What's the use-case for gain selection on a DAC?

Question

I'm looking at the datasheet for an MCP4921 12-bit DAC (PDF), which describes the following feature:

Selectable Unity or 2x Gain Output - The user can configure the full-scale range of the device to be V_REF or 2*V_REF by setting the gain selection option bit (gain of 1 of 2).

This doesn't give me much of an idea as to why anyone would need such a feature. It seems to me that one could just increase V_REF, instead of using the gain selection. So I read further into the sheet:

In some applications, precision digital control of the output range is desirable. Example 6-4 illustrates how to use the DAC devices to achieve this in a bipolar or single-supply application. This circuit is typically used in Multiplier mode and is ideal for linearizing a sensor whose slope and offset varies. Refer to Section 6.9 “Using Multiplier Mode” for more information on Multiplier mode.

And, from section 6.9:

The MCP4921 uses an external reference, and these devices are ideally suited for use as a multiplier/divider in a signal chain. Common applications are: (a) precision programmable gain/attenuator amplifiers and (b) motor control feedback loops. The wide input range (0V – VDD) is in Unbuffered mode, and near rail-to-rail range in Buffered mode. Its bandwidth (> 400 kHz), selectable 1x/2x gain and low power consumption give maximum flexibility to meet the application’s needs.

This seems pretty vague - I don't see how having the output range go to 2*V_REF would be a common enough requirement to warrant making it a feature of the device.

What am I missing? Is there a common use for DACs that makes such a feature useful / beneficial? Why wouldn't we just set V_REF to a higher voltage if we wanted that increased range?

Answer 1

Simple use case: you want an output higher than the internal VREF (2.048V) but don't want to add to your parts count with an external reference.

The less obvious case is what the datasheet calls "multiplier mode". In this case you keep the digital signal fixed and feed an analogue signal into VREF. That gives you a programmable gain amplifier, and the 2*VREF mode lets you have a gain between 0 and 2. That could be useful for analogue control loops or sensor biasing.

Answer 2

In general there are two good reasons in the ideal case:

1) there is a noise floor that you have to get over, so with an increased gain you limit output range but gain back on the bottom end. IN this case the gain is applied at the front of the conversion chain and is carried through (above inherent noise from amplifiers and KTC) at the risk of clipping on the MSB.

2) internal architecture could be fully differential and the gain of two is a freebee by converting to side ended on the output. yet another feature with the benefits of #1 above.

However, there is nothing in the datasheet that I could see that verifies this. Looking at the DNL curves they only show the gain = 1 mode. And there are no measurements shown (that I can find anyways - after a cursory scan) that shows the performance differences at the different gains.