AD5791 based DAC platform

[sahko123]

can you post the comparison between the two as well. It's interesting to note how the r2r section varies

[branadic]

It's not that one can extract the real datapoints from the diagram, even though the pdf contains scaleable vector graphics, so the answer is no.

-branadic-

[jaromir]

Now that we have ADR1399/LTZ1000/ADR1000 I wonder if someone has finished one of the designs already?

Back in early 2021 I made simple voltage source with LTZ1000A and AD5791B as the main components.

Design requirements
Intended as voltage source for linearity sweeps, with short term stability and low noise as key requirements. INL was expected to be good enough for 5-digit-ish precision, but anticipated use was with external voltmeter.
Apart from +-15V DC voltage, AC range with maximal 10VAC and 50/60/100/400 and 1000Hz frequencies is provided, too.

How it works
This is achieved in a single voltage range, +-16V, so it has some margin over desired range.
Base of this unit is AD5791B in datasheet configuration. DAC has voltage reference +-7V, taken from LTZ1000 source and opamp inverter to obtain negative reference. Output range +-7V is amplified by amplifier with gain of 2,3 to obtain desired +-16V range. Output capability of this amplifier is increased by LT1010 buffer.
It's driven by underutilized dsPIC33 microcontroller, display is bubble starburst, DL1414 type. RS232 for remote control is brought to the back panel.
PSU is unsurprising, toroid transformer with separate windings for earth-referenced logic and +-15V for floating analog section.
Each section (analog, PSU, digital) has separate compartment in the enclosure, with reference and DAC having its own plastic shield.
AC is done with simple sinewave lookup table (no DDS), digital interface is plentiful fast for this task.

Results
I ran a quick DC INL check and found no surprises over what I expected, see attachment.
Despite being designed as test source for short term stability, I found it was quite stable. It drifted approximately -4ppm after being continuously powered for three months (most of the drift happened in first week, likely relieving stress from soldering) and stayed there since then (~18 months) within 1ppm range. I can only compare it against a single Solartron 7081 and bunch of 6 digit meters with LM399 reference, so not much of a metrological value in those.

Things to take a look at
I made this as a tool for my personal use, from what I had lying around. No design files attached here, as this project is somehow untidy and half-baked, it takes a bit of effort to make it publishable. On the other hand - I really doubt anyone would replicate it and no particularly big surprises are hidden in this design to learn from.
I collected a bunch of higher-resolution photographs in an imgur album here https://imgur.com/a/W15Z97J and shaky video of output switching by 1V steps

Resume
AD5791B works as datasheet says. LTZ1000A too.

[r6502]

Hi all,

It is an interesting Project, similar to LT AN86 and Xdevs's realisation with 18bit DAC's instead with 16Bit DAC's in LT's AN86.

I wonted to realise this in 2023, but  I think it will be difficult to realise as the needed parts are not available  ...

Guido

[Mickle T.]

These are the results of scanning the MSB 9-bits of the AD5791 DAC with voltage measurement using the recently adjusted R6581.
Attention is drawn to the bifurcation point in the middle of the scale, after which the differential nonlinearity increases.

[macaba]

It looks like you've got a AD5791 that has a better matching between the 6-bit segment and the R2R segment compared to the 2 that I tried. Keep it safe and use it well! 

[branadic]

Thanks for the results shown so far. I wonder how a relatively simple 20 bit DAC like in the Data Precision 8200 performs to that in comparison. Sure, the reference in there is "only" an LM299 with its rather large noise, but assuming it would be replaced by the lower noise ADR1399, mh?

-branadic-

[Mickle T.]

The AD5791 configuration register has semi-documented LIN COMP bits that can be used to compensate for 3rd order residual non-linearity.
The graphs show the amount of compensation for different values of the LIN COMP bit field. (Advantest R6581 PLC100, AD5791B, LTZ1000 10V).

[NNNI]

Has the 'Gain of Two' configuration from the AD5791 datasheet (https://www.analog.com/media/en/technical-documentation/data-sheets/ad5791.pdf), Fig. 53 on page 25 been tried before? Does it have particular advantages in terms of drift or tempco, since the resistors are part of the same package?

[iMo]

Provided the two resistors are wired as a divider (it seems they are) it will help as the "ratio TC" with both resistors on a single die could be low.

[Mickle T.]

In the last two versions of AD5791-based multifunction calibrators, I have included support for both DAC modes (Gain 1x with Ib compensation and Gain 2x) to implement unipolar and bipolar ranges. During the tests, I did not find any measurable differences. Even the INL curve matches up to 0.1 ppm.

[felixd]

Hi Jaromir.

Any updates on the project?

[jaromir]

Not really. I did no adjustments or other tweaking to the design since 2021, so I guess it reached its maturity. It serves me well for four years, mostly as a precision voltage source for linearity measurement sweeps.

If I were to build it again, I'd go for lower common mode noise PSU and would add option for voltage dividers to obtain +-1,6V or +- 0,16V ranges.

[Mickle T.]

Inspired by the great projects of Jaromir and Micke (https://www.eevblog.com/forum/metrology/diy-acvdcv-calibrator-with-ad5791/) I also made a simple ACV/DCV/DCI/OHM calibrator, slightly reworking the old DC-Source based on ADR1399+AD5791 (https://youtu.be/0AOd9971DGY).
The modification primarily affected the transition from software to hardware SPI ATmega328P and switching the /LDAC signal from high to low level (I had to raise the pin of the DAC chip).
Two ACV ranges 0...0.7 and 0...7 V RMS with a frequency from 10 Hz to 1 kHz were added. A table in RAM is used to generate a sine wave. At frequencies up to 1 kHz, the table has 400 samples, at 1 kHz the table length is reduced to 183 samples (SPI frequency is only 8 MHz).
The rest of the capabilities remain unchanged: two DCV ranges +/- 1 V and +/- 10 V, five DCI ranges from 10 µA to 100 mA, six standard resistors from 10 Ω to 1 MOhm.

[derZockerMP]

 Nice project
Do you have photos from the PCB without the cover. Which precision resistors do you use?

Best Regrds

[Mickle T.]

Nice project
Do you have photos from the PCB without the cover. Which precision resistors do you use?

Best Regrds

Selecting resistors is the number one headache. Since the sealed oil-filled Vishay BMFs are currently unavailable for order due to embargo, I tried to select from the available stock of metal-foil and precision wire-wound resistors. Unfortunately, their overall dimensions are often beyond good and evil

[Mickle T.]

One of the advantages of the AD5791 based DAC platform is the ability to easily expand its functionality. For example, when I needed to reproduce a current greater than 103 mA, I quickly assembled a simple and dirty voltage-controlled current source unit. It uses an ALMA 0.1 Ohm current shunt, whose residual temperature coefficient of 3 ppm/C is compensated by a copper resistor installed inside the ceramic tube. This, in turn, minimizes the error due to self-heating of the shunt. The slope of the transfer characteristic of 1 A / V is provided by a Hammon divider of 6 sealed metal-foil resistors C5-61, selected to sub-ppm TCR.