AnsweredAssumed Answered

AD9834. Expected spur performance??

Question asked by asol on Nov 3, 2011
Latest reply on Nov 22, 2011 by BobC



I’m involved in some development work on a VHF synthesiser.


The basics of the synthesiser are:-

The final frequency range for the synthesiser is approx 100MHz to 170MHz, the channel spacing is approx 5KHz.

The DDS is fed with a 30MHz crystal TCXO.

The DDS is programmed to generate a signal of between 10MHz and 11MHz.

The DDS output is used as a reference for an ADF4002 PLL to generate the final frequency.

Final frequency output  = DDS frequency x (N/R) . N/R  is programmable and ranges from approx 10 to 17.

The PLL loop bandwidth is approx 50KHz.

Since the DDS, N and R are all programmable there are many ways to achieve the correct frequency. 


Although the system works I have issues with in band spurs.

I understand the DDS produces spurs but don’t understand how many low level spurs to expect.


The spurs that worry me most are spurs that appear within approx 20KHz of our output frequency, say 150 MHz +/- 20KHz.

On a significant number of channels we see these in band spurs on our output signal (150MHz) at a level of approx -70dBC.

Our main question is, should we be seeing these or not? The position of the spurs change in a complex manner as the DDS frequency is changed by a small amount.


I think that within our loop bandwidth (within 50KHz of the carrier centre frequency) the spurs from the DDS will actually be amplified by the overall frequency ratio, approx 13, this relates to an increase of 21 to 25dB. Am I correct here?

So the spurs at -70dBC at the 150MHz output are caused by spurs of approx -93dBc at the DDS frequency. Is this correct?


If all this is correct I have a few options. We are trying to remove all critical in band spurs to below approx -75dBC.

1)   1)  Understand how to calculate the position and level of each spur for each channel to ensure I have no spurs in critical areas

2)   2)  Accept I have used the wrong part which cannot guarantee the performance we need.

3)   3)  Make some unknown changes somewhere to improve things.

4)   4)  Use a completely different solution, (fractional N synth)


If the level of the spurs and amount of them doesn’t seem to relate to expected performance of the DDS I may need to make further investigation that the spurs are caused by something else.


What are your comments here. Do you think the spurs are likely to be from the DDS and can such low level spurs be realistically be pre-determined to optimise programming for each channel. Am I doomed to failure with this architecture?