Post Go back to editing

Timing Specifications

Thread Summary

The user is concerned about the AD7328 ADC's DOUT timing at 20 MHz SPI clock, where t₄ (SCLK falling edge to DOUT valid) is 43 ns max. The final answer clarifies that in SPI Mode 3, data should be sampled at least 8 ns after the falling edge of SCLK to ensure stability and avoid sampling the previous data, within the 43 ns t₄ window. t₄ and t₇ timings are differentiated: t₄ is the data access time, and t₇ is the data hold time.
AI Generated Content
Category: Datasheet/Specs
Product Number: AD7328

Hello,

I am working with the AD7328 ADC in an SPI interface, and I need clarification on the timing specifications related to DOUT.

In the datasheet, t₄ (SCLK falling edge to DOUT valid) is specified as 43 ns (max). At the same time, the device supports an SPI clock frequency up to 20 MHz (SCLK period = 50 ns), but im using 12.5 MHz spi clock

My confusion is:

  • At 20 MHz, the next rising edge of SCLK occurs after 25 ns. This is earlier than the maximum t₄ = 43 ns, so in the worst case, the data may not yet be valid when sampled at that edge.

  • Does this mean that at 20 MHz the system is expected to sample DOUT at a later point in the cycle (e.g., on the falling edge, or with a shifted clock), rather than directly on the rising edge?

  • Or should we rely on the typical t₄ values (which are faster than 43 ns), meaning the “20 MHz max” spec assumes that in practice DOUT will usually be valid in time?

  • Could you clarify the recommended way to reliably sample DOUT when operating at 20 MHz, given these timing numbers?

  • And also want to clarity of differentiation of t4 and t7 timings

Thank you for the guidance

  • Hi,  .

    The AD7328 operates in SPI Mode 3, where data is updated on the falling edge of the SCLK.

    In this mode, the timing parameter t4 refers to the maximum data access time after the falling edge, which indicates how long the device may take to make the new data valid and accessible. Meanwhile, t7 defines the valid hold time of the previous data, showing how long it remains stable after the clock transition.

    Based on these specifications, I recommend sampling the data at least 8 ns after the falling edge of SCLK to ensure stability and to ensure that you won't be sampling the previous data, as long as the sampling occurs within the 43 ns window defined by t4.

    Warm regards,
    Jo

  • Thanks Nathan, thanks for the reply