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.

TMUX1574: Switches & multiplexers - INTERNAL forum

Part Number: TMUX1574
Other Parts Discussed in Thread: TXB0304

Hi Team,

My customer have a SPI signal lost issue, I mentioned in below ptt.file, and I also attached customer schematic as below.

Could you help on this issue? 


Cascade several MUX Level shift signal lost issue 20220517.pptx




  • Hey Kai,

    Cascading multiplexers can certainly degrade the signal. Adding any component will add more degradation and noise to a signal. There seems to be quite a bit of oscillation or noise here. It seems that the SPI signal is going from the MB side through some 8 or so different chips or connectors. This will certainly be impactful on the signal. Every connector and chip will add some capacitance, inductance and resistance. The traces themselves will also impact this (so it's important to try and minimize the trace lengths).
    That being said, if I'm interpreting the waveforms correctly the MB side itself where the TMUX1574's are function properly and get fed into the TXB0304. Typically cascading the muxes does reduce the bandwidth significantly but the TMUX1574 has a bandwidth of 2GHz and the SPI looks to be at 20MHz so we definitely have room to budge here to cascade a couple of the muxes. Additionally, while the TXB0304 isn't in multiplexer portfolio, I took a look at it and it seems that it has internal buffers so the signal is redriven.
    From my understanding of what is happening, it seems that the signal is fine at the MB side because it's being redriver but on the DB side the load is just too high and signal doesn't look so good because of it. 
    Your last suggestion in the ppt would be my suggestion as well. A buffer of some sort on the daughter board side to redrive that SPI signal and isolate the MB load and DB loads.


  • Hi Rami,

    Thanks a lot for your comment, very useful!

    If follow the suggestion and put a buffer at DB side, do you think is a good solution here?

    Please help to suggest a part to me, since I'm not familiar enough with the signal chain so far.

    Thank you!


  • This problem is probably not related with the muxes; it belongs into the logic/voltage translation forum.

    The TXB is bidirectional. The output drivers on both sides are always active, but they have a very high output impedance (1 kΩ) to allow other devices to override the voltage. Due to the low DC drive strength, it has edge accelerators, but those have a short timeout. So the TXB cannot be used for loads larger than about 100 pF (see section 8.4.2 of the datasheet).

    You can use any buffer for unidirectional signals, but the four data lines of Quad SPI are bidirectional. There is no good solution to buffer bidirectional signals without a separate direction signal. Quad SPI should never be used with long traces or cables.

  • Hi Clemens,

    Thanks a lot for your comment, learn a lot from you two ~