• Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

DDS AD9851 questions....

Thread starter #1
Hi there,

I am working with the AD9851 DDS chip from Analog...

As I understand it I need to attenuate the signal just above the frequency I enter (best option). Given I am using an AD9851 with internal clock set to 180MHz I am running the DDS at 60MHz.

I have two broad questions....


1. If I choose a frequency say 100kHz where is the first harmonic and why do I have to block it out? Its been suggested I use a variable low pass filter - why?

2. Could someone run their eye over the datasheet AD9851... it says you can create squarewaves from the chip.. how do you get square waves out of the chip?

Hope someone might answer...

Thanks

Simon
 

JimB

Super Moderator
Most Helpful Member
#2
As I understand it I need to attenuate the signal just above the frequency I enter (best option).
Due to the way that the DDS generates its output signal, as well as the wanted output frequency, there will be other "image" frequency components generated. Using a low pass filter will remove (within the limitations of the filter) these image components.


Given I am using an AD9851 with internal clock set to 180MHz I am running the DDS at 60MHz.
I am not sure what you are saying here, the AD9851 has an internal 6x clock multiplier, if that is set, then a 30Mhz reference clock input will give you a 180Mhz internal clock.


1. If I choose a frequency say 100kHz where is the first harmonic and why do I have to block it out? Its been suggested I use a variable low pass filter - why?
The "first harmonic" will be at 200khz. Actually the correct term is "second harmonic".
There will also be an image frequency at (180Mhz -100khz) = 179.9Mhz.
Both the harmonic and the image will be very low amplitude compared with the wanted 100khz.


2. Could someone run their eye over the datasheet AD9851... it says you can create squarewaves from the chip.. how do you get square waves out of the chip?
There is a comparator on the chip.
To generate square waves, connect the pins IOUT and IOUTB to VINP and VINN (dont forget that IOUT is a current output, so you will need a resistor between IOUT and IOUTB to give a current to voltage conversion).
A squarewave output will then be available from pins VOUTP and VOUTN.

JimB
 
Thread starter #3
Thanks for your reply Jim,

If I am running this chip at a very low frequency say 500Hz Should I have a low pass filter just above this? I see a number of designs and they seem to only have one filter at the max frequncy of the DDS chip... Yet I get the impression that to get a perfect sign wave I need to have an adjustable low pass filter set to just above the frequency I have the DDS set at... If I had only one low pass filter set to say 65MHz then wouldnt I get those other harmonics interferring with the output frequency..

Simon
 

JimB

Super Moderator
Most Helpful Member
#4
For most practical purposes, the 500hz output with a 65Mhz LPF will be OK.
Not knowing your exact application it is difficult to comment further.
How spectrally pure does the signal have to be?

JimB
 
Thread starter #5
I am putting together a function generator... I have programmed the pic chip and that works LCD etc - PIC16F877A.. I am now faced with the analog part and I am no electronics genius... so I had the mind of splitting the signal into multiple legs based on the frequency... thing is.. I am not sure if I have to add an adjustable low pass filter... not sure what this contributes in light of the way the DDS works..

I guess what I am trying to ask is what effect not having a filter that rolls off just above the set frequency will do for the sine wave and how critical not having a filter would be to the final signal purity etc...

What do I get if I dont have a filter that rolls off just above the set frequency? What do I get if I add such a filter in?

Difficult to couch this succinctly... Simon
 

JimB

Super Moderator
Most Helpful Member
#6
OK, I think that you are worrying too much.
Just put a lowpass filter set for the highest suprious free frequency, about 60Mhz.

You dont say what the maximum frequency of your function generator will be, if you limit yourself to say 10Mhz, then make the LPF for 15Mhz.

Just to give you an idea of what you can expect, I have taken a couple of spectrum analyser traces of the output of my DDS signal source.
The analyser was connected on the output of the DDS chip (AD9850) before the LPF and amplifier.

JimB
 

Attachments

Last edited:
Thread starter #7
Hey thanks for showing me those pics and taking the time to do so... appreciated...

So its the Sampling image that I should be worried about. Looking at the datasheet I see in Figure 12 the image is Fc-Fo where in my case I am running the DDS9851 at a max of 60MHz.... so if I am running at Fout at say 10MHz then the image will be at 50MHz... I need to filter that out?

What happens to the image when you exceed Fc/2 or in my case 30MHz (more than half way?) do you need a band pass filter?
 

