We are having some trouble with implementing a simple DVI-D (HDMI) switch function. This is an embedded application. Our PCBA is an intermediary intended to select between 3 sources and source forward to a Silicon Image sink device. Si91250, on another board. Two of the TMDS361B sources come from the outside world (outside our enclosure). Typical cable runs are 10m to 20m. So, we choose the TMDS361B with its adaptive cable equalization, extended reach. The third source is local inside our chassis from a Silicon Image HDMI transmitter device routed over 150mm of HDMI cable, twisted and individually shielded, to our board with the TMDS361B. The output of the TMDS361B is routed to another Silicon Image HDMI receiver device using another 150mm segment of HDMI cable. Typical clock rates are a 148.5Mhz. Data rates are around 1080 Mbps.
The TMDS361B appears to do a pretty good job performing equalization. Looking at the output of the TMDS361B it looks okay. Except, there is some deviation, jitter, in the amplitude of its outputs, clock and data pairs. It appears that the Silicon Image device is having trouble staying locked onto the output of the TMDS361B. Problem occurs with all three sources passed through the TMDS361B.
If I feed the S.I. receiver other sources, including short or longer up to 10m of cable, it works. If I take the output of the TMDS361B run it over 300mm of HDMI cable to a DS34RT5110 evaluation kit, then through another HDMI cable of any reasonable length, it works. This leads to suspecting that the output of the TMDS361B is unable to drive even a very short 150mm segment of cable well enough to satisfy the S.I. receiver. Unfortunately, we do not have access to the receiver schematics nor the datasheet for the S.I. device.
Testing the TMDS361B with different TMDS rise and fall times does not cure the issue. Works best with the default setting, fastest. Testing indicates the TMDS361B performs good cable equalization. We suspect the source outputs are not up to the requirments of this “system”.
The TMDS datasheet says; “TMDS361B is not intended for source side applications such as external switch boxes.”, on the first page. Okay. But, what does that mean? Does this mean that the drive strength of TMDS361B is not adequate to drive cables? Does this mean that it cannot drive our very short 150mm cable? Is it only capable of chip to chip on the source side? If this is the case, what data points in the datasheet should I inspect to determine this with this device and others?
It is interesting that the TMDS361B can drive a DS34RT5110 over a longer 300mm cable. I have not tested all lengths. But, it works at 2m. Yet, cannot satisfy the S.I. device over only 150mm (.15m) of cable. Comparing feeding other sources directly to the S.I. device verses the TMDS361B using various lengths of cable, the TMDS361B works over longer runs than the other device. This indicates that the adaptive equalization of the TMDS361B out performs the other device. But, our testing and results don’t explain why we can drive the “other” part, the SI 91250 from the TMDS361B over any length of cable?
We’ve tested using various sources including a pseudo random pattern originated from an SDI generator and converted to HDMI. The pseudo random test is the most telling. Eye patterns are what you would expect. Except, the previously mentioned amplitude jitter on the output of the TMDS361B. There is some clock jitter, U.I. interval. But, it appears reasonable. We have been unable to measure at the sink due to lack of schematic and datasheet to the other device. We do not have control over the sink. So, we have to find a way to satisfy what it wants with our TMDS361B circuit. We are considering a board spin that would add a DS16EV5110A to the output of the TMDS361B. But, this is not yet tested. (It appears the DS34RT5110 that we used for testing is in the process of be discontinued. Why?)
Suggestions and advice?