AM-noise versus PM-noise One of the big questions remaining now is why ADI cannot reproduce the noise plateau that I can measure so easily with two different but simple and reliable methods in my home-lab! There is a remarkable resemblance with this graph and the observed AD noise plateau! The notch filter method shows rather similar, actually slightly worse results, in comparison with the reciprocal mixing method. In fact the single sideband noise of the AD is about 20dB stronger than we could theoretically hope for given the residual measurements at Therefore I was not too hopeful that the 4-layer evaluation board would behave much better than my 2-layer design. ADI had been somewhat reluctant so far to accept my measurements, which is understandable given the reputation of the signal source analyzer from Agilent they use.
|Date Added:||17 September 2010|
|File Size:||48.49 Mb|
|Operating Systems:||Windows NT/2000/XP/2003/2003/7/8/10 MacOS 10/X|
|Price:||Free* [*Free Regsitration Required]|
The idea is that the noise of the analog power supplies, especially the 3.
At that point I decided to homebrew a simple PM-noise test-set in order to validate the results with yet a third and completely different measurement method: A low noise, wideband opamp is used as the error amplifier with unity closed-loop gain. ADI had been somewhat reluctant so far to accept my measurements, which is understandable given the reputation of the signal source analyzer from Agilent they use.
Therefore I put especially the notch method through quite a lot of scrutiny, but nothing wrong could be found. My AD prototype board is a simple double sided board. Therefore, the reference clock is a possible although unlikely suspect of introducing the excess noise measured with the AD prototype board. Add to a parts list. In order to eliminate the reference clock as the source of the problem the following simple tests have been done: Save to an existing parts list Save to a new parts list.
Suspicion existed regarding the actual DDS chip revision being used on my prototype board, because it is a pre-production chip obtained through the early sample program before the AD was officially released. AD chip revision Suspicion existed regarding the actual DDS chip revision being used on my prototype board, because it is a pre-production chip obtained through the early sample program before the AD was officially released.
Therefore I was not too hopeful that the 4-layer evaluation board would behave much better than my 2-layer design. But still the evidence is quite convincing that a big capacitor is really needed on a “DACBP” pin to take out the noise plateau caused by the internal band-gap reference of the DAC.
It is unclear at this moment why the squarer on the frontend board does not seem to evalation suppress the AD’s excess AM-noise. It also shows that the reciprocal mixing method is working quite well, because the notch filter method confirms its results.
The answer to the question whether it is an insignificant design flaw or a pretty serious one that the AD produces orders of magnitude more AM-noise than its documented PM-noise, depends on the application where the chip is used for.
Analog Devices Direct Digital Synthesiser (DDS) Evaluation Board for AD9910
Furthermore the really low spur levels gave confidence in the double sided PCB layout and the other design choices made with this board. The reason why ADI did this “optimization” is unclear. Given the lower working frequency of This is new information.
The reciprocal mixing method and the notch filter method however measure any sideband noise; AM-noise and PM-noise. The reciprocal mixing method showed the noise plateau from the start.
Therefore an improvised way of measuring phase noise was needed. The noise level of such a voltage regulators is very low, roughly identical to the input noise voltage of the opamp, which in this case is only 4. They do not have a DACBP pin either and therefore the expectation that those chips will be better regarding the AM-noise issue is rather low. That is dB relative to the 1. The side of the chip that is related to most of the analog functionality has actually no less than evakuation unused pins available in the current chip revision.
Notch Filter Method results Another elegant and simple method to measure sideband noise, including phase noise, is the Crystal Notch Filter Method.
My homebrew regulators are using a band-gap voltage reference that is de-noised with an RC filter cutting off at 0. Furthermore, the Agilent SSE also uses the mixing method. If the internal band-gap device is not thoroughly filtered, then the noise will end up directly in the output signal of the DAC. Second, the cutoff frequency offset changes by evaluatiion much as an octave from KHz to KHz, which is very significant! That is still not good enough, but it is the best we have seen so far!
Thank you for your feedback.