I am looking for an ADC for my project and there are a few features that I don't quite get. On TIs product selector. The default filters includes INL, SNR, SFDR. I assume they are important since but some of them have value some dont, so I assume its ADC-spefic (or maybe they did not just write it) not all ADC have them, being just initials makes it harder to search.

Here are others on which I need some advice:

Programable Gain Amplifier(PGA) - Assuming same sample speed, Which would be better a 16-bit ADC using internal gain or a 24-bit (or 32-bit) ADC that is using no gain. Gain also amplifies the noise but assuming the isolation is perfect, wouldn't the 16-bit with a gain of 2 or more be better to the 32-bit using no gain? Of course there is gain drift but at low gains it's not much.

Simultaneous or Multiplexed Sampling - Although this one is self explanatory, Isn't simultaneous sampling (for multiple channels) always better? Are there other notable reason why to choose multiplexed over simultaneous?

Here are 2 ADC that have very peculiar features about which I have no idea:

Those two interest me because what I plan to use my circuit is to read voltages from a current transducer so that I can get the current consumption of a mains line.

Also what are all these gold for? are there really applications for it? gold, more gold

  • \$\begingroup\$ Simultaneous sampling is always better, sure. It just costs n times as much, takes up n times the space, requires n times the wiring, and n times the power...which is a lot. For your PGA question, it all comes down to how much noise is in introduced by your amplifier and what the effective number of bits really is for your ADC after you account for the INL. \$\endgroup\$
    – DKNguyen
    May 12, 2020 at 18:09
  • \$\begingroup\$ @DKNguyen so true, if it was cheap to implement all multi channel adc would have them \$\endgroup\$
    – Jake quin
    May 12, 2020 at 18:11
  • 2
    \$\begingroup\$ Search up general articles on ADC accuracy. It's a bit of a rabbit hole. INL stands for integral nonlinearity. ENOB is effective number of bits. LSB is least significant bits which means the same thing as counts. DNL is differential nonlinearity. \$\endgroup\$
    – DKNguyen
    May 12, 2020 at 18:14
  • 2
    \$\begingroup\$ SNR - Signal to Noise Ratio, SFDR - Spurious Free Dynamic Range. \$\endgroup\$
    – jpou
    May 12, 2020 at 19:22
  • 2
    \$\begingroup\$ Glossary of Frequently Used High-Speed Data Converter Terms (Maxim Integrated tutorial/appnote 740): pdfserv.maximintegrated.com/en/an/AN740.pdf ... and more: maximintegrated.com/en/design/technical-documents/index.html/… \$\endgroup\$
    – MarkU
    May 12, 2020 at 23:02

2 Answers 2


Here is everything that i have gathered.

INL - Integral Non-Linearity

Basically, this is error of the reading of the ADC (and DAC) to the actual values. The value is calculated by:

 INL = Vreal - Vmeasured

Generally the lower this value is, the better. Higher pecision bit ADC suffers from high linearity reducing their Effective Number of Bits (ENOB), consult datasheet to know more about the specific adc. More info on INL here enter image description here

SNR - Signal to Noise Ratio

It is the ratio of the signal to the noise.

enter image description here

The formula is signal divided by noise. From that formula we would generally want a high signal to noise ratio so that more of the signal can be seen and less of the noise. I do not know why they put this as a filter as this is an application specific measurement, different noise will hit your circuit thus the SNR will vary from application.

SFDR - Spurious Free Dynamic Range

Is the ratio of the fundamental signal to the strongest spurious signal in the output. radio receiver application, Im having trouble putting it into words but you can read more here

enter image description here

From what i understood you generally want a higher ratio, but again this is an application specific measurement as it uses noise in its equation, and noise is something that is application specific.


Try as much as possible to use the least/no gain as a gain amplifier instroduces aliasing and you would not want that, the higher your gain the higher the alias too.

Simultaneous or Multiplexed Sampling

As DKNguyen said multiplexed is best case but will slightly cost you, get it if your constraints allow you.

Notched Filter

A very narrow band stop. And a band stop at 50HZ/60Hz which is also our target working frequency, really not ideal for the application i want to use it with.

AC excitation

the system which is used for providing the necessary field current. Im not really sure how can this be used alongside an ADC. You ca read more about AC excitation here


Please refere to @Sephro Pefphany's answer


I don't think you want a mains frequency notch filter if you're trying to read mains frequency signals.

In answer to your final questions, those packages appear to be special high temperature packages rated for up to 210°C operation. This kind of thing is required for down-hole applications. That also accounts for the hefty (~$600 USD) pricetag for a part that's normally < $50.

They also specify the minimum lifespan as better than 1000 hours only (death due to electromigration).

enter image description here

  • \$\begingroup\$ Do ADC die that quickly? 1000 hours is about 41 days, isnt that a bit short for electronics ? Im planning to use mine for atleast a year of non stop use. Is short life span something that i just have to deal with ? \$\endgroup\$
    – Jake quin
    May 12, 2020 at 21:22
  • 2
    \$\begingroup\$ @Jakequin Keep it cooler than +210°C (410°F) and it will last longer. If you can't keep it cool (like the down-hole boys and girls) then you might have to live with it. \$\endgroup\$ May 12, 2020 at 21:48
  • \$\begingroup\$ Ohhhh 1000 hours at +210°C. That temp is enough to melt generic solders. I wont be buying that anyway they are priced ridiculously for my application \$\endgroup\$
    – Jake quin
    May 12, 2020 at 21:52
  • 1
    \$\begingroup\$ @Jakequin There is a lesson here though, for those who like to do naughty things with outputs like drive excessive current through them on a continuous basis. High current densities in the on-chip interconnects kill chips, and high temperatures exacerbate the effect. At 125°C current must be reduced to 1/10 for the same reliability at 25°C for the same conductor size. \$\endgroup\$ May 12, 2020 at 22:02
  • \$\begingroup\$ I do hope i will be able to get to design a circuit that can justify those use cases. Thank you for that trivia \$\endgroup\$
    – Jake quin
    May 12, 2020 at 22:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.