[time-nuts] Re: DMTD Questions

[Jim Lux]

I think this stuff is fine for time-nuts - we've talked about it before. The mods will tell us otherwise if needed.

Generating precise frequency signals for comparison is kind of where we're at (I guess we could be "frequency-nuts" as well, although I confess we do seem to spend more time on Allan Deviation than on Phase Noise).
 

Yeah, I think the Keysight patent is probably a good description - on the other hand, patents often are a bit deliberately vague - in theory, the disclosure should enable someone "skilled in the art" to duplicate the invention. In practice, your patent attorney helps you not be too specific.   Enough specificity that the examiner believes you actually did it (practiced the invention), it's not technically required, but boy does it simplify the review process. Then the examiner focuses on prior art and the details of your claims.   And of course, you might say in your disclosure something like "a low noise operational amplifier, such as the AD1234 or LT9876" when in fact, the AD1234 won't work over all conditions, and you used another one, or the FET you describe only works if it's a special one. Or some aspect like how the power supplies are configured. 
(My patent on artificial tornadoes 5,971,765 is full of details, but not all details - commercial company; while my patent on radar heartbeat detection 10,201,278 is substantially more detail filled - US government funded)


On Tue, 14 Jul 2026 23:17:09 -0400, Alex Denner <alex.denner at gmail.com> wrote:

Hi Jim,

I don't want to get too off-topic on timenuts. I think this is
relevant because many of us use these generators for frequency
metrology purposes, however, I am happy to take the topic off-line if
there is further discussion.

Thank you very much for your comments and the references. I have read
a couple of them already. I think you are on the right track. Also,
there are a few patents which seem very consistent with Keysight's
descriptions of Trueform. Also they say it is a "patented"
technology--I understand that doesn't mean very much as it might be
that only one small part is patented but the patent below does outline
a process that could be relevant.

The second one which originated at a university in Florida is
reasonably well written and describes a whole process for waveform
processing which is pretty interesting.

Here is the abstract and the references for the patents. Do you think
these could be relevant to Trueform? I think Trueform was announced in
2012.

US Patent number: US7768355B2. Issued Aug. 3, 2010 (Agilent patent)

Abstract:

A polyphase numerically controlled oscillator is disclosed. An input
signal is received at a phase accumulator. The phase accumulator
provides a phase to a phase interpolator. The phase interpolator then
provides a plurality of output phases. The plurality of output phases
are provided to a plurality of phase to amplitude converters. Each of
said plurality of phase to amplitude converters process one of said
plurality of output phases.


US Patent number: 7,952,396 B1, Issued May 31, 2011. Provisional
applications date to 2008. (DME corp patent but licensed)

AWG having arbitrary factor interpolator and fixed frequency DAC sampling clock

Abstract

An AWG includes a waveform memory providing a digital waveform signal
at a sample rate and an arbitrary factor interpolator (AFI) coupled to
receive the digital waveform signal or a processed digital waveform
signal. A complex mixer for carrier modulation is coupled to the AFI
which outputs a complex band pass signal. A DAC is coupled to an
output of the complex mixer for receiving the complex band pass signal
to provide an analog output signal. A fixed frequency sample clock
clocks the DAC to provide a fixed DAC sample rate. The DAC provides a
data clock signal to a sample request controller that generates a
sample request signal that is coupled to the waveform memory for
requesting the digital waveform signal from the waveform memory. The
interpolated digital signal is sampled at the fixed DAC sample rate
independent of the sample rate of digital waveform signal.

