How can I design an output filter for a DDS generator using the AD9834 IC?

[Jdo300]

Hello All,

For those of you who haven been following my last two threads, I finally decided to go with the Microcontroller/DDS combination to make my three-channel sine wave generator (since I have yet to find one in existence at all that doesn't cost megabucks).

I have the microcontroller part worked out already (I'm working with the dsPIC30F2020 since I already have some code written for it from my pulse generator project). The DDS IC I plan to use is the AD9834. I have two of them in house tight now on surfboards and I just finished breadboarding up a basic test circuit to check that I A. didn't burn the IC up when I soldered it and B, can actually communicate with it.

I'll be doing that testing tomorrow but I an curious to know how to properly design the output reconstruction filter for the output waveform. I've done some searching around online to see what other people are doing and It seems like some kind of nth order elliptical filter type thing is a popular choice. However, I am very much filter illiterate (beyond the basic AC Circuits 101 High pass and Low pass filters).

I do at least know that in order to begin the design, I have to know what the maximum frequency and clock speed of the IC will be. At this time, I plan to use an input clock that has a frequency of 53.6870912 MHz. The oddball frequency was calculated so that I would have a frequency resolution of exactly 0.2Hz through the entire range. I plan to make the top frequency of the generator 10.737Mhz to keep the resolution from getting to bad. So I'm wondering what is the first steps to design a filter that can adequately smooth out the steps in the waveforms.

In case anyone is interested here's a link to the product page for more info on the DDS:
http://www.analog.com/en/rfif-components/direct-digital-synthesis-dds/ad9834/products/product.html

Any help/assistance/pointers would be greatly appreciated as I have no clue where to even start with this.

Thanks,
Jason O

 

[Pumblechook]

http://www.circuitsage.com/filter/cheby.pdf

 

[Jdo300]

Hello Pumblechook,

Thank you for the reference material. I looks like it would be very useful. though about 50% of it looks like greek to me, lol. Could someone here help me work though this as I have never taken any classes on DSP design...

Here's some questions I have after reading the document:

1. I see that they define a passband and stopband corner frequency. From checking Wikipedia, I understand that the definition of "corner frequency" is the point in a filter where the power drops to half the power of the pass band. So how does that apply two the two above terms?

2. In my specific case, the upper frequency that my DDS will produce is 10.737MHz so how do I use the information in that PDF to design an appropriate filter to work for me?

Thanks,
Jason O

 

[Pumblechook]

I take it you just want a low pass filter flat to 10.8 MHz. You need to decide the slope of the filter above cut-off. Usually a simple 3 element filter has roll off of about 20 dB per octave (10 times voltage, 100 times power) so at 22 MHz it will be 20 dB down at 44 40 dB down. By 3 element .. a pi network of 2 capacitors and one inductor. 5 element... 3 C and 2 (L) ...40 dB per octave.

There are probably websites which will calculated the component values.. Google..low pass filter and maybe +"cut-off frequency".

 

[bitrex]

Analog Devices has a nice "Active Filter Design Tool" if you want to implement an opamp output filter. You can find it here: http://www.analog.com/en/design-too...ign-sim-tool/Filter_Wizard/resources/fca.html edit:but I guess your frequency is pretty high for using an op-amp filter...

 

[Pumblechook]

Have to beware that active filters have power handling limits and may actually generate harmonics, the very things you are trying to reduce.

 

[Jdo300]

I take it you just want a low pass filter flat to 10.8 MHz. You need to decide the slope of the filter above cut-off. Usually a simple 3 element filter has roll off of about 20 dB per octave (10 times voltage, 100 times power) so at 22 MHz it will be 20 dB down at 44 40 dB down. By 3 element .. a pi network of 2 capacitors and one inductor. 5 element... 3 C and 2 (L) ...40 dB per octave.

There are probably websites which will calculated the component values.. Google..low pass filter and maybe +"cut-off frequency".

Thank you for your input. I basically want the filter to smooth out the steps in the sine wave and also filter out the clock frequency which, according to the datasheet, will feedthrough into the output as well. How do I know how sharp of a cutoff is sufficient to make the output work? I know that some people use 7th or 8th order filters (I don't confess to understand what that means), but I'm guessing it has to do with how sharp the cutoff is?

Analog Devices has a nice "Active Filter Design Tool" if you want to implement an opamp output filter. You can find it here: http://www.analog.com/en/design-too...ign-sim-tool/Filter_Wizard/resources/fca.html edit:but I guess your frequency is pretty high for using an op-amp filter...

In a perfect world, it would be nice if I could just find some IC out there that I could set and have it do the filtering for me. I don't want to over complicate or over engineer this at all, but I do want to have a clean sine wave. I'm just not sure how 'good' a filter I need to design.

Thanks,
Jason O

 

[Jdo300]

OK I just thought of something else, How does one determine how much attenuation is enough to make the output waveform 'smooth'?

- Jason O

 

[Pumblechook]

What is the application? What is the output of the DDS used for?. What level is output is needed?

The 'smoother' the output is the nearer to a sine wave you get and the lower the level of the harmonics. Even a very simple RC would smooth out the output so some extent.

Looking at diagrams of DDSs used in radio receivers they often use fairly simple low pass LC filters .. 3 or 5 element.

The ARRL Handbook is very good. Probably will have a chapter on DDS and another on filters.

 

[Jdo300]

Hi Pumblechook,

I am planning on using this DDS for a three-channel function generator application. This will interface to RF amplifiers that will output a power sine wave to drive a set of coils for some high frequency rotating magnetic field experiments. The most important thing is that the output sine waves are pure and without distortion. Especially at high frequencies above 1Mhz. But honestly, if the output sine waves are at least as decent as an analog function generator with the same range, that is more than good enough for what I need to do. Distortion is the most important thing that I have to watch out for to ensure that these rotating fields are properly implemented.

Again, I don't want to be dogmatic on the filters as I am not sure what is considered reasonable. The guy in this video here also built a DDS function generator using the same IC that I am working with:



I contacted him to ask what he did for the ouput filter and apparently, all he did was use a 20pF capacitor in parallel with the 220Ohm resistor on the output DAC. His sine waves look just fine on the scope from what I can tell. The only thing I noticed was that the output amplitude diminished significantly once he got up into the MHz range. I'm assuming this is due to the simple RC filter that he was using. So if I could design a filter that basically works as well as his, only without attenuating the amplitude significantly below the max operating frequency, that would be more then good enough for what I am doing. You say a 3 or 5 element filter is what is commonly used? that would work fine for me. I'm just wondering what the best type of filter to use is and how to determine the component values.

Thanks,
Jason O

 

[Pumblechook]

I would get one of these.. Older, second hand edition will do.

Available around the World...

I have one (somewhere) from 1982 and I think it contains tables for filter design..

https://www.amazon.com/dp/0872591395/?tag=pfamazon01-20

 

[Jdo300]

Hi Everyone,

I finally found something that will work for me. I found a schematic for a DDS generator called the DDS20 which several friends of mine own. On the schematic, it shows a nice filter stage that I can also use for my design. You can check it out here on Pg. 6 of the PDF file here:

http://www.elv-downloads.de/service/manuals/DDS20/53665_DDS_20_Board_V6_2_KM.pdf

They use the AD9835 rather than the AD9834 but the two ICs are virtually identical except one has a higher tuning resolution. So I'll just go with the IC that they use to simplify things.

@Pumblechook,

I have seen a few references to the ARRL handbook before, so I'll definitely look into getting a copy.

Thanks,
Jason O