Designing simple DIY Voltage Calibrator 0-10V

[slavoy]

I haven't designed anything in many years, and I'd like to get back into it. For my first project, I chose to build a workshop DC 0-10V calibrator using a reference voltage source, DAC, and Arduino. The project should be simple, fairly cheap, easy to replicate, and ideally not require initial calibration for someone who doesn't have the appropriate equipment.
I'm aiming for 0-10V regulation with 1 mV step and precision. I want to see that 5,000V on a multimeter xD The accuracy and stability... well lets say a proper for this job and price 
I chose the following components for this project:
AD5696R: 16-bit DAC with a built-in 2.5V reference voltage source and 1x/2x gain selection
https://www.analog.com/en/products/ad5696r.html
OPA192 as an output amplifier with 0.1% resistors
https://www.ti.com/product/OPA192
Dac will be controlled by I2C with Arduino (nano?), a simple 2x16 display, buttons, an encoder, and user calibration capability. Battery and external powered (?).

 I'd like to stick to a budget as low as possible, preferably below $50.
What are your suggestions regarding this project and what do you think about the choice of components?

[Kleinstein]

Is there a real need for 4 channels ? There are cheaper ( 1/3 the costs) versions with 1 channel (e.g. AD5683/5693).
I would still consider an external reference, if one really wants to rely on the absolute accuracy of the reference.
One may consider just an extra 2.5 V or possibly 2 x in series as an extra to have a check.

For the OP-amp the OPA192 is a nice FET type, but for precision at low cost I would consider the OPA202.
There are sets of equal resistors available as a set  (ACAS series or Susumu RM). These may be an alternative to single 0.1 % ones.

[MasterT]

I tested AD5689 vs DAC80501:

Vertical scale in ppm. Both above +-1 LSB INL, but TI getting close to linearity of 20-bits, very impressive.

 Stuck with attachments, remove - add does not seem to make difference.

 Try to "open image in new tab" to rescale properly by browser

[jwet]

You don't mention how the power supply will be implemented.  Battery is a great way to go, you get "free" isolation from ground.  If you really want 0V volts output, you'll have to have at least a small negative supply or pseudo ground, etc- an amp powered by single supply can't make 0v- it can't sink.

You'll need to make a 10+ volt rail, 12-15v is convenient, you'll want a negative rail, perhaps -5v from a simple charge pump inverting the +5V.  You could also use a pseudo ground scheme that used +5 and +17 say, and bring the +5 out as the analog common out as analog ground.  This might make debug tricky with the arduino initially but you can work around this.  For a decent portable instrument, you'll want low battery indication,  backlight control (if applicable), auto shutoff, etc.  You'll need to figure out how much battery you need based on your Icc's and conversion efficiency.

You'll have to look at PSRR to determine how quiet your rails will have to be-  luckily the currents are pretty low so filtering won't be too difficult.  I would recommend separated analog and digital supplies and extra care given to the grounding scheme.

The reference in the DAC you chose is surprisingly good at 2 ppm (typ, 5 max) for the B grade.  A biploar precision amp would make more sense to give you low offset and offset drift.  You might want to have a short circuit protection scheme if the op-amp's isn't robust enough.  Bench equipment takes a lot of abuse.

Sounds like a nice project- good luck.

[pqass]

Rather than a charge pump,
a bipolar supply from a battery (or single source) can be simply generated via diode(s) or zener and redefining circuit ground.
See attached.

[jwet]

I assume you realize that ground can't sink current for this circuit- you need an active device.

[pqass]

I assume you realize that ground can't sink current for this circuit- you need an active device.

Seems to work fine (simulation here).

EDIT:  You are partially correct. 
The circuit only works to the extent that the negative rail current draw can't exceed the positive rail current draw (ie. until 0mA through the ground).  But then the context was to provide a negative rail for an opamp with everything else (DAC, reference, Nano, etc.) on the positive rail.  You could also add a dummy resistance on the positive rail.  You're trading more power consumption for switching noise and complexity of a charge pump.

[slavoy]

Is there a real need for 4 channels ? There are cheaper ( 1/3 the costs) versions with 1 channel (e.g. AD5683/5693).

No need of course. I just need some better source for selecting components for the project

I would still consider an external reference, if one really wants to rely on the absolute accuracy of the reference.
One may consider just an extra 2.5 V or possibly 2 x in series as an extra to have a check.

Will do! Thanks.

For the OP-amp the OPA192 is a nice FET type, but for precision at low cost I would consider the OPA202.
There are sets of equal resistors available as a set  (ACAS series or Susumu RM). These may be an alternative to single 0.1 % ones.

Thank you sir 

[slavoy]

You don't mention how the power supply will be implemented.  Battery is a great way to go, you get "free" isolation from ground.  If you really want 0V volts output, you'll have to have at least a small negative supply or pseudo ground, etc- an amp powered by single supply can't make 0v- it can't sink.

You'll need to make a 10+ volt rail, 12-15v is convenient, you'll want a negative rail, perhaps -5v from a simple charge pump inverting the +5V.  You could also use a pseudo ground scheme that used +5 and +17 say, and bring the +5 out as the analog common out as analog ground.  This might make debug tricky with the arduino initially but you can work around this.  For a decent portable instrument, you'll want low battery indication,  backlight control (if applicable), auto shutoff, etc.  You'll need to figure out how much battery you need based on your Icc's and conversion efficiency.

You'll have to look at PSRR to determine how quiet your rails will have to be-  luckily the currents are pretty low so filtering won't be too difficult.  I would recommend separated analog and digital supplies and extra care given to the grounding scheme.

The reference in the DAC you chose is surprisingly good at 2 ppm (typ, 5 max) for the B grade.  A biploar precision amp would make more sense to give you low offset and offset drift.  You might want to have a short circuit protection scheme if the op-amp's isn't robust enough.  Bench equipment takes a lot of abuse.

Sounds like a nice project- good luck.

Thanks for pointing it out, seems like I have a lot to think about now  But that's great. I want to learn 

I tested AD5689 vs DAC80501:
Vertical scale in ppm. Both above +-1 LSB INL, but TI getting close to linearity of 20-bits, very impressive.

That looks nice. I think I might go for DAC80501 now.
Thanks for any advice