Using an OP amp as output buffer of a DAC

[OM222O]

Hello, I'm building an electronic load and I'm using the 12 bit DAC6551 and a 4.096 voltage reference, to set the voltage refrence for an op amp which basically controls a mosfet gate voltage (via a 100 resistor) according to the voltage drop across the ballast resistor. The DAC seems to have an integrated op amp on it's output so I was wondering if I need another buffer op amp as I've seen many people use it. I really need a rail to rail output so I think I should use a split supply for the external op amp if I use one. I'm also considering using a 128 bit 5k digital potentiometer to calibrate out any errors and possibly have a x2 gain functionality as well.
I would really appreciate any feedback as if it is possible to just use the internal op amp of the DAC to get the desired functionality as it saves me quiet a bit on buying an external op amp.

[_Wim_]

I was wondering if I need another buffer op amp as I've seen many people use it.

According to the datasheet the internal opamp is specified as this:

Output Amplifier
The output buffer amplifier is capable of generating rail-to-rail voltages on its output, giving an output range of
0 V to VDD. It is capable of driving a load of 2 kΩ in parallel with 1000 pF to GND. The source and sink
capabilities of the output amplifier can be seen in the Typical Characteristics. The slew rate is 1.4 V/μs with a fullscale setting time of 8 μs with the output unloaded.

If you do not exceed this loading, no addtional opamp is needed.

I really need a rail to rail output so I think I should use a split supply for the external op amp if I use one.

When using a single supply rail to rail opamp, a split supply can be avoided in this application.

[SiliconWizard]

You should specify the slew rate you're aiming for for your dac-controlled load.
Unless you require very fast rise and fall times, the integrated opamp is more than enough. The series resistor to the gate will limit the current while charging the gate's capacitance. I would increase the resistor's value up to 1k though depending on the MOSFET you use.

[rstofer]

I have always thought that "rail-to-rail output" is marketing.  How in the world can the output driver not have at least a little voltage drop when loaded - even with just 2k?

I would always try to design away from the rails.  I guess it depends on the DAC.

Having said that, using split rails guarantees you can get to true 0V on the output if you get to 0V on the input or if you calibrate out any offset caused by the DAC.

I just don't trust working in the corners.  To be fair, I haven't tested any single rail op amps working at the rails nor have I spent a lot of time scouring datasheets to see what the devices will really do.

[coppercone2]

i need to read more but a rail to rail transistor level topology explanation would be interesting to see

[David Hess]

The DAC is fully buffered so there is no reason to use an external buffer unless greater capability is required.

I have always thought that "rail-to-rail output" is marketing.  How in the world can the output driver not have at least a little voltage drop when loaded - even with just 2k?

That is a very real limitation of "rail-to-rail" outputs.  They do not really operate rail-to-rail but for many applications they get close enough.  When the output transistors operate in saturation, the output resistance without feedback increases to 10s or 100s of ohms.

[OM222O]

Well I do need the absolute 0 volt (or even negative) output to propetly turn off the output fet, otherwise I would have a few mA at the output no matter what. The high rail won't be a problem as it's powered from 5V but 4.096 is the maximum output.

[SiliconWizard]

The DAC is fully buffered so there is no reason to use an external buffer unless greater capability is required.

I have always thought that "rail-to-rail output" is marketing.  How in the world can the output driver not have at least a little voltage drop when loaded - even with just 2k?

That is a very real limitation of "rail-to-rail" outputs.  They do not really operate rail-to-rail but for many applications they get close enough.  When the output transistors operate in saturation, the output resistance without feedback increases to 10s or 100s of ohms.

Yes. Another thing that should be considered is R-R inputs vs R-R output. On some R-R opamps, their operating range on inputs may be narrower than their output voltage span.

[SiliconWizard]

As far as I've seen, this DAC guarantees a true 0V output and in case the 0V setting still had too high an offset, you can apparently put this DAC in shutdown mode and have its output pulled-down to GND. So that should allow a complete shutdown of the MOSFET.

[_Wim_]

Well I do need the absolute 0 volt (or even negative) output to propetly turn off the output fet, otherwise I would have a few mA at the output no matter what. The high rail won't be a problem as it's powered from 5V but 4.096 is the maximum output.

You could use an AD822 for this, this can go below ground using only a single rail:
http://www.analog.com/media/en/technical-documentation/data-sheets/ad822.pdf

From the datasheet:
The AD822 is a dual precision, low power FET input op amp that can operate from a single supply of 5 V to 30 V or from dual supplies of ±2.5 V to ±15 V. It has true single-supply capability with an input voltage range extending below the negative rail, allowing the AD822 to accommodate input signals below ground while in the single-supply mode. Output voltage swing extends to within 10 mV of each rail, providing the maximum output dynamic range.