2
\$\begingroup\$

Many times you see datasheets showcasing high speed transciever interfaces being tested using PRBS data. Intuitively, I think the motivation is to test the medium using a signal of somewhat even distribution across the spectrum - which aught to make things more challenging and prove the system functions under these conditions (let me know if I'm wrong about that)

But anyways, examining closer how the transceivers are built raises doubt in this approach, because those high speed interfaces often have an encoding layer (8b/10b, etc...) that's designed to achieve DC balance and disparity before sending data over the transmission medium. So regardless of how the original data looks (PRBS), after encoding it will be scrambled.

So why bother with PRBS ?

\$\endgroup\$
1
  • \$\begingroup\$ This is a good question. \$\endgroup\$ Commented Nov 21, 2024 at 10:49

1 Answer 1

Reset to default
1
\$\begingroup\$

It depends on the interface and test.

Many chips don't care what you send through it, i.e. non-encoded NRZ bitstream or 8b10b encoded bitstream.

The point is you send a known bitstream, and can receive and verify it has no errors or the allowed amount of errors.

And with that bitstream you can also examine the eye pattern opening and other potential issues in the waveform.

Of course if you have some chip which encodes data for you, you can still receive and verify if you receive it correctly, like broadcasting the pseudorandom stream in ATM frames or Ethernet packets etc which do require a certain format and headers to look like valid frames.

\$\endgroup\$
2
  • \$\begingroup\$ My question is about high speed interfaces that use transceivers - practically all of them use some form of encoding. For example a JESD204B (which is 8b/10b encoded) - you can find many FPGA IP datasheets online that "prove the interface works" by injecting PRBS. What's the benefit in doing so ? Why not simply send a counter as your bitstream ? \$\endgroup\$ Commented Nov 21, 2024 at 9:20
  • 1
    \$\begingroup\$ @shaiko Well if a random data pattern goes through well, so does any other seemingly random-looking pattern like 8b10b encoded data. The PRBS replaces the 8b10b encoded data at the PHY so the PRBS is not again encoded with 8b10b. It tests the "analog" part of the link. Sending a counter directly is not a good idea because your counter is a very poor sequence of bits, that mostly repeat, and may violate the DC wander specs of AC coupled analog link. So it has wrong spectrum while PRNG is white noise with all frequencies present with basically no DC content. \$\endgroup\$ Commented Nov 21, 2024 at 10:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.