2 AAF Tradeoffs at Low Frequency and Low Sampling Rate

How to create more dynamic range (DR) in your system application is simple, just eliminate those pesky spurs in the noise floor. Usually, the HD2 and HD3, that is the second and third harmonic distortion, respectfully, are what typically define the DR limit for the converter. Removing spurious sounds simple on paper, but this concept can lead to other constraints that make this difficult to employ in the real analog world.

We know that having a spurious free dynamic range (SFDR) of -75dB or so is relatively achievable in today’s modern MSPS converter technology without much effort. See Figure 2-1. Assuming clean power, clocking and input signals are provided, the example below, shows the ADC3664 sampling at 125MSPS in the second Nyquist zone with an analog input of approximately 70MHz. For more on sampling in the second Nyquist zone or sub-sampling ADC techniques please refer to Precision labs series: Analog-to-digital converters (ADCs) and Ansys Nuhertz FilterSolutions.

However, if more DR is required for your application, say -85dB or more of SFDR. Then we typically have one of two options to tradeoff in the analog domain. Create a narrow band anti-aliasing filter, or AAF, around the center frequency of interest. Or go for a higher sampling rate ADC, you still need the AAF, but this relaxes the AAF constraints by a fair amount. Let’s explore these concepts deeper.

Say you want to filter the 94MHz IF tone in the second Nyquist zone with 125MSPS ADC. This means you need to design a filter with sufficient rolloff to decrease down the -85dB+ SFDR dynamic range requirement in less than a 62.5MHz (+/-31.25MHz) passband or Nyquist zone. See Figure 2-2. This significantly ups the order of the AAF design, the number of the components and variance in component tolerance also increases therefore, creating a filter that is difficult to make realizable and repeatable.

To prove the point, using a simple filter modeling tool, see How anti-aliasing filter design techniques improve active RF converter front ends, we need to design and create at least a 9th order filter even to start to get close to that -85dB SFDR requirement. See Figure 2-3 for the simulated frequency response plot of the filter design. This is a 9th order Butterworth topology, centered at 94MHz, with 10MHz (or ±5MHz) of passband.

As discussed, the number of components can also increase with such a high order filter. In this case, 28 components can need to be configured as per Figure 2-4. Keep in mind, this is just for the filter, this does not include any resistive elements that can also likely need to be added, depending on the ADC’s common mode voltage needs, any back terminations or other data sheet recommendations. Filters of this size also take up a significant amount of real estate on the printed circuit board, or PCB, using at least a 1085 mil x 200 mil total space for this specific example. See Figure 2-5.

One other note worth mentioning, low frequency filters typically run up against the Physics limit in terms of their size. Therefore, components sizes in the nH and uH region typically run much larger to accommodate lower frequencies and are typically available in 0805 package sizes or larger. This is another reason why so much area needs to be dedicated to this type of high order, low frequency AAF design.

Lastly, to be fair, the AAF design shown in Figure 2-5 is differential, which is a commonly used in implementations between an amplifier and ADC interface. If using a single-ended AAF approach and a balun to interface with the ADC’s analog inputs, this can half the number of AAF components required, as shown above. However, the size of the balun and extra ADC components to finish off this type of interface can vary. For a deep dive on AAF design, please see How anti-aliasing filter design techniques improve active RF converter front ends