Adding DCS to a conventional layout

Phil, do you have a DMM and have you measured (DC Volts mode) the voltage at the Z4000 output and TIU variable output? 

If you're getting the whistle ON continuously I'm thinking you'll see DC voltage on the track.  I don't think it's glitches, spikes, or fast-edges that is the problem but a "slow" DC offset that will be detectable/measurable by a basic meter ... no oscilloscope required.

Stan:  Done that already:  DC voltage = 0. [Done with no trains on track, max Z-4000 voltage.]

To be clear, you get 0V DC on the track at the TIU Variable Output (with Z4000 set to 22V AC at the TIU Variable Input)?

Did you happen to measure the track voltage DC with the Horns blasting away?  And if so does it change as you lower the Z4000 voltage feeding the Variable Input to 16V AC (when it the Horns stop)?

Phil, You never answered my question about what type of trains.  I did find your original post of the problem on page 2.   It is possible you have a control board problem on the Z-4000.  A leaky capacitor that is signaling a whistle command at those higher voltages.  I would think it would do it with or without TIU, but maybe you never approached those high voltage directly from the Z-4000 without the TIU because the trains would fly off the track.  Not sure why it would reflect on both tracks as described, but I would go back and make sure you have this track wired properly to the terminal block and from there to the track sections.  NO Loose connections. or crossed wires.  To confirm hypothesis you can swap in another Z-4000 and than swap in different TIU to see if symptoms change.  Similar to how we isolate TMCC board problems on those engine on the phone.   G

To GGG & Stan:  B 4 replying to your latest Q's,  I have 2 of my own, after stating that the chopped waves that GRJ reported with the whistle ON clearly show an average voltage [over one cycle, say] that is positive.  [It's 0 when the whistle is off.]

Q1.  Does this average voltage correspond to, or even DEFINE, the DC "offset"?

Q2.  How do/can the various types of engines [PW, TMCC, DCS] detect this in order to blow the whistle?

            Now to answer some of yours:      Stan:  A1:  Yes.   A2.  No

George: The spurious whistle occurs for any & every engine tested;  even an MTH PS 3.   As reported, I have swapped the TIU.  We have no access to another Z-4000;  tho' I suspect your "leaky capacitor" hypothesis has lots of merit and will be further strengthened by my latest experimental results [to be released 'soon';  after I'm more confident in them!]  Also, track wiring is 'perfect'.

Today's expts were full of surprises.   Like 'hysteresis'?  I'll try to summarize them tomorrow. 

Stan & George:  Have you seen my 2 Q's?

Meanwhile, I'm working on a report of yesterday's confusing expts!!

Here's the promised summary of yesterday's efforts:

1ST SERIES

1. Put DCS PS3 engine [SP 5672, GP-9, which was already in the remote] on TK 1, with cap on TIU output. Power-up the Z4000 to ~ 20VAC, with remote pre-set @ 21.5. Engine started in CONV , as it should. [Because TIU watchdog/handshake signal was long gone.]
Now the SURPRISE: It would not start-up in DCS; ENGINE NOT ON TRACK……[Tried several times.]
2. DELETE ENGINE [Done], followed by ADD DCS ENGINE: “No engine to add.” I also tried this several times, to be sure.
3. Remove cap. ADD DCS ENGINE. Success! [Engine is now back in the remote.]
4. Move engine to TK 2, with cap on TIU output; set remote to 21.5; power-up the Z to ~ 20VAC. Result: ENGINE NOT ON TRACK
5. Remove cap; try again. STILL no success. [ENGINE NOT ON TRACK]
6. Power-down, then re-power-up: Engine started properly [DCS mode]. TIU was not shut down.

Tentative conclusion: Caps on TIU output preclude [or at least interfere with] DCS signal. If they’re on TIU input, all is ‘cool’. [Perhaps the problem is associated with high-voltage on the Z. I don’t think I tried these with lower a setting….] Is there a memory/hysteresis effect?

