Question about frequency response of DCS locomotives...?

Actually, I thought about just measuring it with a keysight PNA-X vector network analyzer but I don't trust that the measurement is reflective of the truth becasue the train isn't powered in this condition. Like... you can't power the train in parallel because the 20V would toast RF input of the VNA.

Figuring it out from inside the locomotive or TIU would mean cutting board traces with a knife to cut the part where the power and digital signal branch apart so that I could get separate current readings.

Anyways, hope someone doing this longer than me can offer insight...

The impedance and transmission line aspect of this has always been a bit of a curiosity to me. 

Obviously the 3 rail track we are using doesn't look anything like a 50 ohm piece of coax.  Maybe it's closer to some sort of twin lead or ladder line?

Since the track isn't 50 ohms, did they necessarily design the engines and TIU as we would normally for a 50 ohm system?  Unknown to me.

There is a member here (Susan Deats) who has a son who is an RF engineer - he worked up some RC filters to try to enhance the DCS signal for people that were having problems a while back.  Here's a link to the page regarding the filters he came up with.  Prior to that people were using a "lighbulb trick" to help with the signals making it to the trains.  Newer versions of the TIU seem to be much better than the original from almost 15 years ago that caused these measures to be more consistently needed.

Agree on not wanting to fry the PNA!  That would be an expensive experiment!  One problem with the trains running on AC, I guess.  Can't easily design a safe bias tee to protect the analyzer from the power but simultaneously let the RF signal through.

Thanks for all your all of your very well explained posts (welcome to the Forum!)!  I've enjoyed reading them.

-Dave

Hey there,

This is actually super helpful. Thank you greatly!

I simulated the filter from that post (221ohm and 0.1uF) and it's like Z=221 - J(nothing) up at a few MHz. I could calculate T=1/RC by hand and draw the bode plot....  but ADS is faster and I'm lazy

It's a low pass filter that converges to 200 ohms at AC with a 3dB corner around 8 KHz so we can infer stuff. If the train's own impedance was 9 ohms beyond 7 KHz, then this filter wouldn't actually do anything since 220 ohms in parallel with 9 ohms is 8.6 ohms (the situation has not changed to a coupling spur or noise since the impedance is basically the same).

I guess the idea was to eliminate high frequency noises from coupling into the track. I guess that kind of parasitic signal coupling would be weak so their Thevnin impedance would be high (like >Kohm) so that's why a 200 ohm impedance in parallel would make them go away.

Wavelengths at these frequencies are so long, you would have to assume it's capacitive coupling, not electromagnetic wave coupling... so you can find the effect with a voltage divider

Vtrack = Vnoise  X [200 ohms / [ (a few K of thevnin) + 200 ohms]]

Anyways if that filter really does improve things then it's a safe bet the train impedance is on the order of 200 ohms at the DCS frequencies.

Thanks!

~Adrian

Good info! 

But what about us that use DC power only??

Casey Jones2 posted:

Good info! 

But what about us that use DC power only??

If everything is at DC then I guess it's just track voltage divided by locomotive current, (back to simple V =I/R) The frequency response is still the same  but you're only exercising the DC part of it in this case....

Here's an actual measurement of the track resistance of my locomotive at each frequency from 0.5 MHz to 10 MHz. I came up with a way using a voltage divider to get the measurement without blowing up the network analyzer.

-Adrian