Antialiasing Filter - Differential Low Pass Filter​

My company is using the 16-bit SAR ADC ADS8422 to sample electrical signals coming from a Transimpedance Amplifier (TIA) stage that takes incoming light from a FBG-based sensor. Voltage range: 4V. Input Signal Bandwidth: usually 0-2kHz.

Previous engineers (now gone) have implemented the typical application configuration single-pole AAF (Antialiasing Filter) of ADS8422 datasheet, but over time, they changed values of it based on access of passive components (small companies have more restricted resources). Thus, the cutoff frequency started at around 30MHz (as per datasheet typical application), then went 36MHz and at last 80MHz. Well, this is what I could deduce as there is no written justification behind these changes other than supposing a dearth of certain components and replacing them with the closest value available. It is often said the AAF is optional and that also could be the reason that made them think it was not that important to tamper the recommended parameters for the AAF.

Anyways, here is the datasheet differential low pass filter (1-pole) configuration. Also, I chose the configuration where we convert a single-ended signal (0-4V) to a differential one (-4V to +4V) using common-mode @2V to level shift.

I have perused various articles/white papers regarding the subject of anti-alias filter (AAF) at the front-end of ADCs. It is either a differential low pass filter, a common-mode filter or the combination of both. While the main purpose of the filter is to restrict the alias of higher order Nyquist zones to fold back into the operating band and creating the out-of-band noise, it seems there are far more to it than this.

Here are some of the AAF properties of the AAF that should be pondered before designing it (from what I have read):

• Out-of-band noise filtering: Cutoff Frequency choice (main aim of AAF)
• Charge reservoir: Instantaneous charge at the start of each sampling phase to the shunt capacitor.
• Ratio between AAF differential capacitor and shunt capacitor: it seems to matter in certain cases.
• Type of differential capacitor for AAF: C0G type for instance.
• Passband gain vs. ADC resolution: the higher the ADC resolution, the smaller is the effective region of the passband before the roll-off (there is no such flat passband) gets higher than half LSB error.
• AAF Settling Time vs. ADC Conversion Time: The ADC can be affected by the AAF settling time if it takes longer to settle than the conversion time. This might lead to gain loss.
• TDH consideration...
• Input Signal Bandwidth vs. Nyquist Limit: Oversampling considerations...
• Stopband roll-off steepness.

Please correct me if I am wrong in certain concepts written above.

I know the decision of setting up a 12ohms/12ohms/220pF differential filter is likely influenced by many factors besides the cutoff frequency @30MHz.

Can someone help me to understand what brought this configuration as typical application and what would happen if the cutoff frequency was to be moved at higher frequencies (36MHz with 10ohms/10ohms/220pF; 80MHz with 10ohms/10ohms/100pF)? If things were simple, the cutoff frequency would be 2MHz as it is the Nyquist limit (upper limit of first Nyquist zone). Evidently, many other parameters are thrown in.

The datasheet is quite silent regarding the matter or it speaks louder how novice I am. Anyways, it only states to limit yourself with a low bandwidth input signal and put a low pass analog filter at the front-end of ADC.

Many thanks! I hope we can have a fruitful discussion.

Shamash

 

As you noted, at a minimum, you want to rolloff the signal above the Nyquist frequency to limit noise aliasing, but normally you also want to limit the bandwidth to be no more than needed for the highest signal frequency of interest.
Thus for a 2KHz signal bandwidth I would roll-off the signal some above 2kHz (depending upon how flat you still want the signal at 2kHz).
A higher frequency roll-off just allows additional noise into the signal, even if it's below Nyquist.