[hpsdr] new DDS

[Hans Summers]

> > Some serious limitations on the technology;
> >
> > 1. The output frequency must not be any lower than 20% of the
> > reference oscillator.
>
> They claim that the spurs are negligible for such output frequencies.

Leon, I don't think that is correct. From my reading, they state that
the spur killer has negligble benefits when operated at below 20% ref
clk, where the spurs are already very low (not necessarily
negligible).

I am a novice on this stuff. I did build an AD9851 DDS (see
http://www.hanssummers.com/radio/dds ) using your board Leon,
remember?

Anyway, from my reading of DDS datasheets (quite a bit over recent
weeks), it seems that the reason Analog are still developing their
newest devices with 10 bit DAC's rather than continuing on from the
14-bit AD995x series might be because the wider DAC contributes to the
residual phase noise. If you look at the 10-bit parts their phase
noise specification is better than the 14-bit parts, but the spurs are
worse. So it seems to be a trade-off between low phase noise and low
spurs (high SFDR).

By phase noise, my understanding is that they mean "residual" phase
noise, they are specifying the phase noise added by the DDS. The final
result would be the combination of this DDS contributed phase noise,
plus the phase noise of the crystal reference oscillator. So to take
advantage of a part with lower phase noise does require a high
performance reference oscillator.

About the spur killer: it seems that it probably has a very limited
range of applications. One of those things which sound good, but are
less useful in practice. Read: "Marketing gimmick"? Notice from the
spectrum plots in the datasheet that when the spurkiller is enabled,
the three worst spurs can be attenuated, but the remaining noise floor
is INCREASED? So the SFDR specification of the device is improved by
the spurkiller, but the actual average level of spurs is worsened.

I didn't understand whether the spurkiller automatically removed the
three worst, or whether the chip must be told which spurs to kill. But
it seems to me that in many applications the narrow band spurs will be
more of an issue since they mix with incoming receiver signals which
go right through the input bandpass filter and are close to the wanted
signal. Often spurs further away might be less of a problem.

It seems to me that the spurkiller might be useful in some limited
circumstances but even so requires careful babysitting by whatever is
controllign the DDS.

So it seems that the design choice is still the tradeoff between phase
noise and SFDR, like it was before the AD9911. I think the main
benefit of the AD9911 would be if you wanted the other features it
offers, e.g. multitone and so on.

But as I said, I am a DDS novice so feel free to ridicule and/or
ignore the above!

73 Hans G0UPL
http://www.hanssummers.com

 1148999725.0