• TI Thinks Resolved

TS3A27518E-Q1: QSPI MUX Needed

Part Number: TS3A27518E-Q1

We need to add a multiplexer for the QSPI Lines on our board. We are looking for something which is automotive grade and can handle 80MHz data rate for QSPI Signals. We need 5 channels (4 IO and 1 Clock). Having 6 would be OK too , we can add CS line as well but if not 5 will work since CS is low speed. Do you have any part that we can use which has low input capacitance and low propagation delay which can handle QSPI Signals? Our QSPI lines are 1.8V .

We have our QSPI lines shared between MCU and EQ4 and we need to add some isolation in between so that when one IC is reading the flash over QSPI, the other IC doesnt interfere with it.

Any suggestions on this would be great.

  • Hello Wes,

    Thank you so much for the details you have provided us. This really helps us in how much support we can provide and how quickly we can do it without assumptions.

    Based on the part you mentioned, this will meet your bandwidth requirements. However, some things you will want to mention to use to help us better provide more support:

    1) What is a good range for acceptable propagation delay?

    2) What is an acceptable range for input capacitance?

    3) Is the operation of a 5 or 6 channel 2:1 MUX what your system needs? As of now it is only clear based upon the device mentioned.

    I have highlighted what the TS3A27518E-Q1 can offer below:

    Let us know how else we can help!

    Louie

    If I have answered your questions, please hit Verify Answer.

  • In reply to Louie:

    Have you heard/seen of anyone using this for QSPI or CAN applications?

  • In reply to Joe Howard:

    Hello Joe,

    The switches in our portfolio are general purpose switches. We fall in a large category of applications due to this definition. Although, I have not personally seen this, I also would not be surprised to see it. QSPI and CAN, like many signals have their own specs that need to be met. It is up to the design/system engineer of said application to decide whether or not our switch meets the spec(and protocol) that allows for QSPI or CAN applications. We will simply help guide and understand whether or the spec requested is reasonable of our switch.

    Thank you,
    Louie
  • In reply to Louie:

    Hello Louie,
    my customer wants to use TS3A27518E-Q1 to switch between two QSPI masters, which both have to access a single QSPI Flash device. Looking into the datasheet I am missing the propagation delay from NCn/NOn to COMn, which is critical to the maximum SPI speed. Do you have exemplary propagation delay figures? Or shall this information be derived from IBIS simulation? If yes, does the IBIS model support extended temperature range for automotive application?

    Thanks,
    Peter.
  • In reply to Peter Aberl:

    Hello,

    Unfortunately we currently do not have propagation delay between input(NCn/NOn) to output(COM) pins characterized for this part. 

    However, I took this part into lab and did a quick test with this part by feeding an i2c clock (SCL) signal pulled up to Vcc with a 10k resistor and fed it into the input pins.

    Here are the scope captures that I have for the rising and falling edge probed at the input and the output pins:

    the rising edge takes a little longer to pull up, but that can be fixed if i used a smaller pull up resistor. But if you look at the falling edge capture that I have attached, you can see that the delay is almost non existent (with respect to a 10ns time scale). 

    Unfortunately you will not be able to do temperature simulations with the IBIS model, these models are only for functionality and how the internal parasitic may effect your signals.  

    Could you provide us with more details to what exemplary propagation delay would be for your customer? 

    Thank you,

    Louie

  • In reply to Louie:

    Hi Louie,
    thank you for the measurement. Though, I2C is by far less critical compared to the QSPI use case. The QSPI read speed is 40 Mbps. During QSPI read the TS3A27518E-Q1 switch is two times passed, first the QSPI clock propagating to the QSPI flash device and 2nd the QSPI flash device data signal propagating back to the QSPI master. Having said this, the propagation time through TS3A27518E-Q1 is critical and needs to be in the low nanoseconds range. Could you please repeat the measurement with a square wave signal (tr/tf < 1ns)?

    Moreover, the QSPI signals in the customer application are 3.3V. The TS3A27518E-Q1 datasheet is not very clear on this. Does the TS3A27518E-Q1 allow switching of 3.3V digital signals (V+(nom) = 3.3V) ?

    Thanks,
    Peter.
  • In reply to Peter Aberl:

    Hello,

    Although we do not specify directly in the datasheet, we do support signals of Vcc and digital at 3.3V. Please see the snippet below:

    The input signal you can put into the switch is determined by the Vcc, but in this case it is limited to what we spec in the abs max specs in our datasheet.

    As for the propagation delay, it is actually determined by your rise and fall time. If you look at the scope traces above, you will notice that the fall time signal has a fall time of about 10ns. If you were to test the signal with a fall time which is faster than the one I have shared, your results will be almost identical, meaning the propagation delay will be almost nonexistent, with respect to your input signal's speed. The same can be said about the rise time, in my captures the rise time is slower, but the propagation delay for that signal is also nonexistent with to the input signal, which means if you make the rise time faster, your propagation delay will be smaller. This happens because you charge and discharge any parasitics that cause propagation delay faster with faster edges. If there isn't noticeable delay in these falling or rising edge speeds, there will not be noticeable propagation delay with faster rising or falling edges. 

    Thank you,

    Louie

  • In reply to Louie:

    Louie,
    thanks for the confirmation. This answers my questions.

    Peter.