two isolated 5V supplies, one for each side (5Va and 5Vb).
I still need to think about the switching element. I'm on the fence between a simple jfet, and those TI analog mux products. Either way there needs to be bias under the TMCC signal to get the transistor to the operating point. Probably just (R+R)||C.
It's 25ns on the one-shot chip and 15ns on the coupler so we're about 15-20% into the period of the first bit already (266ns). Parts are ordered so I guess I'll breadboard it next week.
Is that the fastest one-shot you can find? How much delay in the filter?
gunrunnerjohn posted:Is that the fastest one-shot you can find? How much delay in the filter?
The filter is negligible, like 2-3ns. That one-shot is DIP so it's good for prototyping!
I look forward to the results. I doubt the brief interruption of the TMCC signal should affect TMCC/Legacy, they send multiple copies of commands.
gunrunnerjohn posted:I look forward to the results. I doubt the brief interruption of the TMCC signal should affect TMCC/Legacy, they send multiple copies of commands.
Some slow progress and updates on this. Here's the CMOS variant of that one-shot (CD74HCT123E) chip triggering off a DCS waveform through that RC filter. Good clean edges around the packet and no misfires in the middle. Blue is the output, yellow is input DCS. The initial delay is like 8.5ns end to end, but I still need to propagate this through the idolator and into the switch (about 18ns of margin remaining). I used Cx=10n, Rx = 10K, so that's nominally a 50us box per trigger received.
Overall shape:
Leading edge behavior:
Trailing edge behavior:
Packet procession behavior:
This detector is the hardest part (I think). The next is to transfer the reference from the layout ground (for DCS) to the earth ground for TMCC using a galvanic isolator, and then developing a graceful switching circuit that doesn't short out the TMCC base. It's going to look weird because it needs two 5V supplies referenced to the different grounds.
So for everyone else following this the idea who may be a little lost....the concept is we're building something that detects when DCS (the yellow waveform) is communicating between the TIU and train, and generating the blue box which will eventually inhibit the TMCC signal. Since the DCS packet is only about 1/500th of a second long the TMCC shouldn't notice the short time it's being inhibited. At the same time the DCS decoder performance improves because there isn't a 450 KHz TMCC signal inflating the apparent noise floor.
This looks like something that could be incorporated into my upcoming TMCC buffer, kill two birds with one stone.
Speaking of the TMCC Buffer, I was thinking about this issue and the increased amplitude signal from the TMCC buffer.
The basic idea behind the TMCC Buffer is to provide a lower impedance drive to drive the capacitance of a large layout. It increases the TMCC signal amplitude and uses a lower impedance driver to accomplish that goal. Adding this circuitry would allow us to simply cutoff the TMCC signal from the buffer during the DCS packet interval.
gunrunnerjohn posted:This looks like something that could be incorporated into my upcoming TMCC buffer, kill two birds with one stone.
Speaking of the TMCC Buffer, I was thinking about this issue and the increased amplitude signal from the TMCC buffer.
The basic idea behind the TMCC Buffer is to provide a lower impedance drive to drive the capacitance of a large layout. It increases the TMCC signal amplitude and uses a lower impedance driver to accomplish that goal. Adding this circuitry would allow us to simply cutoff the TMCC signal from the buffer during the DCS packet interval.
So how about this...
1. When I do my board I'll put terminals with the Earth-referenced TTL out.
2. Then on your booster you can have a TTL in port that inhibits the driver.
That way we'll be compatible-ish.
Of course this clever trick will only work with 1 TIU. I think in our club the physical delay (speed of light) between the 5 TIUs is already way more than 25ns.
Hmm... The whole point of this is mostly for large layouts, that's the only place that the TMCC Buffer makes sense.
I haven't researched this, but if you have multiple TIU's, don't all the commands go out on all channels? The TIU doesn't know which segment that the locomotive is, and it could have transitioned to a different place. I suppose there's enough time for the signal to have skew based on each TIU has to process the incoming remote command individually.