2ND SERIES

1. No caps. Old [not DCS] MTH engine [Pennsy 6743, a 4-8-2] on TK 1, new MTH engine [SP 5672] on TK 2 and started in DCS. Both in neutral. Z set to ~ 20 VAC; remote @ 21.5. Results: NO spurious whistles; another SURPRISE. However, the Pennsy whistled briefly, every 10 sec or so…..SP 5672 was silent.
2. Ran the Pennsy via the remote: Continuous whistle.
3. Added cap to TIU input on TK 1. Whistle stopped.
4. Deleted SP 5672. Tried to re-add: No engine to add. [Z voltage to TK 1 was 0.]
5. Remove cap. Removed Pennsy from TK 1; replaced by SP. Z voltage to TK 1 set to 22; Z voltage to TK 2 set to 0. ADD DCS ENGINE: No engine to add. Ditto with TK 2 Z-voltage @ 22….But now it’s followed by OUT OF RF RANGE.
6. Power-down RR. Re-power-up, 22 VAC on Z, 21.5 on remote. Now engine 5672 was found, and started successfully.
7. Install caps on TIU input. 5672 still runs fine…..as does Pennsy onTK 2, controlled via remote with 22 VAC from Z.
8. If cap on TK 2 is removed get spurious whistle unless Z-voltage is < ~ 17.

3RD SERIES

1. No caps. SP 5672 removed. Pennsy on one track , Lionel SFe steamer 8642 on other. > 19 VAC from Z. Run engines with remote.
2. Press whistle on remote: TK 1. BOTH whistles blow! Yet-another surprise.
3. Reduce both Z -voltages to ~ 16 and whistles stop.

4TH SERIES

1. Add caps to both TIU inputs. AGAIN, need Z -voltage < 18 to preclude spurious whistle.
2. Switch caps to TIU outputs. Again, both whistles blow @ > 19 VAC from Z.
3. Caps @ TIU input, again. Now, no spurious whistles. But,
4. Hit whistle on remote for TK 1: BOTH whistles blow.
5. Hit whistle on remote for TK 2: Only engine on TK 2 blows.
6. No spurious whistles if Z-voltage < 16-17.
7. Now Z-whistle command blows both unless Z-voltage < 17 VAC. Another surprise?

I think that covers it. The only thing I can think of doing next is to replace the Z-4000 with the old Lionel ZW and run these again…….

Phil, NEVER-TRUST-AN-ELECTRON, Gresho

The cap is known to affect the DCS signal, it's for running conventional. 

A HA!! Now he tells me.  Do you have advice for running both?

E.g. 1. Retain caps on TIU inputs.

2.  Keep Z-voltage < 17.

GRJ:  Can you answer the 2 Q's I posed for Stan & GGG yesterday?

I'd be curious what a 22uh choke would do in series with the cap across the rails.  That might handle the spike while still letting the 3.27mhz DCS signal through.  OTOH, it might damp the spike as well, I'd have to give it a try.

phil gresho posted:

A HA!! Now he tells me.  Do you have advice for running both?

E.g. 1. Retain caps on TIU inputs.

2.  Keep Z-voltage < 17.

Well, maybe now "he" tells you, but you were told of this phenomenon:

https://ogrforum.com/t...73#59713325364564273

Time is up, put down your #2 pencil, and close you test booklet.  From what I can tell, you can meet your objectives by doing exactly what you say:

1. Put caps on TIU inputs

2. Keep Z-voltage <17V.  As was pointed out earlier, having the DCS handheld displaying an incorrect voltage is immaterial...and you get better speed control resolution if starting from 16V AC (or whatever) at the TIU input.

phil gresho posted:

To GGG & Stan:  B 4 replying to your latest Q's,  I have 2 of my own, after stating that the chopped waves that GRJ reported with the whistle ON clearly show an average voltage [over one cycle, say] that is positive.  [It's 0 when the whistle is off.]

Q1.  Does this average voltage correspond to, or even DEFINE, the DC "offset"?

Q2.  How do/can the various types of engines [PW, TMCC, DCS] detect this in order to blow the whistle?         

Q1. Yes, for purposes of this discussion the arithmetic average or mean value of the alternating voltage is the DC offset.

Q2. When the DC offset exceeds some positive threshold the whistle should blow.  Obviously there is an assumed "integration" interval or, in electrical terms "filtering" that exceeds 1 full cycle so that both polarities of the 60 Hz voltage can balance each other out.   Otherwise the detector would fire 60 times per second.  There are guidelines with voltage thresholds and timing but getting into specifics opens a can of worms so I'm not going there.

