I have to ask this question since it has been bugging me for some time. Why is it that decoder manufacturers provide straight-line acceleration functions rather than providing proportional acceleration? By proportional acceleration I am referring to the kind of acceleration we experience in real life. If we put the pedal to the metal, we accelerate rapidly at first and the rate of acceleration drops as we reach top speed. If we don’t apply as much throttle, we don’t accelerate as rapidly. In the straight line type that we find in all decorders that I am familiar with, the rate is constant no matter how much throttle is applied, or how close to max speed we are. This is unrealistic.
In the early 90’s I designed and built a controller and decoder from scratch because I hadn’t heard of DCC yet. It incorporated proportional acceleration and that part worked well. The train accelerated and decelerated realistically.
My scheme failed because I was attempting to use pure DC track voltage and to send commands via a digital signal applied on top of the DC. It worked but I lost control as the train began moving, probably due to electrical noise. I didn’t think anyone would accept an AC signal on the track. Lenz got it right.
So the question remains. Why use the straightline accel curve? And BTW, don’t tell me about the speed step table. That does not solve this problem in any way. It is not proportional to throttle and it’s backwards for deceleration.
You can program momentum into decoders. I was at an op session where I used a loco with really high momentum. You turned the throttle up to get going and it started really slowly then as it picked up momentun, you had to scale back the throttle. Stopping was similar. You cut the throttle way back and had to allow the engine distance to stop.
If I understand the dynamics of electric motors like the types used in locomotive traction service, and steam locomotives tractive forces, your analogy to an automobiles acceleration (acceleration in real life?) is probably not correct. Electric motor generate 100% if their torque outputs from very slow speeds, as do steam locomotives. In fact, if you read the specs on a diesel locomotive, you will find that it’s starting tractive effort is much higher on startup than what can be produced over long periods of time. Given all the different variables, I’m not sure what the acceleration rate of a “real train” is like, but I’m fairly certain that traction limitations (wheels slipping on the rails) affect this number far more than than we notice in an automobile.
So factoring all these things together, I could assume the starting speed from zero mph, and the acceleration speed from say 20 mph are actually pretty close, but I don’t know that for a fact. Now if real locomotives had rubber traction tires, my whole concept would be off… [;)]
You may very well be correct on the dynamics of the low end. My point is just that the velocity curve is not a ramp for any accelerating object except for a rocket in space. It is an exponential curve. The corresponding graph of acceleration is not a flat line that suddenly drops to zero when top speed is reached. It is a curve of some kind that dwindles to zero as top speed is reached.
Decoders, if configured to simulate acceleration, ramp the velocity up at a constant rate from zero to top speed. And that ain’t realistic. They have those speed tables in there that you can program with an ‘acceleration curve’ but that is a boneheaded approach to the problem. It sorta works when increasing speed, but is completely wrong when decreasing speed. And it is not dependent on throttle position at all.
You may very well be correct on the dynamics of the low end. My point is just that the velocity curve is not a ramp for any accelerating object except for a rocket in space. It is an exponential curve. The corresponding graph of acceleration is not a flat line that suddenly drops to zero when top speed is reached. It is a curve of some kind that dwindles to zero as top speed is reached.
Decoders, if configured to simulate acceleration, ramp the velocity up at a constant rate from zero to top speed. And that ain’t realistic. They have those speed tables in there that you can program with an ‘acceleration curve’ but that is a boneheaded approach to the problem. It sorta works when increasing speed, but is completely wrong when decreasing speed. And it is not dependent on throttle position
Nope. Changing the momentum CV just changes the slope of the ramp. My point is that it shouldn’t be a ramp, it should be an asymtotic curve. I wish I could draw it here to demonstrate.
Are you sure about that? I doesn’t react like a “ramp.” It acts like a geometric curve. Very slow acceleration in the beginning and increasing acceration as you near speed. With sound, it is a high rev that in the beginning that flattens out as you reach “speed.” With the throttle you have to back off considerably when get near the speed you want to travel.
I know what you mean, ghofmann. Initially, there is an acceleration of the acceleration itself, and then the acceleration becomes linear. As the high speed intended is reached, the rate of acceleration declines until there is zero acceleration at the intended speed. However, engineers drive their trains like we drive cars; we ma***he gas to get the rate of acceleration (and adhesion on the rails vs. slippage) and back off on the throttle when we reach our intended speed. That is how I drive my DCC trains. I hit the throttle hard, with CVs 2 and 3 programmed for about 3/4 slow, and I get the laboured chuff and rumble of both types of engines. When I reach speed, I often back off by 5 or six clicks on my DT400.