JimB

Super Moderator
Most Helpful Member
#8
Looking at the datasheet I see in Figure 12 the image is Fc-Fo where in my case I am running the DDS9851 at a max of 60MHz.... so if I am running at Fout at say 10MHz then the image will be at 50MHz... I need to filter that out?
I think the quick answer is yes.

What happens to the image when you exceed Fc/2 or in my case 30MHz (more than half way?) do you need a band pass filter?
When Fo is about Fc/2, on a spectrum analyser you will see two frequencies close together, if you look at it with an oscilloscope you will see something which looks like distorted AM, the modulating frequency being the difference between the two frequencies (the wanted frequency and the image).
At the moment I cant see a practical use for such an output. (Afterthought, maybe it could be used as a two-tone test for a transmitter linear amplifier - lets not persue that idea, it could get messy).

Have a look at page eleven of the datasheet, figure 8 suggests a method for using the high side image using a bandpass filter.
However I think that this is getting a bit advanced for someone who claims to be "no electronic genius". (Not that I am claiming genius status for myself, I have just played with RF for a long time).

JimB
 
Thread starter #9
Some final clarifications Jim,

In Figure 12 of the AD9851 datasheet, "Output Spectrum of a Sampled Sin(x)/x Signal" it shows the clock frequency as 100MHz. So if I use the 6x option and set the oscillator to 30MHz then my clock frequency will be 180MHz... this is Fc isnt it?

So setting my maximum frequency to 60Mhz represents 1/3 the clock frequency which I now understand is the maximum frequency for the provision of a good sine wave using a DDS...

Under this scenario, If I set my output frequency to 60MHz then this means that the image will be Fc-Fout which is 120MHz so by having a good elliptical low pass filter that filters just after 60MHz I should never have the trouble of having the image on top of Fout... that would occur at 180MHz/2 or 90Mhz.... Does this all sound essentially correct??

...and, if I did split Fout into two legs (I want some pretty low frequencies so I would dispense with a capacitor in series on the low leg) would setting the low pass filter on that leg to just above the low freq leg's maximum freq reduce those 2nd, 3rd 4th etc harmonics that add noise to the signal? ....it also seemed to me from the TPC1 on page 6 of the datasheet that these harmonics all occur below -70Db and could be comfortably ignored anyway however by contrast the image is much larger..

Wondering if you could make some quick indications (short answers etc) either yes or no for each of these questions/paragraphs...

By the way sincere thanks for taking the time to steer me straight in these issues...

Kind Regards

Simon
 

JimB

Super Moderator
Most Helpful Member
#10
In Figure 12 of the AD9851 datasheet, "Output Spectrum of a Sampled Sin(x)/x Signal" it shows the clock frequency as 100MHz. So if I use the 6x option and set the oscillator to 30MHz then my clock frequency will be 180MHz... this is Fc isnt it?
Yes

So setting my maximum frequency to 60Mhz represents 1/3 the clock frequency which I now understand is the maximum frequency for the provision of a good sine wave using a DDS...
Yes

Under this scenario, If I set my output frequency to 60MHz then this means that the image will be Fc-Fout which is 120MHz so by having a good elliptical low pass filter that filters just after 60MHz I should never have the trouble of having the image on top of Fout... that would occur at 180MHz/2 or 90Mhz.... Does this all sound essentially correct??
Yes

...and, if I did split Fout into two legs (I want some pretty low frequencies so I would dispense with a capacitor in series on the low leg) would setting the low pass filter on that leg to just above the low freq leg's maximum freq reduce those 2nd, 3rd 4th etc harmonics that add noise to the signal?
I assume by "legs" that you mean signal paths.
One path for frequencies less than (say) 1Mhz, and a separate path for frequencies above 1Mhz.

Yes

....it also seemed to me from the TPC1 on page 6 of the datasheet that these harmonics all occur below -70Db and could be comfortably ignored anyway however by contrast the image is much larger..
Yes, the harmonics generated by the DDS are "in the noise" anyway and can be ignored for most practical purposes.

Wondering if you could make some quick indications (short answers etc) either yes or no for each of these questions/paragraphs...
I think you have just turned me into a "yes man".:D

By the way sincere thanks for taking the time to steer me straight in these issues...
You are welcome, but is will cost you a bacon sandwich.:D

JimB
 
#11
Good stuff you guys,,, Your comments shed a better understanding... Remember, others are looking for answers, long after you've moved on.... Thanks JimB
 

Latest threads

EE World Online Articles

Loading

 
Top