The methods of detecting this DC offset (in the presence of a 60 Hz waveform of much larger magnitude) have changed over the years.  Some of the electro-mechanical methods were described in detail earlier.  I don't think any new design in this century would use an electro-mechanical detection method.  Instead it would be electronic with a simple resistor and capacitor performing the integration of the AC track voltage; the result would be a DC voltage that simply triggers the whistle when it exceeds, say, +3V.  If the engine has a microprocessor with an analog-to-digital converter (as in any modern command control engine), then the processor can sample the track voltage magnitude and perform the integration in software.

Stan:

1.  I'm sorry; where was I "told of this phenomenon"?

2. It seems that both of us overlooked the point that the caps, ANYWHERE, I guess, would preclude running a DCS engine!

          Am I back to 'bare bones' ops:  NO caps; voltage < 17?

".....with a simple resistor and capacitor performing the integration of the AC track voltage"

This is VERY INTERESTING....Now I need to figger out the mathematics of this trick!  Do I start with R & C in a series ckt, or is it a parallel one?  I DO know how to solve the associated differential eqns.  I'll also need initial condititons, of course..... for the 2nd-order DE.

Please give me [another] home work problem!

I might have screwed up the link in my recent post which I changed so may work now, but the link should be back to an earlier post where I said, quote:

"While you are running conventional mode on your Variable channels, placing a cap right on the track has the annoying behavior of degrading the DCS command signal.  So for some it could be a bother to remove/install that cap if alternating between command/conventional on a Variable output."

From what I understand, you can put the caps on the TIU Inputs and can still operate in DCS command on that channel.  Additionally, this cap can apparently improve conventional behavior (reduce or eliminate the spurious whistles) irrespective of track voltage.

phil gresho posted:

".....with a simple resistor and capacitor performing the integration of the AC track voltage"

This is VERY INTERESTING....Now I need to figger out the mathematics of this trick!  Do I start with R & C in a series ckt, or is it a parallel one?  I DO know how to solve the associated differential eqns.  I'll also need initial condititons, of course..... for the 2nd-order DE.

Please give me [another] home work problem!

Are you serious?!  If I say Laplace transforms are involved will you go away quietly?  LOL.

It's a series RC circuit.  It would be extremely difficult to mathematically model your exact situation because the chopping of the sinewave is difficult to represent mathematically - especially since the Z-4000 is a close-but-no-cigar to a mathematically ideal "A sin (x) "

phil gresho posted:

Problem solved!!  From GUNNRUNNERJOHN:   "Another thing that tames the beast is a 10uf non-polarized cap across the tracks, it apparently rounds the chopped waveform just enough to prevent the false triggering."

THANK YOU GRJ; AND STAN.  Installation of the capacitors,  via banana plugs across the 2 Z-4000 outputs, has cured ALL ILLS.  I can even set the handles at 22 VAC or more,  with no spurious whistle.

One last point:  The remote will not blow the whistle when set at MAX [22 VAC].  Back it off 1 click to 21.5 and all is well.  [I've been told that this is normal.]

My plan,  which I shall put into s short OPERATIONS MANUAL for the club members, will be to set the Z-4000 handles to 18VAC before using the remote for speed control.

          Thx again, to all who contributed.

 

Why is this thread still going on? Seems like the OP had his answer a long time ago.

Stan:  I don't need no stinkin' transforms!  I solve these in the more 'conventional' way....at least for a sinusoidal driving force.  I've done it many times...In fact,  I wouldn't be surprised if I've solved this exact problem.  I'll look thru my old notes.  Now I think I see what you meant by  "performing the integration".  You mean simply SOLVING the 2nd-order ODE;  right?

Keith:  If this thread 'bores' you,  just bug out.  Personally, I'm enjoying it immensely....Different strokes for different folks.  [What's an OP?]

OP is the original poster; the person who started the thread.

You.

phil gresho posted:

Stan:  I don't need no stinkin' transforms!  I solve these in the more 'conventional' way....at least for a sinusoidal driving force.  I've done it many times...In fact,  I wouldn't be surprised if I've solved this exact problem.  I'll look thru my old notes.  Now I think I see what you meant by  "performing the integration".  You mean simply SOLVING the 2nd-order ODE;  right?

Keith:  If this thread 'bores' you,  just bug out.  Personally, I'm enjoying it immensely....Different strokes for different folks.  [What's an OP?]

