A.D.C Anti Alias Filter !!!

Status
Not open for further replies.

jardenblack26

New member
Does anyone know if the Teensy 3.x boards have an analog anti-alias filter as part of the 2 built in ADC's and what it's bandwidth would be?

Thanks much!
Jarden
 
There is no built-in anti-alias filter. You have to design it as an external circuitry depending your needs and the sampling rate. For example, the signal bw which you want to sample is 20kHz (typical audio application) thus, the sampling rate must be at least twice that bw, that's why most audio applications sample at 44.1kHz or 48kHz. Usually, a 3 or 4 pole Bessel (better phase) or Butterworth (better passband flatness) filter, build around two op-amps as Sallen-Key filters will be sufficient.

If you don't want to struggle with that, you might consider using the Teensy audio shield. Its codec IC, the SGTL5000, has integrated filters (and a higher resolution, 16bit instead of the Teensy's 12bit).
 
there was a discussion on anti-alias-filter in the audio board here:

https://forum.pjrc.com/threads/44111-Anti-Alias-filters-on-SGTL5000?highlight=alias+filter

However, I am still not convinced the SGTL5000 has a built-in anti-alias filter. The reason is that in three of my projects I use the SGTL5000 audio board with sample rates higher than 44.1ksps and all frequencies come through. [1.) bat detector up to 48kHz, 2.) longwave time signal detector up to 96kHz, 3.) Software defined radio up to 192kHz]. But maybe the SGTL5000 codec has digital anti-alias filters that automatically match the sample rate ? Can someone help here?
 
Jarden,

depending on the bandwidth of the signal you are interested in, you could also use a higher-than-necessary sample rate for the ADC. By this oversampling the anti-alias filter could be reduced to a very simple RC filter, maybe?

All the best,

Frank DD4WH
 
AFAIK, the SGTL has a FIR filter with fixed coefficients as an anti-aliasing filter. It is clocked with the word clock, thus, it adapts to the sampling rate automatically.
 
AFAIK, the SGTL has a FIR filter with fixed coefficients as an anti-aliasing filter. It is clocked with the word clock, thus, it adapts to the sampling rate automatically.

I tried to check after your first claim of this, but unfortunately, I could not find any indication in the TRM, any hint where you have this info from?
 
However, I am still not convinced the SGTL5000 has a built-in anti-alias filter. The reason is that in three of my projects I use the SGTL5000 audio board with sample rates higher than 44.1ksps and all frequencies come through.

But did you witness any aliasing if you applying an input signal that's above half the final sample rate?

Some time ago I did quite a lot of testing with sine waves from my Siglent SDG5082 function generator. I saw the response rapidly roll off somewhere in the area of 20 to 21 kHz. If I apply a 30 kHz sine, I get I2S input that's pretty much the same as the noise floor with the input shorted. If there weren't any anti-aliasing filter, I'd imagine getting quite a lot of garbage with spectrum around 14 kHz (Nyquist aliasing).

While I can't find any specific mention of this in the SGTL5000 datasheet (which lacks all but the most basic electrical specs its ADC & DAC), I'm pretty sure its ADC is indeed an oversampling delta-sigma converter with digital filter. But I'd be very surprised if that filter turned out to be a FIR type. It's very likely CIC.
 
But did you witness any aliasing if you applying an input signal that's above half the final sample rate?

Paul, thanks for your test results, they are pretty convincing for 44.1ksps sample rate.

To be honest, I did not check or test that systematically for higher sample rates. You are probably right that one would have to check that systematically by injecting frequencies above half the sample rate and look at the output of the codec.

I am not familiar with CIC filtering: is it possible that a CIC anti-alias filter in the codec also scales with sample rate?

All the best,

Frank
 
The CIC filter is a special case of a FIR filter, thus no contradiction. And if it’s clocked with something proportional to the word clock, it would automatically adapt to the sample rate and thus to the input bandwidth:)
 
Paul, thanks for your test results, they are pretty convincing for 44.1ksps sample rate.

To be honest, I did not check or test that systematically for higher sample rates. You are probably right that one would have to check that systematically by injecting frequencies above half the sample rate and look at the output of the codec.

I am not familiar with CIC filtering: is it possible that a CIC anti-alias filter in the codec also scales with sample rate?

All the best,

Frank
If it is indeed sigma delta followed by CIC digital filter, then no steep AAF is needed, only the typical single pole LP filter at the input of the ADC.
 
The CIC filter is a special case of a FIR filter, thus no contradiction. And if it’s clocked with something proportional to the word clock, it would automatically adapt to the sample rate and thus to the input bandwidth:)

AFAIK, Anti aliasing filters have to be analog before the ADC and not digital filters after the ADC as CIC filters are.
 
Status
Not open for further replies.
Back
Top