ADAU1328 DAC output filters (would passive be okay?)

Hi There,

I'm considering using passive filters directly after the outputs of the DACs on the ADAU1328. Is this a bad idea?

I see that the datasheet recommends active filters on the outputs of the DACs.

The expectation for my design is that users will be plugging this into devices with high-impedance inputs, so I'm assuming this would be okay. (?)

Also, I couldn't find the output impedance of the DAC's in the datasheet. Does anyone know about what that would be?

Any thoughts would be greatly appreciated.


-Pete Lewis

  • 0
    •  Analog Employees 
    on Apr 19, 2021 7:00 PM

    Hello Pete,

    I wish we did have the output impedance spec but basically, you do not want it to drive too low of an impedance. The THD and other specs will gradually start to fall off once you get below around 3.6K-ish. It is not an edge of a cliff but a gradual decrease. So I would not try to drive too much of a load. The active filter is good in that it will be a high impedance input and it is easy to design a 2-pole filter for very good out of band filtering. 

    You can look at the ADAU1962A datasheet. In that datasheet we have some examples of single ended passive filters. 

    Now you mentioned "plugging in" devices. If the outputs are going to connectors and long cables then you should buffer it. There will be shorts and static charges and lot of cable capacitance in the outside world that you really want to isolate from the codec. Much easier to replace an 8-pin Op-Amp than the 48 pin codec. 

    Dave T

  • Hi Dave T,

    Wow, Thank you for the fast response!

    Very good point about cables. I should clarify that this design is meant to be more of a development tool. The outputs will be  accessible on only PTH headers, and so the user/designer will be expected to solder jumper wires from these outputs to the next "stage" in their project (most likely a short length connection to an amplifier). Even so, we still need to consider these "outside dangers" and decide on our ultimate goals in terms of product robustness.

    3.6K-ish is kind of surprising to me. But I suppose that lines up just fine if the next destination is a line level input (which I read is typically 10K). I'm curious to learn how this number would be found if it wasn't in the datasheet. Would you have to send some sign waves (or white noise?) through it with various loads and watch it on a spectrum analyzer?

    Thanks for the ADAU1962 datasheet recommendation. I checked it out, and if we go the passive route, we will most likely go with something similar to their suggested single-ended circuit.

    Thanks again!

    -Pete Lewis