Rich Weyand's CoolerCrawler analog throttle 2025 update

[diesel]

Interesting. It would be worthwhile to consider all the options too, not just the CC. I'm puzzled though, why the ITTC would be quieter than the CC, given that the pulse width on the ITTC is shorter. Because the CC uses both halves of the 60Hz cycle and the ITTC only seems to use one.

One annoyance I'm looking at trying to overcome - the CC relay "sings" at a high pitch, not loud but it rankles me. It's because it's running on rectified but unfiltered AC.

 

[gasnbrass]

Here are two waveforms with the same volt range and time scale. Both are running an Atlas/Kato loco at a scale 3mph. The pulse width of the ITTC is longer and flatter than the CC. The CC will make noise at 120hz, while the ITTC will be a lower pitch of 60hz, so the lower pitch is part of why it sounds quieter. Also the peaks of the CC are higher amplitude near 4v, which will cause a stronger noise pulse while the ITTC is around 2.7v. The ITTC fixed the PWM noise issues of abrupt voltage changes by slewing the voltage up and down and rounding the leading edge of the peak to reduce resonance. The tradeoff is that the "peakiness" of the CC wave along with the adaptive cap makes the CC slightly better at very slow speed starts and stops, but the ITTC is slightly quieter. It's splitting hairs at this point casue both work very well, probably the best two throttles I have.

As for the relay, how about a magnetic latching relay. Only needs a power pulse to flip the circuit.

CC

ITTC

 

[diesel]

Thank you, that is informative. I tried to find info on the ITTC, it looks like it's not made any more and not much out there. But it is worth forking to a separate attempt. Meanwhile, I thought it would be useful to update the circuit to all MOSFETs:
(found a serious mistake so I'll repost later...)

 

[ctclibby]

Thank you, that is informative. I tried to find info on the ITTC, it looks like it's not made any more and not much out there. But it is worth forking to a separate attempt. Meanwhile, I thought it would be useful to update the circuit to all MOSFETs:
View attachment 247997

Nice clean schematic. KiCad I take it? Maybe think about labeling those in/outputs to what they are and not the part#. You might also move J2 to the left edge of the schematic? Also, a damping diode across the relay's A1/2 connection. You might start thinking TvS's for the inputs and outputs, pretty cheap protection against the HBM.

L8r

 

[diesel]

Had some trouble with the current limiting circuit, now it looks like this. It's not all FET, it has one PNP transistor for current limit detection and the driver and final stage are MOSFET. Also I decided to try rectifying the DC supply that runs the relay and biases the control circuit.

 

[ctclibby]

Better.
I am a little confused, maybe you can show me the way. J1 I am assuming it is power in ( AC,DC+-,DC-+) connects to V_DC ( half wave rectified? ) and continues to V_DC on J2. The Bridge connects to V_Pulse for your circuit power. J2 connection a simple POT and DIR Switch? You have J2's DIR coming back in for relay control. Why do you need R7 and that connection at all? I mean if DIR is controlled by a simple toggle switch, it is DC+ or GND and the DC+ should be enough to energize the relay even considering some wire length loss.

Did that make sense?

To me, it looks like you only need the Bridge + Smoothing Cap to power everything. Also as it now is, shouldn't C3 be connected directly to GND and not through the track circuit/motor back to GND?

Why so many poles for the relay? Is there more stuff to be connected? Might be just me - ima cheap!

FYI, GBU4M's come in 2 flavors according to Digikey; 2.8A and 4A.

There is not a part number on the relay so I couldn't look up the current rating on the points. When relay's Make or Break there are arc's generated and you are connecting/disconnecting an inductive load. At even 1.5A you could have large transients in the hundreds of volts for a very quick time. QuenchArc devices might be something to look at. Scope at the points with motor running, then switch DIR. Ya, motor ( train ) is supposed to come to a stop, then switch DIR. Sooner or later, won't happen and DIR is switched at some VDC to the motor. Hmmm, maybe some circuit glue to prevent the DIR switch if above a certain VDC?

L8r

 

[diesel]

J1 is low voltage AC input. With the MOSFET circuit the voltage loss is not so great, so it may work between 8-16VAC subject to experimentation. J2 (and quite a few other things) are cloned directly from the original CoolerCrawler. J2 is a walkaround throttle plug, having the connections for speed and direction. The reason for the R7 circuit, is so the throttle maintains the same relay state when the walkaround controller is disconnected (e.g. to walk around LOL). If you just had a switch on the controller direct to the relay, the train could instantly change direction when you unplug. R7 might also serve to reduce the current appropriately, as the relay is 12VDC and the supply might exceed that.

The relay socket is an Omron_PY14-02. I used a HH54P MY4N-J relay from AliExpress, contacts are rated 10A. When the project is further along, it will have a proper BOM and things like that would be listed. The reason for the 4 poles has to do with current capacity and sourcing semi-obsolescent parts in 2025. I looked at lots of 2 pole relays and none of them had a high enough current rating. It's a prototype now, so if a more suitable relay appears, I would consider it. I have the impression, this is a very common industrial relay so very easy to source.

This throttle design uses the halfwave rectified AC as a source of pulsed power, so rectifying all the power would mostly remove the pulses and defeat the low speed motor control in general, but also render the "magic capacitor" circuit completely non-functional. So I rectify only the first stage control circuit power.

There is no problem with inductive kick back, as there is a proper relief diode in place, D7.

