AD7675 family, timing questions

Hi all,

 

I'm working with an ADC AD7675 and I'm misunderstanding few parameters about timing.

 

The application is quite simple : a microcontroller is dealing with the ADC and memory. Datas are recorded from the process under test continuously at the beginning (50ksps during about 200s) and after, when some events detected by the microcontroller occur, few points are recorded(could be 10 or 20 points in 200ms, every 30s - 50s).

 

I started tests with 8bits A/D integrated in the microcontroller. It works but more precision is required. My problem is simple. When I worked with the integrated A/D, timing was simple and correspond to action : Start acquisition, stop acquisition, start conversion, interrupt at conversion end, data in a given register.

 

With the AD7675, I understood that CNVST start conversion and stop acquisition, BUSY show the end of conversion.

If I use PD (power down) to save energy, how long will it take to warm up the converter (Tension ref always ON)? After a PD, are input pins at '0' ?

 

What about starting acquisition ? Acquisition must length at least "t8", 8.75µs. On the diagramme 11, MODE is defined as acquire or convert, no wait period. If I want to get 1point every 10ms during 200ms (total 20 points, each point acquisition time 10µs), how can I make it ?

 

The only way I see would be to make a RESET. Is 10ns (t9) is enough to make a true '0' on the input pin ? Is it possible to keep RESET active for a longer time ? Does RESET only affect input state ? (no effect on data out for example)

 

Is this timing correct ?

PD end -> warm up ->RESET ->wait 10µs->CNVST->poll BUSY LOW-> RD->RESET hold during 10ms (approx, 10ms from the previous RESET)->((repeat 20 times to get 20points))->PD start

 

The same thing to work at 50ksps, without PD and holding the RESET during a much shorter time. Is it correct ?

 

Many thanks for your help!

Kind regards,

Adrien

  • Hi Adrien,

    Why bother powering the part down? The power of this device scales very well with throughput so if your not converting very often it is very low power. If you look in the datasheet it consumes 17mW @ 100ksps and 15uW @ 100sps. In power down it consumes 7uW max so not really much of a saving by powering down completely. See figure 10 for even more power dissipation numbers. It looks like if your not converting the power is in the region of <1uW. This way you don't have to worry about power up times.

    Best Rgds,

    Alan

  • Hi alan,

    .

    Thanks for your answer. It's true that I could not bother with power down. I use the ADC with parallel port. I need to put databus on high impedance (to read back data). As I need to deal with CS or RD, I thougth why not PD ? It's true that RD will solve power up time trouble. Do you think it's the rigth way to use RESET to start the acquisition ? Troube to keep it active for a "long time" ?

    Best regards,

    Adrien

  • Hi Adrien,

    You do not need to reset the part to start acquisition. The acquisition time is basically the time that the part is not doing a conversion. So at slow throughputs you have a very long acquisition time. As you can see from Figure 10 the part can convert at very low throughput rates (10sps) and even lower if necessary. I am not aware of any issues with very slow throughput rates.

    Best Rgds,

    Alan

  • Hi alan,

    In fact I'd like to get the same acquisition timing during the different way of sampling, and keeping it around 10µs of acquisition duration. This for 2 reasons : first I'll keep an "integration time" constant. At the begining the signal varies fastly. So I'd like to take around 50kSPS. After signal vary slowly so lower sample rate is possible. But I'd prefer keeping an acquisition time constant to get constant signal treatment. With a much longer acquisition time, I think that sensor noise will be filtering in a different way.

    And as a second point, the most important to my mind, in the  process timing I reserve a time sequence with nearly no activities (switch communtation, DAC transition, memory exchanges, ...) to try to reach low noise level. With an acquisition lasting several ms, logic transitions perturbations may to be taken into account during acquisition. With a constant 10µs acquisition time, I could get a single routine for the microcontroller with an easier control on the 'nearby' activity.

    That's why I was looking for a way to control the acquisition timing. The only way I see is to program a reset to get only 10µs of acquisition, even if I get 10ms between the end of conversion and the starting of the next acquisition. I'm not sure about the way the reset works or if it's the rigth way to do this.

    Best regards,

    Adrien

  • Hi Adrien,

    A longer acquisition time will not integrate more noise, the input is sampled at the CNVST edge. The bandwidth of the AD7675 is 3.9MHz so it's time constant is fast relative to your measurement interval. If you really want to keep the acquisition time at 10us you can use reset or do a dummy conversion first, then give 10us acquisition time and do the conversion you want to capture. If you use reset it may reconfigure the device to a state you don't want.

    Best Rgds,

    Alan