Sorry Phil; I thought you were becoming annoyed, frustrated. My misunderstanding.

phil gresho posted:

Stan:  I don't need no stinkin' transforms!  I solve these in the more 'conventional' way....at least for a sinusoidal driving force.  I've done it many times...In fact,  I wouldn't be surprised if I've solved this exact problem.  I'll look thru my old notes.  Now I think I see what you meant by  "performing the integration".  You mean simply SOLVING the 2nd-order ODE;  right?

More power to you (and good luck with that) if you think you can perform circuit analysis without the use of transforms.  Solving ODE's is used for only a brief time in a college design course - just enough to make you think all that time spent studying calculus or whatever has practical application.  (Is this the LOL emoticon?)  Anyway, if coming from a mathematical world the wake-up call is that very little in circuit design engineering is linear in the mathematical sense - particularly when semiconductors are involved.  So now enter the world of non-linear differential equations which are hard to "solve."  Instead, circuits are simulated (not solved in the closed-form sense) in the design phase using non-linear models. 

So in the case of a basic DC offset detector, the actual "decision" is made by a semiconductor (very non-linear) device.  Or even if an electro-mechanical relay makes the "decision" it is a challenge to represent its non-linear behavior with an ODE - for example relays typically turn on at one voltage but turn off at a lower voltage.  Call it hysteresis or whatever but it is non-linear. 

Stan:  Thx for the additional info.  2 more points:

1.  The ODE is simpler that I thought.  It's only 1st order, not 2nd. [I was thinking of RLC ckts.]

2.  I could, and probably will, perform a LINEAR analysis by utilizing  GRJ's result for 21.5 volts with the whistle ON.  It's just several segments of a sine wave.  I can generate a solution in each segment and combine them to get a full cycle.

This I would propose as a linear approximation/model for a [nonlinear] chopped wave....Make sense?

But, FIRST I must be sure about semantics.  If I solve the ODE for a series RC ckt with a segmented, piece wise, sine wave driving function, is this EQUIVALENT to what you call 'integration'/filtering?

This was an exhausting topic to follow...gonna go hug my z1000 and tiu

Where is this going? I guess we should all leave this to the mathematicians. I've lost track of the original subject.

As long as you make the simplifying assumption that it is a linear system you can do as you say.  This is a fundamental principle of circuit analysis called superposition which means you can break down the input signal into parts, solve for the responses to the individual parts, then add the result. 

As for integration/filtering, what I mean is you need to solve the equation for a long enough time until there is no more useful information learned.  I don't know what it's called in mathematics but in electrical engineering the differential equations are arranged in a manner where coefficients represent the real-world parameters such as the value of resistors, capacitors, inductors, etc.  Then the back-of-envelope rules-of-thumb are applied - for example for a first-order response of just an R and C, you might want to run your solution to, say, 10 times the product of R (measured in Ohms) and C (measured in Farads).

Note that to calculate the DC response to an asymmetrically chopped sine wave (as the TIU variable channel creates), you need to know the R and C of whatever is in the circuit in the whistle offset detector.  You don't know these values!  I don't know or want to know all the different engine types you have but I assure you your PS3 engine does not use a R-C circuit to detect the DC offset in conventional mode; it uses a microprocessor sampling the track voltage and calculating the offset in software.  The 10uF capacitor that you have been using is part of a completely different circuit that has been speculated as a means to "round the corners" of any spikes or glitches that may (or may not) exist on the track voltage.

GRJ:  Have you [Will you] found [find] the time to replace the Lionel power supply with a Z-4000?

Final[? Ha!] summary & plan:  More testing with the caps on the Z4000 output [ = TIU input] seems to be working fine.  I can start & run an MTH PS3 engine, and I get no more spurious whistles....even at 22VAC on the Z.

So my advice to the club guys, with some conservatism, will be to set the Z handles @ 15-17 VAC when using the remote for voltage control.

BUT;  since I'm having some REALLY-INTERESTING dialogs with Stan,  I will not yet let this thread fade away.....Many of you will, quite understandably, bow out, as future posts may be a bit too mathematical.

Stan:  In response to your last post,  while I have begun the analysis of GRJ's scope pix @ 21.5 v with the horn ON, I DO see the possible 'uselessness' of the analytical results.....UNLESS it's true that the time constant, RC, is much less than the oscillation period, which of course IS known [1/60 sec].

