Safeguarding UTP Cables in Elevator Shafts: Filtering Options

[mishobg12]

Hi,
I am making a device with 8 INPUTs that will be connected to a long (5-25m) UTP cable running trough an elevator shaft (some interference, I imagine). Could you propose what kind of filters should I use on my inputs. The thing is, the HIGH level on these cables will be 12 to 19V but I need to convert them to TTL. I imagine some sort of voltage divider and low-pass filter. Or probably something with a transistor. I don't want to take any chances, though.
Thanks,
Misho

 

[berkeman]

Hi,
I am making a device with 8 INPUTs that will be connected to a long (5-25m) UTP cable running trough an elevator shaft (some interference, I imagine). Could you propose what kind of filters should I use on my inputs. The thing is, the HIGH level on these cables will be 12 to 19V but I need to convert them to TTL. I imagine some sort of voltage divider and low-pass filter. Or probably something with a transistor. I don't want to take any chances, though.
Thanks,
Misho

You should use differential signalling on the unshielded twisted pair (UTP). The easiest conversion would be to use RS-485 transceivers at the ends, and be sure to terminate each end in the Zo of the cable (probably around 100 Ohms).

 

[mishobg12]

Thanks for your answer, berkeman.
There is a problem, though. I guess I should have said that the communication is parallel. This does NOT depend on my device.

 

[berkeman]

Thanks for your answer, berkeman.
There is a problem, though. I guess I should have said that the communication is parallel. This does NOT depend on my device.

What do you mean by parallel? Are there 8 UTP cables for the 8 signals? Are you saying that you don't control the driving of the signals onto the UTP(s), but need to deal with receiving them and converting them to TTL? What is the driving source impedance and current capability? Will you be able to forward terminate the UTP(s)?

 

[mishobg12]

My device is an elevator display. One UTP cable has 4 twisted pairs = eight cables. every cable is connected to one of my inputs. They have a 5-bit binary number - a code for the respective symbol that should be displayed. The other tree are for the direction. There might be only one display, or there might be one at each stop. So I can't terminate the signal. I also must support different type of control boards (which output the signal) and I can't possibly know the source impedance. I must support a large diapason of characteristics.

 

[berkeman]

My device is an elevator display. One UTP cable has 4 twisted pairs = eight cables. every cable is connected to one of my inputs. They have a 5-bit binary number - a code for the respective symbol that should be displayed. The other tree are for the direction. There might be only one display, or there might be one at each stop. So I can't terminate the signal. I also must support different type of control boards (which output the signal) and I can't possibly know the source impedance. I must support a large diapason of characteristics.

What about power and ground? So you aren't using the UTP as transmission lines, just as individual conductors... Have you built a 25m setup in your lab yet to look at what ringing does at the display?

 

[mishobg12]

power is 19V stabilized to 5 with 7805 circuit.
I was planning just to use my best bet on the prototype and install it on a real elevator. Since it's a digital circuit it can't be that sensitive, right?

 

[berkeman]

power is 19V stabilized to 5 with 7805 circuit.
I was planning just to use my best bet on the prototype and install it on a real elevator. Since it's a digital circuit it can't be that sensitive, right?

The issue is ringing on the signals, caused by the uncontrolled tranmission line and inductance that the signals will see. If the rising and falling edges are slow enough, then ringing may not be a problem. If they are digital logic speed, then you will likely get enough ringing to cause data integrity issues. That's why transmission lines for sending data some distance are usually controlled impedance transmission lines, with appropriate terminations (forward-terminated, back-terminated, or doubly-terminated).

 

[mishobg12]

If the rising and falling edges are slow enough, then ringing may not be a problem.

They are slow.. The signal only changes when the location (or direction) of the lift changes.
De facto, the minimum time in which the signal will not change is equal to the time in which the elevator travels trough one stop. This usually is as much as 4 seconds.

 

[berkeman]

They are slow.. The signal only changes when the location (or direction) of the lift changes.
De facto, the minimum time in which the signal will not change is equal to the time in which the elevator travels trough one stop. This usually is as much as 4 seconds.

It's more the edge slew rates that I was referring to. Regular digital logic can have typical edges that transition in just a couple nanoseconds, and that will cause ringing on uncontrolled long lines. If the signals are latched by a clock signal, then the data that is latched can be wrong, depending on where the ringing lines up.

If the signals are not latched, then there should not be an issue.

 

[mishobg12]

They are not latched. It really is very simple communication, because all of the work is done by the display itself.
So do I think I should put at least a capacitor in parallel to the input. And if yes, what value would you suggest?

I found what you said interesting, though.
http://prikachi.com/images/367/2373367y.png
Is something like this what you mean?

 

[berkeman]

I'm not understanding the graphs. Basically if you think of the Fourier transform of a pulse train with finite slew rate edges (not an ideal square wave), you will get harmonic content that depends on the sharpness of the edges (how short the rise and fall times are). That harmonic content can be a problem when the propagation time in the transmission line approaches the wavelength of the harmonics. So faster slew rates will give problems on unterminated/uncontrolled lines of shorter lengths.

As to a smoothing capacitor at each input, it wouldn't hurt. Use the output impedance of the drivers, and calculate your C value by setting the RC time constant around a millisecond or so for this application.

 

[Carl Pugh]

Are there special standards for elevators?

An electrical fire in an elevator would be a disaster.

I would hate to ride in an elevator where the designer of some of the electrical components had to go to a website for electrical advice.

 

[mishobg12]

hey, I am a student.
berkeman thank you for answering. I think I got it.

 

[berkeman]

hey, I am a student.
berkeman thank you for answering. I think I got it.

So students are allowed to start fires? Ah, the good old days. I remember this one time...

Anyway, Carl's concerns are valid. You're at least using plenum-ratec cables, right?

 

[mishobg12]

what I meant was I am NOT a professional :) high frequencies and active components have always been hard for me. I am not good with advanced math, maybe it's because of that. I'm focusing on digital electronics, but you always need to have some active components...

Since electronics books in my native language derive even the obvious things from a page-long equations, I started reading the Art of Electronics , which is as non-mathematical as it gets, but still I am not confident. Hope I'll just get it somehow