Post Go back to editing

EVAL CN0540. SYNCHRONIZATION OF MULTIPLE AD7768 DEVICES

Category: Hardware
Product Number: EVAL-CN0540

Hello,

I continue working on my design and I have more Doubts. We have started from a one AD7768 (Eval CN0540) but we are designing to four channel because we need 4 IEPE independent channels. Also we need:

    - Each channel must be isolated from the others.

    - We have one crystal per channel so we haven´t a common MCLK. We are going to enable 4 clocks at the same time

    -We need to switch off the unsued channels on each iteration (it can change in each cycle, on client demand) 

We don´t know how to do 4 channels Synchronization because reading AD7768 datasheet we have two options:

1) If a signal that is synchronous to the base MCLK signal cannot be provided:

 

2 If a SYNC_IN signal synchronous to the base MCLK can be provided

We are in the First option (because we don´t have a SYNC_IN signal Synchronous to the base MCLK). The first option involves to have one of the ADC as a main device because it receives the first synchro signal and supply this signal to the others through the Sync_out. I think We can´t use the first option because we must switch off the channels doesn´t use in each measure cycle, so that we can´t choose an ADC as a main device due to It could be innactive in this measure cycle but it could be active 4 cycles after.

So, how can we Synchronise Our 4 channel design? Could anyone explain how and help us?

Thank you very much,     

Parents
  • Hi TSTEngineer,

    We can only recommend the synchronization methods detailed in the datasheet.

    The first thing to resolve is the lack of common MCLK / one crystal per channel. Four separate crystals cannot be easily synchronized**, and by extension, the four devices cannot be synchronized. It may be possible to synchronize the start of the first conversion across the four devices to within one MCLK cycle, but after that (and even over the course of the first conversion), the devices will have slightly different conversion times.

    One approach would be to start with a single oscillator, and fan out to the four devices through high-speed digital isolators. Then use one of the synchronization methods detailed in the datasheet, perhaps using SYNC_IN# (again, a common signal fanned out to the four devices.) Note that in order to ensure exact synchronization across devices, the setup/hold requirements for all SYNC# signals with respect to their MCLKs must be met - accounting for skew through the isolators.

    Regarding activating /deactivating devices, this may not be advisable because these are sigma-delta converters, it may be better to run all devices all the time, and simply ignore data from devices from which data is not required. This question is slightly out of scope for  Reference Designs , consider searching  Precision ADCs .

    -Mark  

    ** Sometimes crystals or other oscillators DO lock to one another, often undesirably, and such behavior can't generally be relied upon: https://en.wikipedia.org/wiki/Injection_locking 

Reply
  • Hi TSTEngineer,

    We can only recommend the synchronization methods detailed in the datasheet.

    The first thing to resolve is the lack of common MCLK / one crystal per channel. Four separate crystals cannot be easily synchronized**, and by extension, the four devices cannot be synchronized. It may be possible to synchronize the start of the first conversion across the four devices to within one MCLK cycle, but after that (and even over the course of the first conversion), the devices will have slightly different conversion times.

    One approach would be to start with a single oscillator, and fan out to the four devices through high-speed digital isolators. Then use one of the synchronization methods detailed in the datasheet, perhaps using SYNC_IN# (again, a common signal fanned out to the four devices.) Note that in order to ensure exact synchronization across devices, the setup/hold requirements for all SYNC# signals with respect to their MCLKs must be met - accounting for skew through the isolators.

    Regarding activating /deactivating devices, this may not be advisable because these are sigma-delta converters, it may be better to run all devices all the time, and simply ignore data from devices from which data is not required. This question is slightly out of scope for  Reference Designs , consider searching  Precision ADCs .

    -Mark  

    ** Sometimes crystals or other oscillators DO lock to one another, often undesirably, and such behavior can't generally be relied upon: https://en.wikipedia.org/wiki/Injection_locking 

Children
No Data