The problem of reversing under full power, exists in the CoolerCrawler and in many other throttles, including all the older MRC at least. It may be possible to build in protection for that, but it's not consistent with the simplistic approach this throttle embraces. However, I am also seriously considering cloning (or imitating) the ITTC throttle which would be based on some digital control circuitry and so offers a chance to incorporate features like that.

I think, the difference in GBU4M's derives from whether they are equipped with a heatsink mounting hole, but I could be wrong. At this stage, 2.8A is more than sufficient, and real life testing will reveal any limitations of it, as well as a few other devices. Really, my choice of heatsink for the main power transistor (Digikey 294-1080-ND) was a little by guess and by golly. But it should dump about 8-10W of heat which I think should do the trick. My recent board layout places the diode bridge and power transistor side by side, so they could use the same heatsink if it's needed.

The current implementation leaves some things open, like the type of throttle plug (I would probably try to place it on board later)... that is because this version is mainly a prototype to iron out any bugs or make any changes that are inspired by its use. So yeah, just four bare connections for external pot/switches.

It now looks like this:

 

[gasnbrass]

So I have found a simple way to make the acceleration and deceleration a linear slope instead of an exponential curve caused by the momentum resistor/capacitor. During accel the voltage is climbing quickly at lower voltages and it's difficult to adjust for minor low speed changes because the voltage wants to ramp up and down quickly with very little pot input. A linear slope makes accel smoother and more repeatable and predictable. On decel the speed drops quickly at high voltages but as it approaches zero the rate of voltage drop slows way down and drags out the decel for long distances. This makes predicting the final stopping point harder, and can cause the loco to repeatedly stutter to a stop at the final moment. A linear braking rate makes it easier to stop where you want it, and drops the voltage fast enough to minimize or avoid the loco stuttering all together.

The fix is two LM334 current regulators on each side of the diode and capacitor to eliminate the exponential input/output of the cap. Here is a basic schematic of what I've tested (I used npn transistors to represent the IC's). This IC puts out a constant current up to 10ma based on the rating of the resistor bridging the adjust and V- pins of the LM334, and can operate between 1v to 40v. Of course one could experiment with different capacitor and resistor rates to get their desired accel/decel/push button braking rates. The LM334 adjust pins could be wired to selector switch with various resistances to change the up/down momentum rates to simulate train weight, and the downstream LM334 adjust pin wired to an additional pot for truly variable braking, just depends on how complex or customizable one wants to make it. Adding an inline resistor or pot between the cap and either LM334 will start to bring back some of that exponential curve.

 

[Canalligator]

I have a limited knowledge of electronics; I can follow along with what you folks are doing, but I couldn’t design it myself. It’s good to see knowledgeable people doing some interesting work.

 

[gasnbrass]

I have a limited knowledge of electronics; I can follow along with what you folks are doing, but I couldn’t design it myself. It’s good to see knowledgeable people doing some interesting work.

I'm just a self taught amateur so only sorta know what I'm doing. Still it's fun to learn how this stuff works, even if it's mostly obsolete in today's modeling world. The website that really helped me to understand how it works and what's actually happening in the circuit is the Falstad circuit simulator.

 

[diesel]

I'm not sure it's right to call analog obsolete, it's still around a lot in N scale and it's the norm in Z scale.

I now have the almost clone CC and also the MOSFET version constructed, but I was robbed for time to scope them today. Mainly, I set up the N scale traverser track for testing, and found some power supplies. I scooped an MRC Tech II Railpower 1400 at the flea market, not only for comparison but for the AC available at the accessory terminals. That can power the throttles. Meanwhile, I scoped the Railpower 1400, and it is definitely the 2-level pulse type, every alternate half cycle is at a lower voltage:
On slow setting:

and on medium speed setting:

 

[gasnbrass]

That was in the context of HO scale, where it appears a majority of HO modelers use DCC these days. I dabble in both DC/DCC.

That 1400 appears to have about the same high/low amplitude ratio as other basic MRC power packs I've tested such as the tech4 200. They do it with a simple resistor tapped from the transformer to the wiper pin of the pot. The MRC tech2 and tech4 momentum throttles use pulses (of various amplitude, some quite high) on top of a base voltage. I did test a Troller power pack that had nothing but half wave pulses til it got past half throttle, then finally the full wave pulses started to show. So trains at normal running speeds were running on half wave almost all the time.

 

[diesel]

Usually, DC locos run quite well on pure DC at medium and high speeds. So it's likely that any pulses are wasted in that condition. Or at least, the pulsing that pulse throttles are producing at other than low speed, is mainly a byproduct of the design. I ran an older Kato N engine, the buzzing was very noticeable with the 1400, and I think that's what you can expect from a roughly 25% duty cycle. I vaguely recall using one throttle that used straightforward PWM, and it made the motor extremely noisy. I don't think it had smoothed waveform edges like the one you indicated earlier on. Also I agree that if both phases are used (120 Hz), it should be quieter. If the suppressed alternate peaks are used, I think you could phase them in sooner. The difference in peaks on the 1400 seems a little more than would be needed to effectively use the 60 to 120 Hz transition. You seem to have to be way up the speed scale before you get much support from the alternate phase.

As an aside, I'm wondering how I can work in a constant AC bias to energize occupancy detection circuits when the train isn't moving. Also, if that function can be combined with the pulse circuit somehow.