On Mon, Jul 13, 2026 at 6:14 PM Jim Lux  wrote:
>
>
> 335xx and 336xx function generators are sort of a hybrid between DDS and Arbitrary waveform generator with, as you say, an interpolating filter. Keysight's documentation doesn't really tell you what's going on in actual mathematical terms. I figured out some special cases by actual testing with a 33622, and later it was confirmed by Keysight customer support as to what was going on (that was about the sinc interpolator).
>
> When it comes to DDS and spur reduction, there's a lot of techniques to get arbitrarily good performance. That is, it's not just using a longer phase accumulator, because you pick up other terms resulting from the other quantizations in the process.
> Analog Devices *does* talk a lot about how their (patented in some cases) schemes work.
> You might also look at the work from Vankka:Direct Digital Synthesizers: Theory, Design and Applications which covers a lot of the nifty stuff. There's papers out there by Vankka as well.
>
> For instance, you can basically run multiple NCOs in parallel, to generate an exact version of a spur generated by your primary DDS that can then be cancelled. Similarly, you can do things to push the spurs away from the center output (this is really handy if the output of the DDS is going into something like a narrow band PLL, so you can push the noise and spurs out of the PLL bandwidth.
>
> Another clever scheme is to change the length of the sin/cos lookup table so that it's an exact submultiple of the desired output frequency. This one is handy if you have a-priori knowledge of the frequencies you want to generate, AND you've got enough memory in your system to accommodate the various length tables (rather than all being 2^N long). Now that I think about it, there's a fundamental relationship between this kind of stuff and mixed radix FFTs.
>
> [1] Vankka J. Methods of mapping from phase to sine amplitude in direct digital synthesis. IEEE Trans Ultrason Ferroelectr Freq Control. 1997;44(2):526-34. doi: 10.1109/58.585137. PMID: 18244150.
>
> [2] J. Vankka, "A direct digital synthesizer with a tunable error feedback structure," in IEEE Transactions on Communications, vol. 45, no. 4, pp. 416-420, April 1997, doi: 10.1109/26.585916. keywords: {Clocks;Read only memory;Quantization;Frequency synthesizers;Registers;Filters;Costs;Sampling methods;Output feedback;Transfer functions},
>
> [3] J. Ketola, J. Vankka and K. Halonen, "Synchronization of fractional interval counter in non-integer ratio sample rate converters," Proceedings of the 2003 International Symposium on Circuits and Systems, 2003. ISCAS '03., Bangkok, Thailand, 2003, pp. II-II, doi: 10.1109/ISCAS.2003.1205898.
>
>
> On Sun, 12 Jul 2026 02:43:55 -0400, Alex Denner via time-nuts  wrote:
>
> Hi Chris,
>
> Wow, that is interesting---we did similar measurements. I have
> included my version of the same measurement. I just did the
> measurement yesterday and I was still thinking about the result as the
> additive phase noise is less than I would have guessed it would be.
> What do you think, is one of our versions approximately correct? My
> additive plot is somewhat similar to the plot that starts lower in
> your graphs (the one you identified as the 8663a.) First, as is
> typical with phase noise plots, one has to use a little bit of poetic
> license to compare them. Also, I am using a different reference which
> obviously makes a difference. (I can try doing the measurement with a
> 10811A if it helps you.) My reference used is an Endrun ultra-stable
> LPN oscillator disciplined by GPS. I've also included a phase noise
> plot of the reference itself. Interestingly, it is pretty much on
> spec. (The Allan deviation is excellent but it doesn't usually quite
> get to the specs listed.)
>
> As you know, the design of a DDS is such that it cannot output most
> frequencies EXACTLY...ie, because of the divisions and multiplications
> it does it can't get exactly to any given frequency. So, when using
> it to adjust between frequencies that are very close together, the
> traditional DDS design may not perform very well--frequency resolution
> is basically =(clock speed)/(phase accumulator size). As you probably
> also know, the Trueform in the 33522B has a modified design which I
> have not ever seen fully described but as I understand it, Trueform
> uses something like a 64 bit phase accumulator with an interpolation
> scheme that allows the waveform generator to vary to more closely
> match the desired frequency. Perhaps you or someone on this list can
> explain in detail how this works....I don't understand it.
>
> Another question: is it better to use a 33522B or 33622a for 10MHz
> offset purposes? The latter has a higher clock speed (1G vs 250 M)?
> and better jitter specs (1ps) but have been told that it might have
> less precise frequency resolution at 10MHz (meaning the residue after
> the least significant digit of microhertz).. Does anyone know the
> details about these generators?
>
> Of course, measuring the effects of tiny offsets are hard to assess
> and measure bc of the long time scales needed to resolve say 10
> microhertz.
>
> I have been trying out the Keysight 5047A phase noise analyzer....in
> my opinion it is a great instrument that is much more flexible than it
> seems. It can do a lot of different types of measurements. The noise
> floor isn't as good as some of the more expensive phase noise
> analyzers but it is quite flexible and has a number of functions that
> aren't advertised.
>
> With all that in mind, I am trying to see if there are phase noise
> differences from small changes in frequency right now. I am also
> testing out a bunch of other instruments including the SRS CG792 clock
> generator which uses a different frequency generation technique which
> is described clearly in its manual. I find it to be very good in
> early testing. (Note as a clock generator it outputs a square wave.)
> There are other approaches such as the SRS SG384 which uses RAFS to
> set frequency exactly to anything in its range within exactly 1
> microhertz. I don't have a Spectradynamics EOG (frequency resolution
> of 10^-20 over a small range) but would like to test one. ;-)
>
> Alex
>
>
>
>
> On Sat, Jul 11, 2026 at 10:33 AM Chris Hastreiter wrote:
> >
> > Alex,
