Ideally yes, but it's more important to try to have the two wires in each pair see the same impedance to maintain signal integrity. In practice for the low resolutions generated by the P2 such as VGA, it tolerates a decent amount of skew. For example, the P2-EVAL wasn't length matched anywhere but it's close enough to still work well. I think in the early days Chip was even able to get something working with some simple flying leads.
The DVI spec mentions this about the allowable skew at the receiver:
Speed of light in a PCB at ~1.6x10^8 m/s means electrical signals travel about 16cm/ns. 0.4 x 1/252MHz would allow 1.6ns of skew or the time for signals to travel ~25cm. So that's a fair bit of trace length difference. Once you crank up the frequency for higher resolutions things begin to tighten more.
I don't know if the traces on the P2 edge are all the same length. That needs to be checked.
HDMI receivers likely have some ability to compensate for delays between the channels. In this experiment http://forums.parallax.com/discussion/comment/1463361/#Comment_1463361 I generated the clock signal from the smart pins. What was interesting was the clock phase didn't matter. It's not used for sampling the data, just as a frequency reference. Obviously, don't go out of your way to unbalance the lengths but at P2 data rates 1cm of trace is much less than 1 bit of data.
I think it's more important that each differential pair be routed appropriately. The length of + and - traces should be the same. The spacing between them will affect the impedance. Although at the <320Mbps the P2 can operate at, none of this should be that critical.
Nice, this should also work for the KISS board. For those who don't want to buy the full set of accessory boards it' good to have a cheaper second source.
Comments
The DVI spec mentions this about the allowable skew at the receiver:
Speed of light in a PCB at ~1.6x10^8 m/s means electrical signals travel about 16cm/ns. 0.4 x 1/252MHz would allow 1.6ns of skew or the time for signals to travel ~25cm. So that's a fair bit of trace length difference. Once you crank up the frequency for higher resolutions things begin to tighten more.
HDMI receivers likely have some ability to compensate for delays between the channels. In this experiment http://forums.parallax.com/discussion/comment/1463361/#Comment_1463361 I generated the clock signal from the smart pins. What was interesting was the clock phase didn't matter. It's not used for sampling the data, just as a frequency reference. Obviously, don't go out of your way to unbalance the lengths but at P2 data rates 1cm of trace is much less than 1 bit of data.
I think it's more important that each differential pair be routed appropriately. The length of + and - traces should be the same. The spacing between them will affect the impedance. Although at the <320Mbps the P2 can operate at, none of this should be that critical.
Edit: Thanks, @rogloh for finding the skew spec.
Mike
I've used a 1kohm resistor with success