Can you [or anyone else, enlighten me here]?  If RC << 1/60,  I could use a steady-state approximation, and thus obtain some useful results!

phil gresho posted:

...I DO see the possible 'uselessness' of the analytical results.....UNLESS it's true that the time constant, RC, is much less than the oscillation period, which of course IS known [1/60 sec].

Can you [or anyone else, enlighten me here]?  If RC << 1/60,  I could use a steady-state approximation, and thus obtain some useful results!

I suppose 'uselessness' is one way to look at it. RC  >> 1/60 for the circuit to work as intended. 

As mentioned previously, "solving" linear differential equations for the purpose of circuit design/analysis is only a fleeting moment in the electrical engineering curriculum.  A practicing engineer would be schooled in other design methodologies.

stan2004 posted:

phil gresho posted:

...I DO see the possible 'uselessness' of the analytical results.....UNLESS it's true that the time constant, RC, is much less than the oscillation period, which of course IS known [1/60 sec].

Can you [or anyone else, enlighten me here]?  If RC << 1/60,  I could use a steady-state approximation, and thus obtain some useful results!

I suppose 'uselessness' is one way to look at it. RC  >> 1/60 for the circuit to work as intended. 

As mentioned previously, "solving" linear differential equations for the purpose of circuit design/analysis is only a fleeting moment in the electrical engineering curriculum.  A practicing engineer would be schooled in other design methodologies.

I solve most of my AC circuit problems with

V=L*di/dt and I= C*dv/dt

stan2004 posted:

As mentioned previously, "solving" linear differential equations for the purpose of circuit design/analysis is only a fleeting moment in the electrical engineering curriculum.  A practicing engineer would be schooled in other design methodologies.

I remember a typical example in the lab working on a DC servo loop in the 80's. (yes, I'm that old).  I had a couple of decade boxes on the bench tuning the servo loop on an aircraft panel instrument prototype.  The chief engineer came in and admonished me for trying to solve the problem empirically with the statement "you have to design this mathematically".  I stepped away and he started computing.  The next day when I came back in, he has more decade boxes than I was using, he was trying to tweak the servo loop!  The devil made me say it, but I had to.  "I don't think you can design this empirically".   He stomped out of the lab in frustration.

FWIW, I kept the job and proceeded to finish the design with the decade boxes.  What he was failing to take into account was all the friction and inertia characteristics of the gear train, that was the real problem.  Trying to quantify them mathematically was a hell of a lot harder than just tuning the loop empirically.  The instrument worked fine and was certified and shipped.

Some 30 years ago I was having trouble employing a logic gate as an oscillator. It didn't always start. My manager sent me to visit a professor that ran through all the math and ended by saying something about black magic. I finally got it to work, through trial and error, by using an inverter gate with hysteresis. That product went to market and I never heard of a single failure.

I confess, typically when I wanted an oscillator, I use a real packaged unit.  I couldn't afford to risk not starting over the -40C to +105C temperature range that most of my designs got used at.

gunrunnerjohn posted:

I confess, typically when I wanted an oscillator, I use a real packaged unit.  I couldn't afford to risk not starting over the -40C to +105C temperature range that most of my designs got used at.

It was a cost and space constraint.

Keith:

Yes, I do too.   But I also use V = IR = R dQ/dt,  and V = Q/C for the ODE of interest here...Viz, the '1st-order lag' ODE,

RCdV/dt  + V = Vo*sin[Omega*t];  V(0) = 0.....For example.

P.S. Welcome back!

Stan:  B 4 I do much more 'damage',  I think we need a telecon. I see you are also an Oregon resident.  Please either send me your fone # or call me [but not today] @ 503 728 2349.

Phil, yes I am just south of Portland and am willing to chat but frankly don't know what more there is to say. 

It seems to me your problem has been solved.  By lowering the input voltage going into the TIU variable channel, the spurious whistles go away.  And, in the specific case of using a Z-4000, a 25 cent capacitor placed across the Z-4000 output can help.

Stan:  I live an hour north of Portland;  between Rainier & Clatskanie

I agree that the problem of my club's layout has been solved.  But I still need to finish up loose ends related to OFFSET and RC-ckt integration.  Will you call me tomorrow, or give me your phone # and best times to call you?

BTW, you live close-enuff to me to come visit my 'pretty-nice', multi-level layout in a 25x40 building,  should you so desire.  I control it with EVERYTHING:  DCS,  TMCC, LEGACY, and CONVENTIONAL via 2 PW ZW's....