XboxHDMI - Update #4 - Part 1

Posted by Dustin Holden on

Hey, Dustin here from MakeMHz with another update. If you haven’t had a chance to read our other posts then make sure to check them all out before continuing on!

XboxHDMI - Update #3
XboxHDMI - Update #2
XboxHDMI - Update #1

This post will be a little different than the previous posts and will be broken up into 3 parts. 1.0-1.1 Improvements, Shipping Update, and Quality Assurance/Testing. Each one will be posted throughout this week. 

* This post was originally written on December 15th and has since been rewritten due to the site and blog being down.

Trouble with the 1.0 Xbox

It was brought to our attention shortly after our last update post that two separate users encountered issues with their 1.0-1.1 install that was an undocumented failure case. Immediately we updated the installation manual with a note for users installing on a 1.0-1.1 to wait until further investigation was done.

“We are currently investigating an issue related to SOME revisions of the 1.0 motherboard. We've narrowed down the issue and will be making it our top priority to have the guide updated to reflect the changes that may need to be done.”

Investigation

With confirmation from two customers of the issue, it was very important for us to drop everything and begin investigating immediately. The first step was to perform the installation on more 1.0 systems and see if we could reproduce the issue. During development and testing, we had installed the XboxHDMI on at least 4 other 1.0/1.1 systems and never ran into this issue. Searching the bin of Xbox boards we quickly found 2 more boards and performed the XboxHDMI install and was able to reproduce the issue on one of them.

Noisy Power and Non-latching Signals

The Xbox GPU and TV encoder run in a Pseudo-Master Interface lock-step set up with the pixel clock going from the GPU to the encoder and back to GPU. During boot the system will hang if the GPU fails to properly latch on to the clock out signal from the encoder. (The GPU is set up in master mode if the system boots to an error screen and bypasses the need for a stable clock out signal from the encoder). 

Example of excessive power supply ripple on a 1.0 motherboard.

 

After extensive testing it was determined that the issue came down to poor power rail design, failing capacitors, and small component variances in Xbox systems. This also explains why the issue was not discovered in our testing as systems that have been recapped, both motherboard and PSU, can work correctly without any additional modifications.

In later motherboard revisions Microsoft added additional filtering to the pixel clock in, pixel clock out, and the field data lines.

Reverse engineered schematic of the GPU and TV encoder section for revisions 1.2 to 1.5

Solution

The solution is to add additional filtering to the data lines the same way that other revisions do. This should clean up the signals and fix the issue of the Xbox hanging on boot. The cleanest solution for this will be via a small QSB, quick solder board, attached to the bottom of the Xbox motherboard.

The QSB board is an easy install on the bottom side of the Xbox motherboard.

Conclusion

In about two weeks we've quickly identified an issue related to some revision 1.0 and 1.1 motherboards, investigated the issue, designed, and implemented a workaround.

While a full recap can resolve this issue in some cases, we're going to be rewriting the installation guide to reflect the additional step of installing and verifying the installation of the QSB.

We just received our first samples of the fix QSB yesterday, December 28th, and are working to verify it on as many boards as possible before signing off on it. We hope to complete the testing this week. If all goes well, and our initial testing says yes, we will be placing a bulk order at the beginning of next week. 

It's very unfortunate that this issue caught us off guard. We developed and tested the XboxHDMI on every board revision throughout the development of the product and never ran into this issue. At the end of the day, we hope that this discovery not only resolves the issue with the XboxHDMI but also the issue of 'coma' 1.0 Xbox systems that have not been modified.

We hope to have this wrapped up as quickly as possible. In our next post, in a couple of days, we will be covering our plans on sending out these boards at no cost to orders that have already shipped out. The QSB will also be included in all 1.0 - 1.5 kits going forwards.