level-shifting daisy-chained data line from 5V to 24V for long distance communication stability

Question

I need to daisy chain several WS2811 chips with distances up to 20-30m in between.

Every unit has a WS2811 chip that drives 3 DIO5151BCD10 LED drivers to control a high power RGB led. To account for voltage drop, a 24V line runs between the units, and every unit uses a step down converter to get 5V. The other two lines that run between the units are the WS2811 communication line and ground.

I want to make sure that the stability of the communication line for the WS2811 chips is as good as possible and so I thought that level-shifting the signal to 24V for the distance between the units could help me achieve that because a higher voltage is less prone to outside interference from what I understand.

This is how I imagine it:

WS2811(5V->24V)-----(24V->5V)WS2811(5V->24V)-----(24V->5V)WS2811(5V->24)...

I have been trying to make a circuit that helps me achieve this and have tried adding a classic 1 n-channel MOSFET bi-directional level shifter on both ends but after looking at the communication line with a scope the signal looks much worse when I shift is vs. when I do not. This is the level shifter:

1. Is trying to shift the signal to 24V for the distances in between the units a good idea or not?
2. If so, how can I build a simple circuit that achieves that without sacrificing rise time on the data edges?

Answer 1

You've gotten into a different realm here, and you'll need to do rather more than you think. Although, you might (might, I say) get lucky.

You are proposing to transmit ~400 kHz digital signals over 20 - 30 meters. You can do this, and it's straightforward, but not the way you suggest. You cannot use just any bundle of wires. Increasing the voltage will only get you in trouble. You'll need twisted pair (coax would be even better, but it's overkill here) and dedicated transmit-receive ICs. Fortunately, both are pretty cheap.

Comments have suggested, and I agree, that RS-485 would be an excellent starting point. You'll only need 5 volts, and you already produce that. A good place to start is here You'll see from Figure 6 that your data rates and lengths are perfectly fine. You can use a single full duplex chip at each WS2811 IC, one part to convert the TWP signal and drive DIN, the other to take DO and drive the next segment of TWP. You'll need to follow the termination resistor requirements faithfully. Note that, of the two wires in a twisted pair, both wires carry the signal, and a third wire is needed to connect the ground between the units.

However, there is a chance that you can get lucky. You can use what is called series termination and perhaps it will work. In this case, all you need to do is put a resistor on the DO output and drive the TWP with the other end of the resistor. It would look like this

^{simulate this circuit – Schematic created using CircuitLab}

The resistor must be matched to the impedance of the twisted pair, and is usually about 90 ohms, so that's what I've shown. YMMV, though. 100 or 120 ohms may work better. In principle this will work just fine, but I'm not certain if the WS2811 DO output has the necessary drive to produce the speed you want. It is worth a try, though. It has the problem that, if your DO output is not quite strong enough, or your termination resistor is not quite the right value, you may have reliability problems. It also assumes that the DIN pin of the IC has very high impedance and will not load the TWP, and this appears to be true - but I make no guarantees. However, since it's cheap and simple you don't have much to lose by trying it.

EDIT - And don't try to twist your own wire. To get consistent performance you need a consistent number of twists per inch, so you'll need to buy your wire.

Answer 2

1. Page 3 & 4 of WS2811 chip datasheet has the timing, the highest frequency may be said 2 MHz (T0H).
   
2. The signal driver capacity is as low as 10mA.
   These two tell the devices are intended to be located (very) close.
   

MOSFET bi-directional level shifters are likely analog switchs that opens the channel, not buffered. Thus, the driver sees the load.

It is right reasoning trying to improve S/N (signal to noise ratio) by increasing the amplitude. Meantime, that cause many adverse effects as well. For your understanding, the signal has to reach to the threshold level in time, which is the percentage of the full swing regardless the voltage. That can easily end up in slowing speed. There are other problems as well, associated to the high(er) voltage.

Thus, the answer, "trying to shift the signal to 24V for the distances in between the units" is not good idea for your applications. It could be a good idea for some other applications. Some of the lighting products have 48V (need to verify) signaling.

Since you are concerned about the speed, that I agree, you would check some of high-speed communications media, like RS422 & RS485. High-speed (>10M) parts are available from DigiKey and Mouser.