Examining HDMI Performance Signature & Collective

The HDMI interface is a somewhat complex interface, primarily due to its multi-function responsibilities. These functions consist of High Speed Digital Video (TMDS), EDID InfoFrames, Audio Packets, Supply Voltages, HotPlug Trigger, tight timing requirements and an I²C Serial interface for communications.

All of these functions are transported by performing five mandatory tasks that must work in harmony for the HDMI interface to support both audio and video reliably. It is the reliably part we are most interested in.

These five tasks become a collective process that must work together as a group for the entire system to work:

• 1. Video (TMDS Balanced Line)
• 2. Timing for Data, Encryption, Triggers, etc.
• 3. Supply Voltage (the life blood of the system)
• 4. HotPlug (Trigger confirmation)
• 5. Intelligence (HDCP, EDID)

Over the years we (DPL Labs) have been testing HDMI products and what we learned many years ago was that the products, cables in particular, had a tightly defined dynamic range in which to operate. Scores of tests were executed on hundreds of products to try to learn just why HDMI had so many intermittent issues and failures. Each reported event was sampled and examined in an effort to discover the root cause.

Adding to the confusion was that in many cases products would function normally in some systems but in others they would fail. After a while, it became obvious that the reason there were so many unanswered questions on these failures was due to the character makeup of the entire system environment.


Boccaccio's book HDMI Uncensored is now available. Check out this video of his book signing at CEDIA Expo 2012.

A system environment consists of all the associated HDMI products in a system: source products such as Blu-ray players, displays, switches, matrix systems, cables and AVRs to name a few. It turns out all have a distinct DNA that makes up each part of the collective.

Since each part of the collective’s functionality is really analog, it was a relatively straightforward process of discovering that the performance of each part can vary quite a bit within the defined dynamic range for operation to be reliable.

As an example, we sampled a host of source products that varied in DC output as much as +/- 8 percent. Some source products were capable of providing as much as 300ma of DC current, where others only supplied the HDMI minimum of 55ma.

So depending on the length and gauge of the cable and the amount of current being consumed by the sink, this can directly influence the final voltage, with the DC low-end limit of just 4.7 volts. You can see how this one part of the collective can change dramatically when considering the entire system environment, especially as more and more products are introduced.

The scenario can and does occur with the remaining parts of the collective also. In the case of video, for example, there is a defined dynamic range in which the data must operate. It spans about 780mv peak to peak measured inside the Eye Pattern that is generated under test. Here again the Eye itself cannot get too big or too small.

If the system is working with a small Eye to begin, with the rest will either pass it in unity, pass it at a dangerously lower level, or pass it with a higher level. It all depends on the remaining bits of hardware that make up the system environment. In many cases the sink, source, and even the cables can have a direct effect on the outcome due to each device and the integrity of their collective.

The bottom line is that because every product has a defined dynamic range within its own collective, and each collective will yield its own overall Performance Signature or DNA embedded within each HDMI product.

Case in Point

We visited a site that had a large matrix setup with seven displays. All but one worked. Here the integrator used Cat 5 ATDs (baluns) and due to the aggregate of all the products within that one display system there just was not enough video data (peak to peak Eye) to establish any image. By reducing the length of the output cable on the ATD by just 1 meter (0.7db), enough output was provided to get inside the dynamic range needed for both audio and video to function.