FAQ: Why would I choose to use DDS for synthesizing a signal rather than a PLL?

Q.

Why would I choose to use DDS for synthesizing a signal rather than a PLL?

-----------------------------------------------------------------------------------------------------------------------------

A.

Let me start with a basic comparison between the two, but keep in mind this can change over time.

In general, it is fair to say that DDS technology provides some specific advantages:  dynamic frequency and phase control is relatively easy, and is not prone to ringing or settling, so if you need to do any sweeping or hopping, DDS is a good option to start.  DDS also provides much finer tuning resolution than even a fractional N PLL can provide.  The digital nature of the DDS also makes digital synthesis more repeatable.

Conversely PLL technology offers other benefits:  Generally speaking it is less expensive and less power hungry than DDS technology.  Perhaps most importantly, you have the ability to 'upconvert' your input frequency, whereas with the DDS is a Nyquist system, and limited to providing outputs that are <50% of the sampling clock frequency that is used.

But if you look beyond the surface, you 'll find a whole lot of softening of these boundaries.  For example:  There exist both 'ping pong' and 'fast hopping' PLLs on the market (including products from ADI) which close the gap in terms of their ability to provide frequency and even phase agile outputs.  Fractional N PLLs help close the gap on timing resolution for that matter.  Conversely, ADI has also closed the gap for DDS products in some cases.  Specifically worth noting is the AD9913, which significantly closes the gap both in power and in cost to PLLs.  ADI's DDS portfolio includes many products with reference clock PLLs on board, allowing customers to upconvert their reference clock before dividing down through the DDS, meaning the output can now be a higher frequency than the reference - note is is also possible to bandpass filter a DDS output isolating a Super Nyquist image to obtain an output frequency that is higher than the reference frequency source provided.

And taking it a step further, where fractional N PLLs closed the gap in terms of frequency granularity, the Programmable Modulus mode implemented in the AD9913 (and eventually most future DDS devices), widens that gap back up.

Comments
Anonymous
    •  Analog Employees 
    in reply to Eleuthere

    Adding to Ken's note, specifically, with the AD9913, you could start from a 10 MHz reference source and get exact 1 Hz steps all the way from 0 Hz to Nyquist.

    •  Analog Employees 
    in reply to Eleuthere

    Not all commercial DDSs are strictly binary in terms of frequency synthesis. Take the AD9913, for example, which offers programmable modulus operation. This allows you to synthesize output frequencies with a non-binary modulus (exactly 1/10th of the system clock, for example).

  • Commercial DDS have a drawback that should be noted: they are definitly binary. It is impossible to get an exact 1 Hz step with an exact 10 MHz reference, for example ...

    It is possible to build a decimal DDS with somme programmable logic, a sine PROM (with 1000 or 2000 or 2500 or 5000 addresses and not 1024, 2048...) and a DAC ... I did it (not with a 500 MHz clock).

    Nevertheless, they are cheap and powerfull components !

Related