SN75DPHY440SS: Use to extend CSI-2 D-PHY interface to over 1 meter

Matthew

The DPHY440 can support max data rate of 1.5G, so 1.2G is within the DPHY440 support capability.

The DPHY has both receiver equalizer and transmitter pre-emphasis to compensate the channel ISI loss. Three RX equalization levels are available at 0 dB, 2.5 dB, and 5 dB at 750MHz. Two TX pre-emphasis levels are available at 0dB and 2.5dB. Depending on the channel loss and the placement of the DPHY440, we can use both RX equalizer and TX pre-emphasis to achieve the optimal result.

You can also download the DPHY440 S-parameter model from TI.com and simulate your system channel setup to estimate the performance.

Thanks
David 

Hello David,

Thank you for your rapid reply.

The question that I am asking goes beyond the information available in the data sheet.  I'm asking for guidance on compensating a CSI-2 D-PHY channel over 1 meter long.  Figure 17 of the data sheet illustrates a design example of a a 12 inch (30 cm) channel, on the high side of generally recommended D-PHY lengths, but the data sheet does not comment directly on the factors limiting the length of the channel.   Using information from the data sheet and the S parameter model, one could simulate a longer channel, or one could build a prototype to obtain experimental data.  TI may well have insight about the efficacy of the SN75DPHY440SS in compensating longer channel lengths. 

Can you provide advice based on either device capability considerations, or on simulation or experimental results on the ability of the device to compensate channel lengths over 1 meter?

Thank you, M.Reich

Matthew

How much is the insertion loss of the 1m+ cable at 600MHz Nyquist frequency (1.2G data rate)? If the loss is within the DPHY440 RX equalizer range, then the DPHY440 will able to work with the cable.

Thanks

David

Hello David,

I don't have an estimate of the cable insertion loss.  This is why I'm asking the question, I'm asking for TI's experience and / or knowledge, since getting reliable loss data is not trivial for an actual cable.  Also, loss at one frequency  probably is not the only factor; differences in frequency response will cause different propagation times between lanes, which push against the equalization range.  If we want to pick one frequency as representative, shouldn't that be at least the 3rd or 5th harmonic?

Thank you, M.Reich

Matthew

The function of the equalizer is to compensate for the early roll off of the cable by boosting the high frequency portion in the response characteristic. With this added high frequency response component, the over-all response is maintained up to the Nyquist frequency, and the original signal can be transmitted in as close to its original form as possible So we ask for the insertion loss of the cable at its Nyquist frequency as the first step to see if the DPHY440 is suitable to this application.

Thanks

David

Hello David,

Yes, the data sheet equalization specifies equalization at  at 500Mhzx and 750mHx, and de-skew betwen 20 and 750Mhzx, both corresponding to he clock frequency (fundamental) range.

If TI has any information of the maximum length of a cable that the device will compensate, or information on expected, observed or typical loss and skew fo typical cables, I'd appreciate it.

Thank you, M.Reich

Mathew

Two cables of the same length could have different loss profile depending on the material, the build, etc of the cable itself. So it is more accurate to define the cable in terms of the loss, instead the length.

Not knowing the loss of this 1m+ cable, you can also order a DHY440 EVM and directly evaluate its performance against the cable.

Thanks

David 

Hello David,

Certainly the profile of loss (including phase) vs frequency depends on cable type.  For directly transmitting  D-PHY for multiple image sensors, some type of ribbon would be the most economical cable implementation.  As you suggest, I could do  experimental development, or use simulation.  A worthwhile first step might be a rough calculation or simulation to determine whether  the loss and differential delay are likely within the DHY440 compensation range. 

As mentioned at the beginning of this dialogue, SerDes is the alternative solution.  At the expense of increased complexity,  it certainly is becoming the more conventional solution, is almost certainly less risky over manufacturing variation, and may be less expensive.  I'm not sure that evaluating cable characteristics to support the DHY440 solution would be worthwhile.

Thank you, M.Reich

Matthew

The cable characteristics include the loss and the differential delay, etc. This is the first step in estimating whether the DPHY440 can support the particular cable. And with the DPHY440 EVM being available, you can also verify the cable interoperability through direct measurement. Second step would be through simulation with the simulation setup match as close as possible to the actual system design implementation. Third step would be the actual design in itself. By the time with the actual design in place, you should already have high confidence that the design will work base on the first two steps.

Thanks

David