SPI Buffers - Any Good Recommendations from TI?

Hi Ivan,

Can you tell me a little more about your application?

What type of speeds do you need?

Is the signal bidirectional (ie is data being sent both ways on individual lines)?

How many signal lines?

What is the actual length of the cable? What kind of cable? Any shielding? What environment? Might there be any noise?

I²C requires special handling because of the bidirectional signals. Other protocols like UART or SPI use only unidirectional signals, and there are many electrical solutions that can be used for all those protocols.

See Comparing Bus Solutions (SLLA067).

Hello Emrys,
The application is for an industrial project.
Speeds are upto 20MHz over a cable that's roughly 4 inches. Cable used is Mirco-Coax.
This is SPI so it's unidirectional.
Signals on SPI are CS, MOSI, MISO, CLK , so the total signal count 4.

Actually, our TI support team had recommended the following for this: ttp://www.ti.com/product/SN74AVC4T774


We've sense proceeded with that.

Thank you,
Ivan

Hello Clemens,

The length of the cable is roughly 14 inces, within a micro-coax bundle (... other LVDS signals exist within this cable). The environment is industrial so yes, there will be noise.

These are SPI signals... we've been recommended to proceed with the following: ttp://www.ti.com/product/SN74AVC4T774

Just curious, what solutions have you used for this?

Thanks,
Ivan

In an industrial environment, my first thought would have been RS-422/485, but if you're already using LVDS elsewhere, you can of course reuse that.

The AVC774 gives you single-ended signals, but for a cable much shorter than 1 m, that's probably perfectly fine.

Thanks for the details Ivan,

Just so you know - there are other versions of SPI that have more data lines and even ones that operate bidirectionally. I never know which 'flavor' people mean when they mention SPI.

Sending a 20 MHz _analog_ (ie sine wave) signal over a 4" coax cable should be relatively safe since the wavelength at that freqency is around 15m. However, sending a 20 MHz digital signal probably has some much higher frequency components (depending on edge rates). If that 20 MHz signal has edges switching in 1ns you can see frequency components up into the hundreds of MHz and even low GHz, which then start to matter over a 4" cable - group delay and reflections can cause over/undershoot issues as well as edge rate problems.

Adding a damping resistor to the output of the line driver is a typical method of reducing ringing from reflections and line capacitance - something around 33 ohms in series with the output typically works well.