19
\$\begingroup\$

How does a differential analog to digital converter differ from a regular ADC?

\$\endgroup\$
16
\$\begingroup\$

A differential ADC will measure the voltage difference between two pins (the plus and minus input). A single-ended ("regular") ADC will measure the voltage difference between one pin and ground.

A lot of differential ADCs can be configured to give you twice the channels in single-ended mode. For example the AD7265 has 6 differential channels and 12 single ended channels.

\$\endgroup\$
  • \$\begingroup\$ I probably should have said half the number of channels when in differential mode :) I added the AD7265 example to my original answer. Also there are various MUX + ADC systems (AD363+AD364) that have 16 single ended inputs or 8 differential inputs. \$\endgroup\$ – jluciani Sep 22 '11 at 0:06
  • 1
    \$\begingroup\$ That's weird to me. Why would you ever use two ADCs to do the job of one? The differential aspect could just be handled by a much cheaper op-amp. \$\endgroup\$ – endolith Sep 22 '11 at 0:23
  • \$\begingroup\$ Dual-input ADCs are handy for those times you need to sample two signals at exactly the same time while keeping parts count, size, and power low. An example application is stereo audio. \$\endgroup\$ – Mike DeSimone Sep 22 '11 at 4:54
  • 2
    \$\begingroup\$ As for converting differential input to single ended before it hits the ADC: a differential input ADC can do a lot better job of common-mode rejection than the external-amplifier version, because it does not have a single-ended stage to pick up noise with, and can be calibrated to minimize offset at the factory. Finally, a differential input can give you 6 dB of gain over a single-ended input, since each leg of the pair can go through the full input voltage range (e.g. with a 5 V single rail, you can sample from -5 V to 5 V, not just 0 to 5 V). \$\endgroup\$ – Mike DeSimone Sep 22 '11 at 4:59
  • 1
    \$\begingroup\$ This is analogous to your dual trace oscilloscope which gives you two single-ended traces (using the ground clip of each probe) or a single differential trace with both probes (X - Y mode, grounds ignored). \$\endgroup\$ – Kaz Nov 23 '12 at 22:44
7
\$\begingroup\$

A regular ADC samples it's inputs in the range 0V to AVcc, where AVcc is often configurable (5V, 2.56V, user input etc).

A differential ADC shifts the lower reference from 0V to some other value - either a user input on a second analog input, or a internal reference. This is helpful for measuring small signals that have a large DC offset - eg measuring changes of 100mV in the range 2.5-2.6V.

Readings for voltages lower than the offset are hardware dependent - can give negative readings, absolute values, or zero.

A typical application is in a load cell which has a small voltage change at some DC offset.

\$\endgroup\$
5
\$\begingroup\$

As others said, it has two inputs for each signal, one of which is subtracted from the other.

This gives you more signal-to-noise ratio because

  • The maximum input level is 6 dB higher
  • Uncorrelated noise of the two inputs combines to be only 3 dB higher
  • It cancels out any common-mode noise. (If the ADC's ground voltage is fluctuating relative to the ground of the thing being measured, for instance, both inputs will move up and down together, and this will be cancelled out. If the two inputs are both driven from the same op-amp and some of its power supply noise is getting into both outputs, that will be canceled out. etc.)
\$\endgroup\$
3
\$\begingroup\$

It's hard to say exactly what you're talking about without a reference, but I'm guessing that you're talking about an ADC that has a differential pair input.

Differential pairs are nifty things that allow you to double the perceived voltage swing without raising the supply and inducing additional noise. Essentially what goes on is that instead of having a signal referenced to ground, the two wires are total opposites; when one line is at +1.3V, the other is at -1.3V. The voltage of either line to ground is only 1.3V, but since the ADC is converting the difference of the voltage on these signals, you have 2.6V.

I assume you're talking about ADCs which sample differential signals.

Differential pairs are used where ever you want to limit induced voltages. Ethernet and USB are both differentially signaled. Lots of RF is differentially signaled. If you do some hunting on Google you'll find LOTS of more information.

\$\endgroup\$
  • \$\begingroup\$ Differential signalling reduces the amount of interference emitted into other things. Balanced lines with differential inputs reduce the amount of interference picked up from other things. They're usually combined, but they don't necessarily have to be. \$\endgroup\$ – endolith Jul 27 '10 at 1:18
  • 1
    \$\begingroup\$ Well if you are on the receiving end of a differential pair, you are doing it for the noise immunity. If you're on the transmitting end, you are usually doing it for the emissions and/or noise immunity. I guess I did not make it clear enough. \$\endgroup\$ – akohlsmith Jul 27 '10 at 13:57
3
\$\begingroup\$

Another point not yet mentioned is that a typical ADC which is designed to resolve 0-3 volt signals to one millivolt precision (12 bits) might not have much better than one-millivolt precision when trying to resolve a 0.1 volt differential signal riding on a two-volt common-mode signal (e.g. it might have 8 bits of useful precision), whereas an ADC which is designed to resolve small differential signals would be able to perform much better; a 12-bit ADC could be designed for such purposes to provide 12 bits of useful precision with a 0.1-volt signal without having to be designed to provide 16 bits of precision on a larger signal).

\$\endgroup\$
1
\$\begingroup\$

A Differential ADC is a two terminal device. In principle it takes the difference between the voltages on the two terminals and converts that into a 2's complement binary number. I would say it's common to see this type of ADC used for signals that vary around GND, since in principle negative conversions have meaning in this context. A single-ended ADC is a one-terminal device, where the voltage is converted to a binary number by comparing it to an internal reference (say ground). Typically these are used for sensors that have output a linear voltage in proportion to the phenomenon they are sensing.

\$\endgroup\$
  • \$\begingroup\$ Just to clarify that what an ADC outputs as digital data is independent of the implementation of the analog circuitry. For example the TI ADS8519 is a differential ADC that allows the user to choose either a straight binary or two's complement output. \$\endgroup\$ – spade78 Feb 4 '11 at 19:20
  • \$\begingroup\$ @spade78 i presume by "straight binary" you mean "unsigned binary", right? \$\endgroup\$ – vicatcu Feb 6 '11 at 20:11
  • \$\begingroup\$ Not so much signed or unsigned, more like how the bit codes are mapped to the voltages. Straight binary is defined in the ADS8519 as binary 0 being mapped to the negative rail, binary 32768 represents 0V, and binary 65535 (its a 16-bit device) is mapped to the positive rail. \$\endgroup\$ – spade78 Feb 6 '11 at 21:04
  • 1
    \$\begingroup\$ Some devices let you choose the encoding, usually between signed and unsigned. Fun fact: all they usually do is invert the high bit. \$\endgroup\$ – Mike DeSimone Sep 22 '11 at 5:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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