This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DS90UB954-Q1: deserializer Isolation options

Part Number: DS90UB954-Q1

We're using a DS90UB954 deserializer with a 20-ft coax cable in an application that requires 4 kV of isolation and 60601 compliance. Our current solution, which works fine, takes the MIPI data into an FPGA which converts it to a wide 20-bit parallel bus that goes into a stack of digital isolators. While this works, we're making architectural changes that mean we need to keep everything MIPI between the deserializer and the processor.

There are two places we could put the isolation barrier: at the high-speed serial input, or at the MIPI output.

Let's start with the MIPI side. So, I've never seen a MIPI isolator solution. LVDS isolators don't work, since MIPI isn't purely LVDS; there are high-voltage single-ended parts of the protocol. I'm not sure you could use transformers to isolate the MIPI bus since there's quite a bit of a DC component to the signal — does anyone have thoughts about this?

Moving onto serial input isolation, the inputs to the DS90UB954 are capacitively-coupled, so we thought about just using those capacitors as an isolation barrier, but they're rated at 0.1 uF, which would cause way too much leakage current with a mains-applied 60601 test (we need to say in the 100-1000 pF range). I tried reducing the capacitance but the deserializer drops out.

So that brings me to transformer-based isolation. We'd like to try something like this:

I ran the numbers on the back of a napkin and I think high-quality 2.5 Gbps Ethernet magnetics will have low enough return loss to pass a signal across them (while it's passable, the numbers aren't great, and I could be wrong on my math, though). I'm just not sure which input configuration to use on the DS90UB954. Do we connect it as shown above? Should we tie the CT pin to a capacitor connecting to GND to establish a reference voltage? Or some other DC operating point?

I assume we have to change the configuration register to put it into STP mode instead of coax mode? (I'm not exactly sure what that register change does....)

Or we could keep the chip in single-ended mode and do something like this:

There's not a lot of information in the datasheet about what the inputs look like in the DS90UB954, so I'm not sure the correct configuration. Or maybe the idea won't work at all for some other reason I'm not thinking of?

We spun a little breakout board that has several resistor/capacitor options that will allow us to evaluate different configurations:

Unfortunately, so far I've been unable to get it working, and there's a lot of variables to change, so I'd love some guidance!

  • Hi Jay,

    For the RX ports on the 954, we have recommended return and insertion loss requirements (Table 227) that you need to meet in order to guarantee robust operation. Every time you're introducing components on those lines, you need to make sure you're not distorting the signal and the return loss requirement is still met.

    Our chip has ESD rating (table 6.2). The RX pins have higher than 4kV ESD rating so you shouldn't need extra components. We haven't seen a use case that uses an isolator before, but I would be concerned if the isolator is able to maintain the FPD-Link signal. You may be able to use an ESD diode on the RX+ line, but once again, this will introduce potential impedance mismatch.

    On the serializer datasheet, you should be able to find more information on the FPD-link signal.

    Best,

    Jiashow

  • Regarding your comments about our ESD diode, sure we can DNI that on the board and add it in later if needed.

    Do you have any substantive comments about the isolation, though? I'd love to get some feedback from a TI engineer who's familiar with the inner-workings of the IC.

  • Hi Jay,

    Placing any active device in the FPDlink signal path is a major risk. We have not qualified this and it is expected to drastically alter the link performance, and possibly kill the back channel entirely. I would not recommend using an isolator on the FPD link lines.

    Best,

    Jiashow