Differential Signaling for High-Speed Digital Transmission by Jason O

[Jdo300]

Hello All,

I have a board that needs to send a high-speed digital signal to a second board about 5-6 inches away. The signal itself is just a 5V square wave with a variable pulse width and frequency that could go as high as 3MHz. The second board has an isolated supply and I plan to route the signal to it using a digital isolator (IL610 to be exact). I was first thinking about driving the signal single-ended, but then thought about using a differential signal since I heard that it has much better noise immunity.

The isolator is located on the destination board and I was thinking about using some kind of line driver (differential line driver?) to drive the signal through some twisted-pair wire to the second board. My question is do I need to be concerned about transmission line effects when transmitting in differential mode? does there need to be a termination of some kind like in a single-ended setup or something different?

Thanks,
Jason O

 

[ynot]

Twisted-pair wire is a excellent way to do it. You can make a differential driver yourself by using an inverter for the second line. When one switches up, the other one switches down, and vice-versa. The noise immunity comes from having a good differential receiver.

For 6 inches of wire in the air, that's about 0.5 nanoseconds of flight time (0.5 x 10e-9 seconds). For wire in air, signals travel pretty close to the speed of light, about 85ps of flight time per inch.

A 3MHz signal has a half-cycle time of 167ns. That means you probably don't have to worry about reflections too much, and shouldn't have to worry about terminating the wiring. If you did want to terminate, probably around 300 ohms (on each side) might work, I think that's about what twisted wire comes close to.

When the waveform switches, the reflections will bounce back and forth every 1ns (counting up and down), and die down pretty quick. You still have 166ns before the waveform will switch again, so it's not a big deal. You usually only have to worry about transmission line effects if the line is long enough so that the reflection time approaches the waveform time.

 

[berkeman]

Hello All,

I have a board that needs to send a high-speed digital signal to a second board about 5-6 inches away. The signal itself is just a 5V square wave with a variable pulse width and frequency that could go as high as 3MHz. The second board has an isolated supply and I plan to route the signal to it using a digital isolator (IL610 to be exact). I was first thinking about driving the signal single-ended, but then thought about using a differential signal since I heard that it has much better noise immunity.

The isolator is located on the destination board and I was thinking about using some kind of line driver (differential line driver?) to drive the signal through some twisted-pair wire to the second board. My question is do I need to be concerned about transmission line effects when transmitting in differential mode? does there need to be a termination of some kind like in a single-ended setup or something different?

Thanks,
Jason O

I think RS-485 may work for that frequency at that short distance. Use Cat-5 cable, and forward terminate with 105 Ohms at the RS-485 receiver input. Double-check the RS-485 chip specs (like from Maxim) to see if they go that fast.

BTW, does your isolator really work to 3MHz?

 

[Jdo300]

Hi Everyone,

Thanks for your help with this. I'll definitely look into the RS-485 IC for a possible differential driver.

@Berkeman,

Yes the Isolators I'm using (IL610 or IL710) can both go over 100Mbps. These aren't like regular isolators. Here's a link to the dataheets:

http://www.nve.com/Downloads/il600.pdf

http://www.nve.com/isopdf/il710.pdf

- Jason O

 

[berkeman]

Hi Everyone,

Thanks for your help with this. I'll definitely look into the RS-485 IC for a possible differential driver.

@Berkeman,

Yes the Isolators I'm using (IL610 or IL710) can both go over 100Mbps. These aren't like regular isolators. Here's a link to the dataheets:

http://www.nve.com/Downloads/il600.pdf

http://www.nve.com/isopdf/il710.pdf

Wow, fascinating devices. Thanks for the links, Jason. Interesting that they already have about 85 Ohms for the input coil -- would have been nice if it were closer to the 100 Ohms of Cat-5 cable, though.