But wouldn’t this only account if you pegged to full throttle. Knowing your engine’s throttle response is all you would need for control at any speed or condition.
Bob K.
Let me preface this by saying I’m not an expert on decoder design - I haven’t studied current decoders at all. I’m just applying logical design engineering to the problem. We live in an analog world; the output of the decoder appears to be using pulse width and/or frequency modulation of “12 volt” pulses to the analog motor, almost directly controlling its speed.
Changing the speed curve, although it theorectically drives the acceleration curve (derivative of the speed curve), depends upon the preprogrammed relationship in the decoder between the 2 methods of calculating speed - using the speed curve, and using the constant acceleration parameter (I am assuming the momentum CV simply modifys the acceleration constant). If the speed curve dominates, and a derivative is computed for the acceleration curve, then realistic acceleration is possible (deceleration remains a problem). If the constant acceleration parameter is weighted more, than you have the unrealistic output cited in the problem.
A possible solution is a programmable acceleration/deceleration curve. The speed curve becomes the integral of the input acceleration curve. The problem with this scenario is that the acceleration curve is going to vary tremendously with train load - the speed curve isn’t. Which means you would have to program in a new CV set for the acceleration curve as your load changed.
The only way out I see would be to have a “fixed” acceleration/deceleration curve moved up and down the vertical axis by a “train load” constant, but I don’t know if this would be realistic enough. This solution would still require reprogramming a train load CV on the fly as the engine uncoupled/coupled from/to the train or otherwise changed loads significantly. I think this would be far too complex for operators to use on most layouts with the relatively short runs. Also, how much extra processing power is required? Would it drive up heat, and possibly even physical size of decoder?
I don’t think real locomotives accelerate anything like a car. If the engineer just yanks the throttle wide open (assuming it doesn’t pull the drawbars out of the cars), you just get a whole lot of wheel slip, or, in the case of more modern controls, the computer will regulate the power to the limit of adhesion. If the loco can generate maximum tractive effort for the speed at notch 4, it’s NOT going to accelerate any faster at notch 8.
The adhesion limits are much lower for steel wheels and steel rails then they are for rubber tires on dry pavement - so there is a much more noticeable band with a typical car where more throttle = faster acceleration, but there is a limit there too. A high powered car can easily overcome the limit of adhesion - just ask any powerful V8 driver who’s been embarrassed by a small 4-cylinder because he got nervous and dumped the clutch and made a lot of tire smoke but not a lot of forward progress while the little car just pulled away. Traction control in cars works the same way - if the traction control has to kick in a 1/4 throttle to keep from spinning the tires, the car is not going to accelerate any faster if you floor it.
As for proportional braking, well, since our model trains do not coast all that well, even if they had freewheeling drives (large die-cast O scale and larger might, but still FAR less than the prototype), that’s goign to be hard to accomplish. Slot cars have had brakes for years - dymanic brakes, as a matter of fact - when in the braking zone of the controller, the resistance is connected across the motor with the power disconnected. They STILL stop on a dime, WAY too fast to be prototypical. The closest we have so far is the newer QSi diecoders that have a dynamic brake - Atlas Trainmaster and newer locos have this. It doesn’t just turn on a noise, it actually brakes the train, release the brake button and the train accelerates back to the throttle setting.
I understand what you are saying, Randy, and agree. I was referring to the toy itself when I mentioned opening the throttle. The instructions say that the "sound of power’ comes on as you open the throttle at a given speed. So, I make sure it DOES come on by over-doing the throttle advance. I agree that no steam engineer would pull back the throttle more than 10-15% on initial start-up because all that pressure on the pistons would just spin the drivers.
Back to the toy: I don’t really think the exaggerated throttle advance that I do makes the loco accelerate any faster. The settings for CV2 should control its acceleration. But, I sure get better sound. Now, if I could only get some wheel slip and an associated chuff-rate rise…THAT would be cool. [8D]
Can’t speak about DCC but I can tell you operating conventional DC with the momentum turned on makes for a real interesting operation. On long trains you can even watch the slack run out as you start and as has been noted, you better plan ahead on your stops.