Do you have space to include a high-uF electrolytic capacitor across the LEDs to reduce flicker when pulsed at 20Hz?

Hi, this is a good idea. This solution is used specially with internal passenger car light kits because, although the car's wheels pick up the energy, it happens small blackouts across the tracks due bumps or even bad contacts. This eliminates those flickers, and, I guess, should also eliminates flickering caused by the PWM frequency. N scale locomotives are so small so I would prefer a solution at the controller level. In fact, soon I'll migrate to DCC system where the decoders are mounted inside the locos and this kind of solution, I guess, it is in use: an approach to motor and another one for lights, both at the decoder level. For now, I'm using 40Hz which is led flicker free and the buzz sound is on bass band and low, so it's much more bearable and with a little imagination, it seems the beating sound of a diesel motor, unlike at 500Hz which sounds like a annoying mosquito.

 

No, that resistor controls the current that runs from your Arduino out through the base and then the emitter of your transistor when the Arduino pin is high. There are limits to how much current each of these can handle, although in this case the Arduino would be the limiting factor (IIRC it's good for max of 30mA per pin.) You're limiting it at around 5mA right now, and could easily raise that to 10 safely (it might or might not help your high frequency performance - that's a different question entirely, but it wouldn't be a risk to any of your components!)

As you can see, my electronic skills are at the very beginning. Arduino pins output 40mA and, as you said, with resistor the output is limited to around 5mA. I'll make a experience with a lower value resistor to approximate the output to near 30mA. Maybe this can improve the high frequency response of the transistor. That what I said about protection, I think I messed with the information I was reading on the sites. Thanks for the correction.

 

I'd be curious to see if cutting that in half to increase the base current would make any difference, but it's mostly speculation and wishful thinking on my part. Can't find a whole lot of info on TIP120 switching speeds, but what little I found looked reasonable, and was based on Ic/Ib of 250. Hard to know without better motor specs, but I think you might be running higher than that now, which is what made me wonder.

Looking forward to hearing how MOSFETs work out. Good luck!

I changed the resistor value to 220 ohms and the performance was the same when using a 1k resistor.
I also tested with a mosfet instead Tip120 and the linearity response was the same.
The answer is in the article on post#10 shared by blocco a spirale. The time-constant described there must be smaller than the PWM pulse interval. If not the loads doesn't reach its full value and feeding, consequently torque, loses efficiency, specially on low duty cycles values. The loss of linearity on PWM high frequencies is due motor's circuit features and not switching capacity of the transistor.

 

I changed the resistor value to 220 ohms and the performance was the same when using a 1k resistor.
I also tested with a mosfet instead Tip120 and the linearity response was the same.
The answer is in the article on post#10 shared by blocco a spirale. The time-constant described there must be smaller than the PWM pulse interval. If not the loads doesn't reach its full value and feeding, consequently torque, loses efficiency, specially on low duty cycles values. The loss of linearity on PWM high frequencies is due motor's circuit features and not switching capacity of the transistor.

Sounds right. That was a great article. Thanks for experimenting and sharing your results.