> >
> > The "additive" measurement method you're describing sounds exactly like what I used to measure the residual noise of my synthesizers. In the case of my 8663A I used its internal reference but measured the phase noise of its output referenced to the "ref out" connector on the back. In the case of my 33250A I used an external 10811A as the reference. Here are my results, it looks like the 33522B is on a similar order of magnitude. (the 8663A is the lower noise trace on my plot)
> >
> > I've also included the absolute noise measurement plots for comparison.
> >
> > Chris
> >
> > On Thu, Jul 9, 2026 at 4:50 PM Alex Denner wrote:
> >>
> >> By coincidence, I happen to be looking at one small part of your
> >> project. I was curious about using a DDS to do frequency offsetting
> >> for another purpose. In these graphs, I used the Keysight 5047A (the
> >> AnaPico PN analyzer) cross-correlation phase noise analyzer. I looked
> >> at the phase noise of a free running Keysight 33522B vs the same unit
> >> with an external reference. Below are the two phase noise curves from
> >> 0.1 Hz to 5 MHz (10 MHz carrier). One can see the free running 33522B
> >> adds a decent amount of jitter if using it for a synthesized offset
> >> generator. I was estimating jitter by the difference in the two
> >> integrals. (I think the graphs have 100 correlations with some spur
> >> removal and averaging, I could try to find the details if anyone
> >> cares.).
> >>
> >> However, I thought of an interesting experiment: I could run an
> >> "additive" phase noise plot of the 33522B where one channel is the
> >> external reference and the second channel is the 33522B's (locked to
> >> external reference) output where the output is set to be in quadrature
> >> to the external reference entering the 5047A. It is not really an
> >> additive phase noise measurement in the classic sense but it may
> >> measure the total phase noise contribution of the DDS synthesis chain.
> >>
> >> As you know, the unreferenced HP 8663a with the OCXO is quite a bit
> >> better---probably some combination of the OCXO, better synthesis
> >> chain, and better power supplies, etc. I will also include a phase
> >> noise plot of my 8663a from 0.1 Hz to 5 Mhz at a 10 MHz carrier.
> >>
> >> Alex
> >>
> >>
> >> On Thu, Jul 9, 2026 at 4:42 AM Chris Hastreiter via time-nuts
> >> wrote:
> >> >
> >> > Time-nuts,
> >> >
> >> > I have been working on some modifications to Corby’s DMTD design [1] in an
> >> > effort to further improve its noise floor and had a couple questions for
> >> > the group. The performance I’ve been able to achieve so far is shown in
> >> > the below plot with the blue trace being the noise of the ProTIC counter
> >> > itself, the pink trace being the combined noise of the counter and my zero
> >> > crossing detector (both channels driven by a 10Hz sine wave), and the
> >> > remaining traces the full system noise floors with 4 different offset
> >> > oscillator options. All noise floor measurements were taken with a carrier
> >> > frequency of 10MHz and an offset of 10Hz. I believe that much of the noise
> >> > floor degradation at long tau is due to thermal effects in the first stage
> >> > amplifier of the ZCD as I haven’t yet gotten the system installed into an
> >> > enclosure and I can worsen the issue by thermally disturbing the LT1028 of
> >> > the first stage.
> >> >
> >> > [image: image.png]
> >> >
> >> > My first question is in relation to mixer noise which is a noise source
> >> > that I have not yet considered. I am using the same SBL-1 mixer as Corby’s
> >> > original design but [2] implies that some mixers exhibit lower flicker
> >> > phase modulation than others and I haven’t seen any literature or
> >> > specifications that point to any devices other than the 10514A in that
> >> > paper. They also discuss adding capacitive loading to the mixer to further
> >> > reduce noise referencing another paper that I have yet to be able to find.
> >> > Is there any expectation that changing to a mixer different than the SBL-1
> >> > will improve my results and should I consider capacitive loading on the
> >> > mixer? If so, how would I determine an appropriate amount of capacitive
> >> > loading? I currently have it loaded with a bridged-T filter holding 50
> >> > Ohms of impedance across the frequency spectrum.
> >> >
> >> >
> >> >
> >> > My second question is in regard to the common oscillator noise
> >> > contribution, specifically for the use case of a synthesized offset
> >> > oscillator. The below plot shows the results I’ve gotten so far with some
> >> > sources that I already have, and as can be seen the system noise floor
> >> > degrades much faster with the synthesized offset than with the 10811A.
> >> > However, it would be advantageous for me to be able to use a synthesized
> >> > offset for tests where the DUTs differ slightly in frequency and the system
> >> > will phase wrap. The reason for this is that when using a free-running
> >> > oscillator the phase wraps are very difficult to remove automatically
> >> > because the divided phase wrap will differ from the expected with any
> >> > frequency error in the offset oscillator. Do you have any suggestions for
> >> > synthesizers or other offset techniques that would help with this? Using a
> >> > 10811A and a slow PLL as a cleanup loop on the synthesizer does appear to
> >> > work well in the below plot, but looking at the first plot in this email at
> >> > least in my initial implementation there seems to be some low frequency
> >> > oscillation in the PLL impacting the noise floor. I’ve also included a
> >> > plot showing measurements I’ve taken of the residual noise for the two
> >> > synthesizers used as well as a comparison of two of the 10811A oscillators
> >> > that I’m using.
> >> >
> >> > [image: image.png]
> >> >
> >> > [image: image.png]
> >> >
> >> > [1] https://www.eevblog.com/forum/metrology/dmtd-board/
> >> >
> >> > [2] https://ieeexplore.ieee.org/document/5075023
> >> >
> >> >
> >> >
> >> > Thanks,
> >> >
> >> >
> >> >
> >> > Chris
> >> > _______________________________________________
> >> > time-nuts mailing list -- time-nuts at lists.febo.com
> >> > To unsubscribe send an email to time-nuts-leave at lists.febo.com
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com
Â