The first part of this series explained how high-bandwidth digital content protection (HDCP) works and offered basic troubleshooting tips.

This article covers extended display identification data (EDID) and how to overcome some of the more persistent problems that arise when setting up a system.

Extended display identification data

Let’s first make this clear: a scaler usually results in a worse picture.

This is contrary to what some people believe. In the analogue days a scaler did make the NTSC/PAL/SECAM scan lines disappear, but that is so 1980s. Virtually all source signals originate as digital now. Even VGA comes from computer digital content.

For digital content, any scaling is bad. Figure 1 shows familiar artifacts that result from scaling.

To avoid scaling, and to achieve the best image quality, the source must generate the signal to match the display’s native resolution. That is the primary purpose of the EDID.

During system initiation (after a power-up, hot-plug detection or HPD status change), the display sends the EDID to the source and requests the video and audio signal format. This process is also known as a ‘handshake’.

EDID data structure is regulated by the Video Electronics Standards Association (VESA), and the current revision is 1.4.

The data is 128 bytes and is required to carry the following: ID manufacturer, ID product code, year of manufacture or model year, EDID version, EDID revision, basic display parameters and features, display x, y chromaticity co-ordinates (phosphor or filter chromaticity), preferred timing descriptor block, extension flag, etc.

VESA defines most of the computer video parameters in the EDID.

The Consumer Electronics Association (CEA), on the other hand, regulates the data of HD video and audio parameters through the CEA- 861-D Standard.

The HDMI Standard calls out the CEA-861-D Extension 3 data, which includes: video data block, audio data block, speaker allocation data block, vendor specific data block, colourimetry data block and video capability data block.

The data is comprehensive and includes the resolution, refresh rate, position, blanking, colour space, format, audio format, encoding format, channels, effects, sample rate, etc. By reading the EDID line by line you would probably know more about the monitor than its designer.

In short, EDID is display-to-source data that defines what signals the source needs to send. Unlike HDCP communication, which happens every two seconds, EDID communication happens once at system initiation.

What’s wrong with DVI and HDMI?

In an HDMI cable, there are 19 wires. Two of them – the serial data and serial clock wires – are used for the display data channel (DDC – see Figure 2). These are probably the most important wires in an HDMI cable because they carry the HDCP and EDID data.

The protocol used in the DDC lines is called inter integrated circuit (I2C). This was invented by Philips in the early 1980s to allow two-way communication over a single pair of wires among multiple ICs on the same board.

The initial bus speed was 100kbps but was raised to 400kbps in 1992 and to 3.4Mbps in 1998.

I2C is a very simple and elegant design for its purpose. It uses only two wires, yet it is a multi-master bus with arbitration and collision detection.

But DVI borrowed I2C for the DDC lines, then HDMI carried it over. This was a short-sighted choice of protocol, because the developers did not foresee that DVI and HDMI would be used way beyond the simple system of one DVD player connecting to one TV.

The maximum DDC capacitance of DVI and HDMI was a relaxed 700pF, but because of the relatively low data rate it is too limited for a pro audiovisual system.

The typical DDC capacitance of a 7.5m cable is more than 700pF, and each electronic input and output would add 50pF or more.

It becomes even worse when we use HDMI over Category 5 baluns for long-distance transmissions. The typical Cat 5e cable has an equivalent DDC capacitance of about 50pF per metre. Therefore, a 100m cable has a capacitance of 5,000pF.

In short, most systems used in the pro audio-visual sector do not meet I2C DDC capacitance specifi cations.

When DDC capacitance is greater than specifi ed, the timing would be wrong for the devices to know when the line is busy or free and whether the receiver is getting the data correctly. This would cause all kinds of data corruptions (Figure 3).

But, what are the symptoms of the DDC data collision? Because the DDC lines carry HDCP and EDID data, the symptoms of the data collision depend on which data is corrupted:

• If the error is at the initial HDCP key exchange bits, the screen will be black.
• If the error is at the de-encryption confi rmation bits, the picture will be cut off after two seconds or will fl ash every two seconds.
• If the error is at the EDID screen resolution bits, the image size or position will be wrong.
• If the error is at the EDID video format (RGB or YCrCb) bits, the screen will be blue, pink or other abnormal colour.
• If the error is at the EDID audio format bits, the sound will pop, or cut in and out.

Are these symptoms familiar? I am sure you have seen them, and now you know the cause.

What is the solution?

Ultimately, the organisations behind HDMI and DVI must replace the I2C protocol with one that suits large multidevice systems. In fact, Luxi Electronics has presented a proposal to HDMI Licensing that would employ a new protocol with backward compatibility for existing products.

However, such a solution will be a long time coming. In the meantime, there are three ways of dealing with the problems.

EDID minders – Remember that EDID is a one-time and one-way communication. The display sends the EDID data to the source when the system is powered up.

An EDID minder is basically a small data recorder. You connect this device to the display, record its EDID data, then connect the device to the source unit. The source, rather than the display, will talk to the device locally through the system to obtain the EDID.

This method works with DVI because there is no HDCP data, but not with HDMI. EDID minders cannot record and playback HDCP data.

DDC accelerators – Several manufacturers have developed ICs that reshape the DDC data’s rising and falling edges, thereby reducing the timing delay caused by the DDC capacitors charging and discharging (Figure 4).

The black line is the original DDC signal after going through a capacitor (like a long cable). The edges are round and the time is delayed. The ICs set a threshold (the blue line). When the signal voltage crosses the threshold, the ICs change state from low to high, or high to low. The output signal then looks like the red line, clean and straight.

This method helps to ease the problem but won’t eliminate it. You can see in the diagram that the first rising point of the red edge is still behind the timing of the black line’s starting point. This means there is still a delay, albeit reduced.

Some IC manufacturers lower the threshold to further reduce the timing delay, but this would make the system more subject to noise interference.

Another problem is that this circuit is usually part of the equaliser IC, which must be at the display end and only processes DDC data from the source to the display. There’s nothing on the source end to process the DDC data. Remember that HDCP and EDID data must travel from display to source.

DDC timing alternative – Because the system DDC capacitance is greater than the I2C specification, the data will collide no matter what you do. This method is aimed at making it happen less, so that the system still works.

Every smart electronic device has software that re-tries the communication if the previous attempt has failed. This is controlled by a software timer. If the second attempt succeeds, you may not even know that the first one failed (or you may notice a delay in the picture).

The problem becomes significant only when every communication attempt fails.

Each device has a different timer; and each cable and device adds different delays. The big collisions happen when the sum of the timer period and delay from one device is equal to that of another.

Every communication attempt will fail because both devices retry at the same time interval, thus they collide again and again.

When any device or cable is changed, the timer period plus delay will no longer be identical between the two elements. The repeated collision will not happen and the system will work.

You must have experienced this phenomenon: you change from a brand A device to a brand B and the system works. You are convinced that the brand A device is faulty, return it to the manufacturer and swear never to use one again.

Yet in another system, if brand B does not work you replace it with brand A and the problem is solved. You curse brand B and fall in love with brand A.

Both manufacturers say their labs cannot find anything wrong with the returned products, and now you know why.

However, once a system is installed it’s very difficult to swap devices. It involves time, money, customer anger and yet another trip to the job site.