ADC16V160, ENOB over VPP

Florian,

What part are you referring to - ADC16DV160 or ADC16V130? We do not have a ADC16V160.

Thanks, Josh

Hi Josh,

thank you for your answer.

Yes, I missed typing the D. It is the ADC16DV160, the one that is mentioned in the tag list, clocked with 160MHz (mentioned in the question) and shows 12.3 ENOB on TI's webpage (easy to check).

The key questions are for both parts the same ;-)

1) Does the ENOB depend on VPP ?
2) Can the ENOB on the webpage or in the datasheet be (roughly) verified with the eval board and WaveVision5's on-the-fly ENOB calculation.
3) Why is the ENOB with VPP = 1.5V below 8 for the aformentioned part.

But I assume it all boils down to the basic question:

4) Is the ENOB (or the SINAD) measured by using a constant input signal (I guess not, otherwise Fig 10 in the datasheet makes no sense) or by using a sinus signal (then, don't you have to specify the amplitude of your sinus) ?

Thanks a lot for your answer again,

Best, Florian

Florian,

ENOB is a mathematical conversion from the SINAD metric. ENOB = (SINAD-1.76)/6.02. ENOB can be calculated with any combination of input frequencies and input amplitudes but it is most commonly reported on datasheets for a -1dBFS input signal. In the case of the ADC16DV160, neither the SINAD nor the ENOB is reported on the datasheet. It is hard to say exactly where the 12.3 ENOB number came from on the webpage because conditions are not included but it appears to correspond to the SNR on the datasheet reported as 76dBFS for a -1dBFS input signal at 197MHz. This is not technically correct because we are looking at an SNR number.

Yes, the WaveVision 5 software will display an accurate ENOB value for the corresponding SINAD under the conditions that you are measuring.

Now, the fact the you are measuring a very low ENOB can be due to a number of reasons. Here are some that you can investigate:

- The calculation from SINAD is not correct

- There is noise or distortion coming into the converter through the analog input. This may be due to a poor noise quality signal generator or it may be because that signal source is not filtered

- There is noise coming into the converter through the clock input. Again, This may be due to a poor noise quality signal generator or it may be because that signal source is not filtered. Phase noise of the clock source is of particular concern here.

- The input frequency is VERY high, resulting in high distortion and noise due to performance limitations of the converter. The -1dBFS 197MHz input conditions should definitely result in ~12 LSB ENOB is good signal sources are used for the input and for the clock.

Regards, Josh

Hi Josh,

thank you for your answer.

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

Josh: “ENOB is a mathematical conversion from the SINAD metric. ENOB = (SINAD-1.76)/6.02.“

Well, you are referring to a simplification. The IEEE spec says [1] (equation 71), that ENOB depends on the VPP of the input signal. The ENOB reduction (based on the IEEE spec) is very much in-sync with the numbers I can measure. Technical papers from Rohde und Schwarz [2] as well as from Analog devices [3] say therefore, quote:

“The measure of ENOB provides a basis for comparing different systems but is only a
useful reference when quoted together with the input amplitude at which the measurement was taken.”

[1] IEEE Standard for Terminology and Test Methods for Analog-to-Digital Converters, IEEE Standard 1241-2010
[2] cdn.rohde-schwarz.com/.../ENOB_Technical_Paper_1ER03_1e.pdf
[3] www.analog.com/.../MT-003.pdf

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

Josh: “In the case of the ADC16DV160, neither the SINAD nor the ENOB is reported on the datasheet. “

Please have a look at Figure 10, which shows SINADs over fin. Calculating the ENOB is then a no-brainer.

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

We are still struggling to reach more than 8 ENOBs with VPP=1.5V @40MHz. The 160MHz oscillator is directly soldered on the eval-board, and the input sinus is coming from a decent lab-function generator with good shielded cabling. I don’t think that any other test setup would dramatically improve the ENOBs to 12.
Any other idea how to verify 12+ ENOBs (@ 1.5V VPP) on the TI eval board ?

Best, Florian

Florian,

The effect of the -log2(A/FSR) term in the equation from reference [1] is to change the reference of the SINAD number. In the case of the ref [1] equation, SINAD is the noise and distortion relative to the signal amplitude, A. The -log2(A/FSR) term changes the reference so that SINAD becomes the noise and distortion relative to the full-scale range, FSR. I didn't clearly state this, but my equation assumes that SINAD is in units of dBFS (dB relative to full scale), so the changing of the reference to the full scale range is already assumed.

It is true ENOB is a reasonible basis for comparing different systems and that the input amplitude and frequency must be consistent to make a fair comparison.

Are you filtering your signal generator that provides the input signal? Have you measured the signal source to verify that the noise is <-160 dBm/Hz so that it does not impact the measurement? Have you verified that the distortion is < -90dBc?

It appears that you are not filtering the clock source, so all the noise from the 160MHz oscillator is impacting your measurement. To measure the full performance of the ADC, the input signal source noise and the clock noise must be reduced as much as possible with filtering.

If you post a screen shot of the spectrum and FFT results values from WaveVision 5, I may be able to comment further.

Regards, Josh