Embedded Systems Conference
Breaking News

Friday Quiz: PAM Is Coming

Page 1 / 5 Next >
View Comments: Newest First | Oldest First | Threaded View
User Rank
Re: Opening the Eye
Chris.Loberg   1/14/2015 2:40:35 PM
The symbol rates for these eyes in answers 3 & 4 is 25GBaud.

The "ugly" eye pattern is neither really bad nor really great :-) . If this signal was captured after transmission over a bad backplane; it could be quite worse. The eye you're referring to was from a normal transmitter connected through a short 1M coax. So, it is an "OK" setup but not an optimal one like a VSR-type setup.

zzglen also asked; How much of the eye closure in that ugly eye can be attributed to the oscilloscope probes and overall measurement bandwidth?  This can be rendered much better as I mentioned above; we intentionally wanted to illustrate some distortion/ISI in the eye for this example. But with the use of a remote sampling head like the Tektronix 80E10B; one can expect even better results.

Another question from zzglen: How does one even probe something like this? I don't expect probes being used in these measurement scenarios, it's more likely the customers DUT will have SMP's or a standard industry connector like QSFP or CFP on the device. In the case of connectors; testing would start by connecting a fixture from company like Wilder Technologies to the device which then interfaces on a differential pair to the micro-coaxial inputs to the oscilloscope or remote sampling head. 

For the nicer eye in the answer to question 4 - it was acquired from a 25GBaud symbol rate signal but we used the remote sampling heads of the 80E10B in the Tek DSA8300 Sampling Ocilloscope for a much cleaner, noise-free eye involving a shorter cable pair.

One last comment - to view a working PAM4 demo, feel free to stop by DesignCon 2015 Jan 28 & 29th. As Martin Rowe mentions; there will be active demonstrations on display in the exhibit areas.

User Rank
Re: Whoa on q 4
Chris.Loberg   1/13/2015 8:13:49 PM
Whoa indeed. Totally true, Bert22306! PAM-4's 1E-5 and added signal complexity is placing more stress on receivers' decision thresholds than traditional NRZ. This was an oversight on my part and I apologize if anyone failed their quiz as a result. The correct answer to #4 is: A

User Rank
Re: Opening the Eye
MeasurementBlues   1/12/2015 10:02:26 AM

I looked at the original images that Chris sent but they too are unclear and too small to read. They are mostly for the purposes of the questions so I didn't ask for larger images.

Tek has an app note that might help. It's about eyes and jitter but not so much about PAM4. Physical Layer Tests of 100 Gb/s Communications Systems.

Here's a short video on PAM4 analysis.

Here's a one-hour webcast on PAM4, Verifying PAM4 Performance in 28G & 56G Designs.

If you really want to know about PAM4, jitter, and the like, then you should go to DesignCon two weeks from now.

User Rank
Opening the Eye
zeeglen   1/10/2015 9:19:56 AM
What is the symbol rate of the eye pattern shown in the answer to question 3?  Can barely make out the sweep as 8ps/div, but cannot see now many divisions from center to adjacent symbol.

At first glance this looks like a very ugly eye pattern, but considering the high symbol rate it is nice to get any eye opening at all.  That makes it a beautiful eye pattern.

A more technical blog as a follow-up would be interesting.  Is this eye pattern typical of current state of the art, or is it being used as a horrible example of degradation?  What were the conditions of the test to get this measurement - backplane, copper cable, optical?  How much of the eye closure can be attributed to the oscilloscope probes and overall measurement bandwidth?  How does one even probe something like this?

The eye pattern of question 4 is much nicer, but again what symbol rate and conditions?

A nice effect of these Friday Quizes is their educational value. Chris Loberg, would you care to write another blog with some more information?

User Rank
Whoa on q 4
Bert22306   1/9/2015 8:42:36 PM
I did just fine until we got to question 4.

All else equal, the BER is expected to be HIGHER with PAM4, or indeed with anything that is more spectrally efficient, than it is with NRZ. A higher BER means more expected errors, in a given data stream.

When you say:

1E-5 BER for PAM4, as opposed to 1E-12/1E-15 for NRZ

certainly 1E-12 is fewer errors, i.e. a lower BER. So the answer to

What are expected BER results for PAM4 compared to NRZ-based signaling?

PAM4 results in more errors, higher BER, all else equal, compared with NRZ. The answer is a.

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed