Post Go back to editing

AD7684 Pre-Processing-Best Methods Approach


I have a couple of questions on this part, and the best way to pre-process the incoming signal.

I have a low frequency AC sensor that is single ended and requires AC coupling. With the present circuit, I'm using an ADA4096 (due to it's excellent OVP) as a buffer. I then go to a AD8616 Bessel-Butterworth LPF. From there, it goes to another programmable gain stage, and then into a 12 bit micro ADC. Total gain is about 80dB.

I want to change this to take advantage of any common mode noise from the sensor by going to a differential circuit, then to a 16 bit ADC. After reading through AD data sheets, it appears the best device for the frequencies I am working at is the AD7684.

I realize that at some point I need to go from single-ended to differential. However, I still want to have the OVP from the ADA4096 in the circuit, at least as a buffer for the sensor. which point does it make sense to go from single-ended, to differential, to take advantage of the benefits of differential input?

Where should the filtering and gain stage be, and should it be differential? I have a requirement for filtering off 60Hz signals, although I am working below that. So if the filter were differential, it would need to be order 8 to 10 to provide the same transition I presently have.

The data sheet for the ADA4941 makes it an ideal candidate for a pre-ADC buffer and gain amplifier.

What will be lost in using differential if I were to input buffer, filter, gain and then go single-ended to differential using the ADA4941, vs differential from the input right through - keeping in mind I still need to AC couple and have OVP.

Thanks for any advice or opinions on this!



  • Hi, Gary.

    I'm looking into this. I'll get back to you in two days.



  • Hi, Gary.

    We have a wide range of ADCs in our portfolio with SE inputs, differential inputs, and others configurable for both. Deciding how to drive it will depend on your application. I understand that you chose to use a differential ADC to take advantage of its rejection of common mode interference and your sensor accepts single ended signals only. We have newer SAR ADCs in our portfolio, may I ask for the reason you selected the AD7684?

    How was your programmable gain stage implemented - PGA, or discrete? High-order filtering should not be performed at the ADC driver as this would limit bandwidth at the ADC input and prevent adequate settling of artifacts resulting from the switching of the cap DAC input of the ADC.  The ADA4941 is a suitable choice for converting the signal from SE to differential, and you could add gain at this stage, taking note that there is a tradeoff between gain and bandwidth.



  • Regarding the ADC choice, what do you recommend? How about the AD7916?

    The "programmable" gain stage is pretty complex - an ADG708 multiplexer selecting the gain resistor in the op-amp feedback loop.

    You said: "High-order filtering should not be performed at the ADC driver..."

    What's high order filtering? 2nd, 3rd....8th? Can you be more specific?

    Sorry, my question still isn't answered, and I was hoping to really get that resolved. I understand best performance can be achieved using differential amplifier and ADC (or pseudo-diff ADC)

    Recap of objectives and problem:

    1 - Single ended sensor (present configuration) that must be AC coupled

    2 - Need to LPF 60Hz by at least 60dB (stop band), all signals of interest 5 to 30Hz

    3 - Need front end gain of 80 to 90dB

    4 - Would like to have 16 bits resolution to discriminate small signals

    Since I'm starting out with single ended, AC coupled sensor, where in the chain, keeping the above in mind, should I introduce the single-ended to differential stage?



  • Must be a difficult subject.

    I guess this can be closed...unanswered....

    Thanks for trying, AD.

  • Hi, Gary.

    Apologies for the delayed response.

    I'd recommend a newer ADC such as the AD7989-1, an 18-bit part with the same throughput as AD7684. You might also want to check the 16-bit AD7916 that has higher throughputs. The best place in the signal chain to implement single-ended to differential conversion is at the ADC driver. The ADA4940 should be suitable for this application. You may refer to this article here, it includes a discussion on a SE-to-Differential data acquisition system with the ADA4940 used as a driver.

    High order filtering, referred to your 60-Hz low-pass filtering, i.e., you should implement the Bessel-Butterworth LPF earlier in the signal chain, and definitely not at the input to the ADC.

    We also add a single-pole RC filter in between the ADC driver and the ADC (when using a differential ADC driver, placing one filter on each output).  This is used to limit wide-band noise and to reduce the settling artifacts caused by the switching of the capacitive input of the ADC, but it would not be useful for the 60 Hz LPF mentioned earlier.  This is because this filter’s time constant must be short enough to allow these settling artifacts to settle to the resolution of the ADC (16 bits for the AD7684 and AD7916 and 18 bits for the AD7989-1).  For example, with a 30 Hz maximum input tone at 5 Vp, a filter using R = 160ohms and C = 2.7nF should work.  Here's an article which has an in-depth look at how to select the RC components.



  • This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

    Thank you,
    EZ Admin