Combining DACs for more current?

[rea5245]

If I use a dual DAC like the MCP4802 (https://ww1.microchip.com/downloads/en/DeviceDoc/20002249B.pdf) which has a typical short circuit current of 15 mA, is it reasonable to tie the two Vout pins together to drive a higher current load? (Obviously, I'd need to set both DACs to the same value.)

Thank you,
   Bob

[David Hess]

Current sharing between the paralleled outputs will be poor because of the low output resistance from the closed loop operational amplifiers.  Voltage regulators, references, and operational amplifiers have the same problem and special circuits are used to allow parallel operation.

[OM222O]

No! You use an amplifier circuit for that 

[Mr. Scram]

No! You use an amplifier circuit for that 

No need to be snippy. He's asking us exactly because he's aware there may be better solutions he doesn't know.

[rea5245]

No! You use an amplifier circuit for that 

Yeah, I know I need an op amp. Buffered DACs have op amps built in, right? So that's what I'm looking at. But even with the built-in op amps, they all seem to have relatively low current outputs. I'm looking for at least 30 mA.

- Bob

[OM222O]

OP amps are low current devices themselves (unless you buy special ones with high current output). They need buffer ICs themselves to boost their current. if your dac already has an op amp on the output, then just connect it to the buffer IC and you're good to go. If you need even more current (in order of a few amps) then you need to use a FET + OP AMP as a current buffer.

[ejeffrey]

The internal opamp is generally still meant to drive signal level / light loads to reduce the effects of parasitic inductance and capacitance.  There are exceptions, but they tend to be more special purpose devices.  If you want to drive substantial power you would still normally use an external amplifier, and one designed for your load.  So without more information of what you are trying to accomplish it is hard to give any other recommendation than use an external amp.

As David Hess explained, paralleling voltage sources doesn't work very well.  You either need ballast resistors to force current sharing or a feedback network that uses a single error amplifier to drive multiple output devices.  Otherwise the different offset errors will lead to unequal current sharing ultimately leading to one device reaching its current limit well before the desired output current.  Since most voltage sources are not designed to have good behavior in the current limiting regime this can really degrade your performance.

[David Hess]

No! You use an amplifier circuit for that 

Yeah, I know I need an op amp. Buffered DACs have op amps built in, right? So that's what I'm looking at. But even with the built-in op amps, they all seem to have relatively low current outputs. I'm looking for at least 30 mA.

The problem is that you only have access to the output from the internal operational amplifier and not any part of its feedback network.  Parallel operation generally requires breaking the connection between the output and feedback network.

[ejeffrey]

One more thing: outside of dedicated line drivers or high speed devices intended for 50 ohm systems datasheet specs for DACs and amplifiers will generally be measured at a small fraction of the maximum output current.  So your Even using a single channel at near the maximum load will likely have degraded performance relative to the datasheet.  For instance on the MCP4802 all specs are at a load of 5000 Ohm // 100 pF unless otherwise noted.  That should give you some idea of what the manufacturer is expecting.  Accuracy, linearity, and settling time may all be diminished when operating near the maximum output current.

[Vovk_Z]

30 mA is not a big current for many of actual opamps, even some precision ones. For example AD823 (not very precision), AD8610 (precision), OPA189 (very precision one).
You can't simply parallel outputs of opamps (and DAC) but you can parallel them through some small resistors (10 ... 100 R). Those resistors will safely help to divide the whole current to two equal parts. I have made two headphone amplifiers with parallel LM6172 (two in parallel with 33 R series resistors). It is an old trick described in many opamp application notes.

[NiHaoMike]

One "hack" you can do with a single output DAC is to supply the DAC via a resistor, then wire in a PNP transistor so it adds to the output current when the DAC starts drawing much more than its quiescent current. You'll probably also want to add a capacitor across the resistor to prevent oscillation.

Another solution is the classic NPN transistor on the output to boost the current, just keep in mind the voltage drop varies significantly with temperature. Using a small linear regulator instead of the transistor will give you a positive offset instead of a negative one, plus it would be more stable over temperature. Then there's the LT3083 that has close to zero offset.

[David Hess]

30 mA is not a big current for many of actual opamps, even some precision ones. For example AD823 (not very precision), AD8610 (precision), OPA189 (very precision one).

High output current is incompatible with precision, even on the chopper-stabilized OPA189, because of thermal effects from self heating.  This is why precision operational amplifiers are characterized with high load resistance, and in applications where it matters, they will be externally buffered.  They implement current limiting only for self protection and not with the idea that they should be operated with high output currents.

The above is also why lower power parts with lower input bias currents may outperform higher power parts with higher input bias currents; lower input bias current allows for higher impedance feedback which causes less self heating in both the operational amplifier and feedback network.

[Vovk_Z]

 You are right of cause in general. But in real life hot high loaded OPA189 will be still much more accurate than cold AD8610, for example. Hot AD8610 will be more accurate then AD820-823, etc.
We know very few about accurasy TS needs, but MCP4802 is a 8-bit DAC - it is hard to make it worse with hot precision opamp. I think any opamp with a maximum offset voltage (at 125 C) less then 1-2 mV should do. There are many of them.  TS may need a low voltage rail-to-rail ones - I'm not familiar with them unfortunately and can't give a list without some investigation.

[David Hess]

The AD8610 is not remotely competitive with the OPA189.  The proper comparison is with parts like the bipolar OP77, OP177, AD708, and similar (1) which have offset voltage drifts of 100 nV/C compared to the typical chopper stabilized amplifier's offset drift of 50 nV/C.

What these parts all have in common is that their offset voltage drift is dominated by self heating and the immediate environment.  Under ideal conditions, chopper stabilized parts have a lower drift but without thermal baffles and without high load resistance, they all perform equally poorly; thermocouple effects will dominate.  With the AD8610, the difference in performance would be much less.

Where precision bipolar parts cannot compete is with lower flicker noise since chopper stabilization removes it.

(1) Since trimming the offset voltage of these types of precision bipolar operational amplifiers reduces offset voltage drift, even parts with higher drift are suitable if offset nulling is used.

[Vovk_Z]

The AD8610 is not remotely competitive with the OPA189. 

- no, it is competitive if we compare only its output current (and costs much less). 
The TS question was only about current. So I proposed several options.