Anti Aliase filter between DAC3152EVM and ADS4125EVM

Hi Prashant,

First, please review the following two app notes to get familiar with the basics of high-speed ADCs and DACs.

http://www.ti.com/lit/an/slaa523a/slaa523a.pdf

http://www.ti.com/lit/an/slaa510/slaa510.pdf

.. Questions..

1. Can I directly connect an SMA cable between DAC3152EVM o/p and ADS4125EVM i/p ??
   1. [Matt] Yes you can
2. Do I need to put Anti-Aliasing filter between DAC and ADC EVMs ??
   1. [Matt] Yes, review app notes above
3. What, if I don't use such Anti-Aliasing filter between DAC and ADC EVMs ??
   1. [Matt] You will see spurs in your captured spectrum that are related to the DAC images, they will alias back into the first Nyquist zone of the ADC.
4. What would be the sampling rate for my sine wave input ??
   1. [Matt] That's up to you, based on your system needs...

Regards,
Matt Guibord

Hi Matt,

Thanks for reply and providing a useful material. Infact earlier I referred the same for starting with basics.

My sampling rate of ADC is 25 MHz and frequency of interset is 5MHz, which is very much less than nyquist frequency (fs/2) = 12.5 Mhz.

As I am directly connecting a SMA cable between DAC3152EVM and ADS4125EVM, so hopefully no noise will be added to the signal.

Questions :

1. Still I need to have an anti-alasing filter ??
2. If yes, what would be cut-off frquency of that filter as my expected Fin is 5 MHz ?? (May be 5.5 MHz)
3. What if I use a low pass digital filter inside FPGA with same cut-off frequency rather using this low pass anti-aliasing filter before ADC ?

Prashant

Prashant,

1. Still I need to have an anti-alasing filter ??
   1. Yes, otherwise you will capture the DAC images
2. If yes, what would be cut-off frquency of that filter as my expected Fin is 5 MHz ?? (May be 5.5 MHz)
   1. DAC Fs/2 or ADC Fs/2, whichever is lower.
3. What if I use a low pass digital filter inside FPGA with same cut-off frequency rather using this low pass anti-aliasing filter before ADC ?
   1. Will not work. The DAC images are caused by the digital-to-analog conversion, they cannot be digitally filtered. Review DAC images and ADC aliasing (undersampling).

Regards,
Matt Guibord 

Hi Matt,

Thanks for answering my questions !!!.

What I understood from theory that, "reconstruction filter" should be used at DAC o/p to smooth-en the analog o/p AND

"Anti-Aliasing filter" should be used before ADC to remove the high frequency images.

But in most of theory says => aliasing comes into picture when we are under-sampling the ADC, but as I mentioned my sampling rate 25 MHz for 5 MHz i/p.

Also I am unable to find an effective document on DAC images. I would be happy if you can share document related to DAC images.

Thanks

Prashant

HI Prashant,

You're correct that the "reconstruction filter" is used at the DAC output to smooth the analog output, however this filter is really meant to remove the DAC images that fall in the higher Nyquist zones. These images will fall at Fs +/- Fout, 2*Fs +/- Fout, etc. So for a 100 MSPS DAC that outputs a 10 MHz tone, we'll have images at 90 MHz, 110 MHz, 190 MHz, 210 MHz, etc. These images are a result of the zero-order hold output of the DAC. Filtering removes the high frequency images, which will leave us with only the tone in the first Nyquist zone.

The "anti-aliasing filter" is used to remove frequency content from undesired Nyquist zones. If your signal falls in the first Nyquist zone, you'll want to use a lowpass filter with a cutoff around Fs/2 to remove frequency content in the 2nd, 3rd, 4th, etc Nyquist zones. If your signal falls in the 2nd Nyquist zone (undersampling), you'll need a bandpass filter to remove content in the 1st, 3rd, 4th, 5th, etc Nyquist zones.

Let's assume we don't use a filter between the DAC and ADC and we're using the 100 MSPS case above. The DAC will have frequency content at 10 MHz, 190 MHz, 210 MHz, etc. The ADC will then capture all signals (10, 190, 210 MHz) and convert them to digital. Although you have captured your 10 MHz desired tone, the 190 MHz and 210 MHz signals are also captured and they will alias back into the first Nyquist zone (based on undersampling theory). This will cause your 10 MHz signal to be corrupted by the other signals falling back. You're correct that aliasing comes into the picture when "under-sampling", but you are essentially undersampling the DAC images because they fall in higher Nyquist zones.

If you use a filter between the DAC and ADC then you will be left with only the 10 MHz tone. Now when the ADC captures the signal, it only sees the 10 MHz signal and there will be no corruption from higher frequency content.

In the case of a DAC hooked directly to an ADC, the frequency content is very predictable. However, let's say we use a DAC to transmit a signal through the air. You may own a certain frequency band that you can transmit in, but regulations require you to keep noise in other frequency bands below a certain amount. If you don't filter after the DAC, then the DAC images will fall into someone else's frequency band and you will not pass regulations testing. For an ADC, you may want to capture a signal at a certain frequency but there will be higher frequency content that will alias back into your band and corrupt your desired signal.

Hope this helps...

Regards,
Matt Guibord 

Hi Matt,

Thanks for replying with most useful contents.

Your answers to my several posts helped me to understand basics of DSP quickly.

Thanks once again !!

Prashant

Hi Matt,

Returninmg back to this post !!!

I have designed an anti-alising filter similar to circut given in slea085.pdf (Application Report  SLEA085 TVP5146 Anti-Aliasing Filters). Almost similar circuit is designed except inductor taken with 1.2 uH. With TINA software I could see it's frequency response as expected.

But useually such filters are kept very near to ADC pins, but now with this approch there is lot off circuit (like transformers) in between this filter and ADC.

Questions::

1. Is this a correct approch for placement of such anti-alias filter ?
2. As DAC3152EVM o/p is just 250 mV (Vpp), the i/p ADS4125 further dropped to 2.2 mV(Vpp) because of this filer. Will ADS4125 accepts this signal?
3. Do we need to amplify this signal to 2 V(Vpp).
4. Where we should place this amplifier circuit (before or after anti-alising filter), if at all required ?

Prashant

Hi Prashant

You may see ripple in the passband due to extra parasitics, but the filter should still apply attenuation at the desired frequencies.

The ADC will likely still sense the signal, but you should try to operate the ADC at higher input levels for best dynamic range. 2.2 mVpp is 59 dB down from full scale.

If you use an amplifier, it would likely go before the filter so that the filter would remove noise from the amplifier. It's possible to split a higher order filter into two lower order filters with one before and one after the amplifier. This is all application dependent and you really need to analyze your system and the expected frequency content in order to design the filter/amplifier circuit.

Regards,
Matt Guibord