# DAC controlled constant current source

Was given the task of creating the simplest constant current source that can supply up to 50 mA with an ~700 ohm load. I can use any IC's out there. I stumbled across the chip here from linear tech: Which seems to fulfill all of my needs, however the problem I am facing is that I need to create 16 of these DAC controlled current sources and all of the loads on the chip are common grounded, while the image suggests that any load attached would need to be connected to the negative supply. Is there any simple fix that would allow the loads to be grounded? My first instinct was making the ground on the diagram the positive rail, but I am not sure if this would alter the input I need from the DAC.

• the load has to be connected to the +ve rail, this is a current sink. This is probably more trouble than it's worth to modify to source a current into a ground connected load, a PNP current mirror per channel would be the minimum needed. Jun 29 '16 at 21:13
• What precision and accuracy do you need? And are you aware that 50 mA through 700 ohms requires 35 volts? Are you prepared to include a 40 volt power supply in your project? Jun 29 '16 at 21:20
• Doesn't have to be extremely precise, looking for a .01 mA precision which just means we need at least a 12 bit dac, and we are using a 16 bit dac. For the voltage, yes, the max output voltage is 50 for this chip Jun 29 '16 at 22:16

The LT3092 itself is designed for full-floating applications -- the only reason the application circuit you mentioned is GND-referenced is because it's assuming a GND referenced DAC output. Using an isolated DAC output that floats with the rest of the regulator is the conceptually simplest way around this:

simulate this circuit – Schematic created using CircuitLab

Caveats:

• U2 needs to be a low-power digital isolator solution (phototransistor output optos are too piggy) that can run off of 2.7-3V
• R1 and R2 are a matched pair -- and should be the best matched resistor pair you can get your paws on
• I wasn't able to fit all the decoupling caps in the above schematic -- C1 is the regulator output cap, and C2 and C3 are representative decoupling caps.
• The Iq of the ICs flows into the load, limiting the minimum current (this is true for the LT3092 by itself with its 500µA Iq -- I picked the parts to not add to it significantly.
• Vin needs to be at least 3V higher than the desired minimum voltage at Iout in order for stuff to work.
• The LTC2641 can be replaced with another suitable (2.7-3V operation, low power, SPI compatible, unipolar 0-VREF output with a min 2V VREF) DAC, and the LT6656-2 can be replaced with any other suitable 2.048V reference
• 2V references aren't a thing, so we use a 2.048V ref instead -- this introduces a portion of the high end of the DAC's range that isn't quite usable unless you want to trim the R1/R2 divider -- you may wish to introduce a trim there anyway to deal with Iq issues as the LT3092 will lose regulation if you try to ask it to flow less current than its own Iq. (i.e. a low end trim)
• U3 needs to be a low-Iq (preferably MOS-pass-device), high voltage (needs to withstand max Vin-Vout differential for the circuit), low noise, low dropout regulator -- these are rather hard to find. (Most micropower PNP LDOs consume way more than their Iq when starting up, and most MOS LDOs aren't high voltage capable.) It also helps if it's stable with ceramic caps -- if C1 being piezoelectric turns out to be a bother, a film cap can be subbed for it penalizing naught but space.
• Analog isolators were considered briefly, but accurate/precise analog isolation is far harder than accurate/precise DAC outputs, and consumes more power as well
• Just goes to show, RTFM! Jun 29 '16 at 23:55
• Depending on application, you can also add a calibration step where you record the relationship between DAC values and Iout, so you can compensate for manufacturing tolerances (even if they are small). Jun 30 '16 at 0:10
• @SimonRichter -- that is definitely possible :) now, if only finding a high voltage, RF-grade (i.e. very low noise) MOS LDO wasn't so hard... Jun 30 '16 at 0:13