I did have another thought.
The TMCC buffer increases the signal amplitude in an effort to combat the increased capacitance of a large layout. The output buffer is capable of driving about a .1uf capacitance of the outside track to earth ground. One of the issues is the amplitude, the buffer has a 3:1 gain. I wonder if dropping the gain to unity and adding a larger isolation capacitor would allow it to drive the same capacitance but have less effect on the DCS signal? This TMCC Buffer Schematic.pdf is what we have now.
gunrunnerjohn posted:Hmm... The whole point of this is mostly for large layouts, that's the only place that the TMCC Buffer makes sense.
I haven't researched this, but if you have multiple TIU's, don't all the commands go out on all channels? The TIU doesn't know which segment that the locomotive is, and it could have transitioned to a different place. I suppose there's enough time for the signal to have skew based on each TIU has to process the incoming remote command individually.
I did have another thought.
The TMCC buffer increases the signal amplitude in an effort to combat the increased capacitance of a large layout. The output buffer is capable of driving about a .1uf capacitance of the outside track to earth ground. One of the issues is the amplitude, the buffer has a 3:1 gain. I wonder if dropping the gain to unity and adding a larger isolation capacitor would allow it to drive the same capacitance but have less effect on the DCS signal? This TMCC Buffer Schematic.pdf is what we have now.
Inside the TIU there's a mux (U502) which steers the FPGA to each of the each of the 4 channels one by one. Although everything happens together, if you look close you can see different channels (VARx, FIXEDx, ..) are on different time slices.
You could try to do a blanket window over all of them, but now your talking 0.2 seconds and TMCC/Legacy might start to take notice. Also physically... if you have two TIUs and either of them is more than 20 ft away from the TMCC input, that's already a good 23ns of delay right there.
Both DCS and Legacy are voltage-mode receivers so you don't want to lower the gain of that driver or you're directly dropping your SNR on TMCC.
A last one we could also think about (but it's really hard compared to the the switching idea) is to induce a copy of the TMCC signal on the center rail that closely mirrors the one induced on the ground. Nominally the TIU and Transformers are in common mode to the injected legacy signal meaning both + and - outputs of the TIU/transformer are going up and down at 450 KHz wrt the Earth ground and with the same phase and amplitude on center and running rails. Of course practically this is not the case since the capacitance of each side (center and running rails) to Earth ground is a little different and the circuitry inside the transformers and TIU aren't perfectly symmetric. Also it can be different at different locations along the layout due to series R and L of the track and wiring.
If the Legacy signal is very well balanced on both rails, the DCS decoder which senses things differentially should not see it, as it's on both ground and power rails. That's the operating principle of DCS when legacy is present.
So could we build a small circuit/board/module that senses the differential 450 KHz at some points along the track and copies the TMCC onto the center rail with just enough amplitude to balance them.
This would probably be feedback based...
-Have a diff amp across the rails with a high-Q 450 KHz filter in front
-Have some type of envelope detector on the filter to get a DC-ish value from the legacy tone
-Use that envelope DC to control a variable gain stage like your booster connected to the center rail and make them balanced.
This is probably an even better idea than what I'm doing, but it's pretty hard.
Adrian! posted:Here's a doodle of how that might look:
Also you're pretty-much relying on the BPF to keep the DCS transients out of the loop. I doubt it'll be so good when the the inputs are being yanked +/-7V with ns rise and fall times. We could make the loop time constant really really really long (like seconds) so the DCS timescales don't affect where it settles.
Also^2 unlike the outside rail, the center rail is in blocks, so you're going to need a lot of these for a big layout.
One wonders if there's not an easier method. How about a low-pass filter that simply couples the outside and center track together. Obviously, it would have to pass the 455khz TMCC signal and block the DCS signal lowest frequency component, that's around 1mhz, right? With that approach, you could do each DCS block.
Access to this requires an OGR Forum Supporting Membership
