Recently I wrote an article which appeared in the Volume 47 - February 2013 edition of Analog Dialogue. This article outlines a method of estimating the noise introduced into an ADC signal chain by an amplifier or driver. Amplifiers or drivers can be used to achieve one or a combination of the following objectives:

• Provide gain to the input signal to increase the ADC resolution.
• Buffer or transform the impedance between the input source and the ADC.
• Convert a single-ended input signal to a differential output signal.

In doing so, the amplifier introduces noise into the ADC inputs thereby affecting its performance (SNR). In this article, I explain how to interpret the datasheet of an amplifier and that of an ADC and using some basic math, arrive at an estimated SNR when a particular amplifier-ADC combination is chosen.

The ADC chosen for this experiment is the AD9268-125 and the Variable Gain Amplifier (VGA) is the AD8352. Using the datasheet numbers for the Noise Spectral Density (NSD) of the amplifier (20nV/√Hz into a 150Ω load), I can prove that the performance of the system that uses the AD8352 to drive the AD9268 will be dominated by the amplifier noise. This means that using the AD8352 to drive the AD9268 will result in an ENOB of about 11.8bits.

This same technique can be used to estimate the performance of any amplifier – ADC combination thereby helping the customer make a more informed decision on the amplifier selection for a particular ADC.

The article can be accessed here:

Understanding How Amplifier Noise Contributes to Total Noise in ADC Signal Chains

Cheers,

Umesh

• Hi Vishnu,

I am sorry I didnt see this comment until now. The NSD for the amplifier is dependent on the design. The AD8375's architecture lends itself to a constant NSD of about 20nV/rtHz across frequencies. I am not sure that you can compare the NF to NSD relationship between a log amp and the AD8375 because internally their designs are separate. ADI's DVGA families usually have a constant NSD because of their architecture. This makes it easy on systems designs to evaluate the noise performance of the amplifier in the user's band of interest.

Hope this helps.

Thanks

Umesh

• hi Umesh,

The article is very informative.

I have a doubt in the calculation of NSD for the VGA. You have taken the NSD as 20nV/sqrtHz (In AD8375 datasheet it is mentioned typically this value). Is there any equation to derive this value. In the AN-Noise Figure and Logarithmic Amplifiers (AD Volume 42 – June 2008), there is table which gives VNSD against NF. There 8.3dB NF corresponds to 1nV/sqrtHz. Here AD8375 typical NF is 8dB, so the VNSD must be somewhere close to 1nV/sqrtHz. That is my confusion.

Regards,

Vishnu

• Hi Jun Zhou,

Thank you so much for your feedback.  I will be sure to update the equation and the calculated value.

Regards,

Umesh

• Hi Umesh,

First of all, very useful article helped me to understand noise calculation. Thanks.

You wrote the correct equation (6), but when used it later with the shape factor of 1.02, the shape factor was not inside the square root. If you put 1.02 inside the square root, you can get the calculated amplifier noise about 148uVrms, instead of 149uVrms.  And 148 is more close to the measured 140 than 149.

Regards,

Jun Zhou

• Got it. Thanks.