kb8tq@n1k.org said:

I don't know how to think about a DDS in this context.

I remember years ago thinking that a DDS was the greatest thing since sliced
bread.  In the context of something like a PPS going into a PC for
timekeeping, that's probably true.  You get long term accuracy and the PC
can't see the short term issues.

But then, somebody mentioned close-in spurs.  They get closer the more bits
you have in the DDS magic number.  (What is that number called?)

Suppose I have a black box labeled "10 MHz" with a cable coming out.

If you plug that cable into your ADEV measuring setup, can you tell if my box
has a DDS in it?

If you plug that cable into your spectrum analyzer, how good a setup do you
need in order to see the spurs?  Do they get lost in the close in noise?  Or
maybe the question should be how clean a signal do I need to start with before
the spurs become visible?  Or what should I be asking?

--
These are my opinions.  I hate spam.

Hi

Well….. the existing synth in the 5065 is no champion spur wise. Coming up with a
DDS that is “as good as” is not all that crazy.

The 5 MHz reference is multiplied to 6.x GHz and then mixed with the synth. The close in
phase noise of the multiplied signal is mixed with the synth. The output is the sum of the two.
Going from 5 MHz to 6.x GHz gets you 20 log (N)  … a bit over 60 db added phase noise.

The synth (at 100 ti 300 Hz offset) needs to be 60 db worse than the noise on the 5 MHz
before it even gets in the game. That’s not a really tough spec to meet.

Bob

Hi,

The existing syntesizer is rich on spurs. Very rich in fact. Fairly strong.

However these spurs is far enough not to cause much of a problem for the
clock, but they will help to demodulate some noise.

The noise being problematic is too far in for the spurs of the synthesis
to generate any real danger.

I replaced the 5.3 MHz syntesis with that from my HP3325B, which for
sure isn't free from spurs either, but it turns out that the far in
noise was filtered by the lock-up, so I ended improving the phase-noise
for the 6.8 GHz synthesis such that the stability out at 10000 s was
improving by a factor of 2.

So, don't over-do it with hunting spurs that ain't going to contribute
significantly to the noise and systematics of the box. Increasing
resolution is good. Enough bits to the DAC to keep the additional spurs
from non-linear mixing down, because that is what truncation of the
phase-accumulator state will do and mix the spurs from the current
phase-accumulator setting. Step the frequency setting and the spurs move
around and the mix products will also move around as with any mixer. As
long as that is linear enough, additional spurs will be sufficiently low
that we have other things dominating.

Cheers,
Magnus

On 2020-10-08 01:33, Bob kb8tq wrote:

Hi

You could also say that with a longer accumulator, you get a closer approximation to a
desired waveform.

If you compare the data sheets, the spur levels on the 48 bit parts

ad9854.html https://www.analog.com/en/products/ad9854.html
ad9912.html https://www.analog.com/en/products/ad9912.html
https://www.analog.com/en/products/ad9956.html https://www.analog.com/en/products/ad9956.html

are more related to the age of the part than the width when compared to the smaller parts

https://www.analog.com/en/products/ad5930.html https://www.analog.com/en/products/ad5930.html
https://www.analog.com/en/products/ad9834.html https://www.analog.com/en/products/ad9834.html

Indeed, if you go through the entire page if parts

https://www.analog.com/en/parametricsearch/11018#/ https://www.analog.com/en/parametricsearch/11018#/

There does not appear to be a spur penalty for wider parts. Newer is indeed better, pretty
much across the board.

Bob

This I suspect is not original, but with respect for the need for 48 bit
resolution, why not, say, a 24 bit, with reduced range such that it's steps
are the same as what  you would see with 48 bits, this stacked (Biased) on a
fixed stable voltage in middle of expected control range. Fixed voltage
could be sourced from an initial courser wide range device.

Since this seems so obvious, who did it first?

Lester B Veenstra  K1YCM  MØYCM  W8YCM  6Y6Y
lester@veenstras.com

452 Stable Ln (HC84 RFD USPS Mail)
Keyser WV 26726

GPS: 39.336826 N  78.982287 W (Google)
GPS: 39.33682 N  78.9823741 W (GPSDO)

Telephones:
Home:                     +1-304-289-6057
US cell                    +1-304-790-9192
Jamaica cell:           +1-876-456-8898
 

-----Original Message-----
From: time-nuts [mailto:time-nuts-bounces@lists.febo.com] On Behalf Of Hal
Murray
Sent: Wednesday, October 07, 2020 6:18 PM
To: Discussion of precise time and frequency measurement
Cc: hmurray@megapathdsl.net
Subject: Re: [time-nuts] Aging 5065A ?

kb8tq@n1k.org said:

bits

I don't know how to think about a DDS in this context.

I remember years ago thinking that a DDS was the greatest thing since sliced

bread.  In the context of something like a PPS going into a PC for
timekeeping, that's probably true.  You get long term accuracy and the PC
can't see the short term issues.

But then, somebody mentioned close-in spurs.  They get closer the more bits
you have in the DDS magic number.  (What is that number called?)

Suppose I have a black box labeled "10 MHz" with a cable coming out.

If you plug that cable into your ADEV measuring setup, can you tell if my
box
has a DDS in it?

If you plug that cable into your spectrum analyzer, how good a setup do you
need in order to see the spurs?  Do they get lost in the close in noise?  Or

maybe the question should be how clean a signal do I need to start with
before
the spurs become visible?  Or what should I be asking?

--
These are my opinions.  I hate spam.

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.

Hi

It would be far easier to just use a (stock) 48 bit chip than to try to
get something like that up and running ….. Keeping all the transitions
“just right” would be tough. If you don’t do that, you have a major added
source of spurs.

Bob

Bob wrote:

The width of the accumulator does improve the waveform, but only if you need a frequency between the
ones possible with the more narrow accumulator.

It goes without saying that the extra bits only have an effect if they are non-zero.

Mike write:

Adding non-linear bits to the DAC increases noise significantly, because it introduces a periodic effect.

This stuff is all easy to simulate with a computer.

Try simulating a 12 bit Dac and a 14 bit DAC where the LSB is only half the magnitude it should be.

If I were trying to build a high-res DDS, I would calibrate the sine-table to the specific DACs imperfections.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.