Original Xbox
Vertical Lines in 480i

January 7, 2023

A few times I had noticed vertical lines in the original Microsoft Xbox boot animation when using a high definition component cable and I wanted to try and get to the bottom of why this happened on some systems but not others.  Here is a comparison of the high definition output with the bars on the left, and another system using the same high definition cable on the right which shows a cleaner picture.  These samples have been stripped down to just the Y luma/Sync channel to eliminate any chance that the colour channels were the cause.

Image capture from the Microsoft Xbox boot sequence showing a jailbar effect.
Image capture from the Microsoft Xbox boot sequence showing no jailbars.

Through a close look at the various systems that I had available I was able to relate the issue to systems which used the Conexant video encoder.  The next question to answer was are the vertical bars in the display typical for a system with a Conexant encoder, or has something deteriorated with the system that has resulted in the bars appearing?

Close up view of a Microsoft Xbox main board, showing the Conexant video encoder.

Despite my best efforts and Google-Fu, I was unable to find any instance where someone had a system with a Conexant chip that wasn't showing the veritcal bars.  Of course you cannot prove a negative, so while open to someone coming forward with a working Conexant system, my money is on a design flaw with the Conexant which causes the problem.

There seemed to be two main solutions offered for the issue that I was seeing.  The first was to force the system to output in 480p which would clear up the problem, and the other was to replace capacitors.

As I poured over the boards for the systems I had which showed the jailbars, the capacitors looked fine.  I did some in circuit testing of the ESR for the capacitors and everything checked out to be in spec.  I decided to remove capacitor C3B2 100uf 25V and C5A4 which was a 680uf 16v as these were both in close proximity to the Conexant.  I checked them both out of circuit and they both came back fine.  I replaced them anyway to give the best chance of clearing up the bars.  When these didn't solve the problem I proactively replaced the rest of the electrolytic capacitors just to rule them out as a cause, but as expected the bars remained.

Next it was time to hook up a Conexant system and a Focus system to an oscilloscope to see if I could capture some waveforms which could quantify what was being seen on the screen.  At first this was a bit like trying to catch a flash of lightning because the only screen I had seen which displayed the problem was the boot animation and with the movement on the screen from the animation it was nearly impossible to grab signals of the same field between two systems for comparison.

A lightbulb went on sometime overnight when I realized that since I had the capability to run an HDTV Test Application on both Xbox systems, I could generate an NTSC test pattern in 480i and then direct the system output through the scope before going to the display.  Since the test pattern is static I would be able to more easily compare the signals!

Oscilloscope output showing two waveforms, the top showing a fairly clean trace, and the second showing a high amount of jitter in the signal.

In the picture above the Y Luma/Sync channel for the Focus encoder which does not show jailbars is on Channel 1, and the Conexant system which suffers from jailbars is piped through on Channel 4.  You can see slight differences in the sharpness of the lines for Channel 1.  A closer look:

A closer image of an oscilloscope output showing that Channel 1 in yellow is a much cleaner signal than Channel 4 in blue.

You can see the amount of activity in the waveform of Channel 4 in the above picture compared to the first few nanoseconds of the signal for Channel 1 which represents the same portion of the field.  These variances are representing the luma variance of the pixels across the scan line.  Stacking multiple fields (lines) up you can begin to picture the result which would be an image filled with varying vertical lines.

Here is a real close up of the signals to further illustrate the issue.  In cases where the colour being shown is consistent across the screen you would expect to see a relatively flat signal line, which is demonstrated by the Focus encoder on Channel 1.

Another close up of an oscilloscope waveform capture showing a stark difference in signal quality between Channel 1 and Channel 4.

Given the testing and results it is my hypothesis that the Conexant video encoder has poor performance in 480i which results in jailbars or vertical lines.  It is not a symptom of failing capacitors, and the fact that it's tied to the Conexant chip explains why some systems with other encoders are able to display 480i just fine with an HD cable.  Hopefully this will help prevent some people from chasing their tails on this problem.

The forum post on OGXbox.com found here is the original thread where I worked through to this conclusion for those who are interested.