Ceiling lights PWM control via a long wire

[MarekB]

Hey guys, I am trying to design a circuit for PWM control of ceiling downlights with a microconroller. Here's a schema I came up with so far:

TLC5940 is controlled by a microcontroller (not shown). It's output is connected to a logic optocoupler (6N135 or similar) whose output is connected to a line driver (74244 or similar). The output of the line driver goes through a ferrite bead and connected to a capacitor to reduce EMI. The 5V PWM signal from the line driver then travels through one pair of cat5 cable (estimating 10m at max) which is terminated at the end with a 100 ohm (+-15ohm) resistor to reduce reflections. It then continues through a 4R7 resistor to the PWM input of a constant current driver (like for example LDD-350H). The PWM input has a 5V1 zener diode on its input to clamp any excess voltage. The other pairs of the cat5 cable are used to carry anything between 12V and 57V depending on the downlight (and appropriate current driver) requirements. The line driver is powered by a DC to 5V module.
The PWM signal on the remote side does not have to be perfect, I don't think it will cause too much variance in brightness (as oppose to a perfect PWM signal). LED downlights won't need more than 700mA. Most of those I looked at are 350mA or 500mA. But the voltage can be quite different depending on the chosen downlight (number of LEDs, technology,...).

So my question is whether this design is any good and what would you improve? Especially in terms of reliability, safety and coexistence with other electronic devices in the house. Here is a few notes and questions:

1. I'd like to keep the remote/light side as simple as possible but I can add as many passive components as needed if it improves any of the three things I mentioned earlier. Maybe add another 5V TVS diodes on both sides to protect against surges? And another TVS diodes on the power lines? I can also add any active and passive components on the side of the line driver.

2. Those cat5 cables are going to create a star like topology on the ceiling (imagine 20 lights) and with those long wires I am bit worried about EMI. Is the cat5 twisted cable going to mitigate EMI to acceptable levels? I can use a shielded cat5 if necessary.

3. If there will be any power lines (230V) crossing those wires e.g. for controlling electrical jalousies, will switching 230V cause the lights to flicker (even for a brief moment)? If so, how to prevent it? Better shielding or some filters? Maybe another 100nF capacitor on the PWM input?

4. Is the line driver necessary? Would the optocoupler be enough to drive the PWM line? I will rather keep the line driver if it's better to use it, I do not want to be changing that optocoupler every month.

5. Anything else I should be thinking of?

 

[Diver300]

It is common to use a series resistor, whose value is equal to the characteristic line impedance, when driving a transmission line. That has the downside of forming a voltage divider with the terminating resistor, so halving the output voltage, but it means that both ends of the transmission line are terminated at the characteristic impedance, so neither end will reflect signals.

The 74LS244 is being overloaded driving 100 Ω to ground. It is only rated down to 133 Ω load. You could probably put a 2 outputs in parallel to get over that problem.

I think that you would do better with a device that is designed to drive a low impedance load. There are are lots of RS485 drivers that would work very well.

I don't know what frequency PWM you are intending to use, but I would suggest at least 1 kHz. 100 Hz PWM lighting can be seen by many people, and has odd effects on others who can't directly detect the flicker. This was noticed with the modulation of fluorescent lighting at twice the mains frequency, but PWM flicker can have a larger effect because the lighting levels turn on and off faster on each PWM cycle. Personally, I won't have 100 Hz lighting in the house, and will go to some effort to modify PWM circuits to remove flicker.

**broken link removed** says quite a lot about it.

 

[MarekB]

Thanks Diver300 for your reply. I intent to run TLC5940 PWM at around 1kHz. I may try the series resistor, will see if the constant current driver will be able to pick up a half voltage PWM signal. Thanks for pointing out overloading of the line driver.
I was thinking about RS485 but it would mean that I need RS485 receiver on the remote side and another 5V voltage source to power it. I plan to use RS485 but only for serial communication between master-slave microcontrollers.
Would TVS diodes on signal and power lines be of any usefulness? Also anybody got any thoughts on points 2 and 3 (or anything else)?

 

[ChrisP58]

EMI

Ethernet signals work through cat5 UTP with low EMI, not only because of the twisted pair, but also because the drive is differential. Meaning that each wire in the pair is driven with a voltage exactly opposite that of it's mate. That. combined with the twisted pair reduces the EMI radiation to near zero. A single ended signal, such as you are showing, will not benefit nearly as much from being twisted.

Ethernet achieves differential to single ended coupling at each end through use of transformers. You could go that route and not need +5v at the receive end, but ethernet transformers will saturate at the low data rate of 1k.

RS422 and 485 transmission systems are based on the same differential to differential concept. And yes, need 5V logic at each end to make them work

But, since your data rate is so slow, I would suggest another approach, that of limiting the slew rate of the PWM edges. Stretching the rise and fall time out to a few hundred nS should reduce the EMI generated significantly. And that can be done with a simple RC filter at the drive end. What are the input characteristics of the LDD-H chip at the far right of your drawing? Is it picky about rise and fall times, or just duty ratio?

 

[MarekB]

This is the **broken link removed**. Not much information there in regards to PWM_DIM input. Anything below 0.8V is OFF, above 2.5V is ON, PWM frequency 100Hz-1kHz. Is it reasonable to assume that it has a Schmitt trigger on its input and thus won't mind slow rise and fall times? Any idea what cut-off frequency for the low-pass filter should I be shooting for?