Other Parts Discussed in Thread: LMP7731
Hello,
We have a new PCB prototype design using the DAC8832IRGY. The 8832 is used as the adjustable programming voltage for a constant current source.
In trying to characterize the responsiveness of the current source, I was led down the path to evaluate the DAC8832 output response/settle time based on a new code being provided to the DAC.
Here is a snapshot of our DAC8832 circuit:
Designed SPI/LDAC behavior: The LDAC line is kept high. We send our DAC code down via SPI, the pulse the LDAC line low for 60nS, then bring it back high. I have verified all the digital timings are being met per figure 1, of the data sheet.
Test Setup & Results:
I was seeing settle times on the order of milliseconds. Thinking it might be the LMP7731, I lifted the DAC-OUT-B pin going into it, so the DAC8832 was unloaded. The following plots are with the DAC8832 unloaded.
A Lake Shore VM-10 was used to make the voltage measurements. Data was sampled deterministically at a 1kHz rate, with the VM-10 set to an NPLC of .06 (1mS @60Hz line) to sync up with the sample rate. A P2060 60Mhz/10MOhm probe was used to attach the VM-10 to the "TP25" test point on the output of the 8832.
The test involves the DAC sitting at a steady 3.5V output. The code is sent and the output is changed to 2.6V:
On the surface, all looks well. However, if we zoom in, we can see the step change begin at the 3.285 second mark, and appear to settle at the 3.294 second mark, a difference of ~9 milliseconds.
If we dig a little further into the plot, and look at the next 100mS of settling (from the 3.300 second mark, to the 3.400 second mark), we see further settling that continues.
At 3.3secs the voltage output is 2.594258626, and at 3.4 second mark, the output is 2.593166206. This is a further change of approx. 0.04%
For context, we want to use this DAC and current source in a high accuracy application (0.03% to 0.04% accuracies, after we perform our factory calibration) where we have a 50mS hardware settle time budget in between measurements of a DUT, that this current source is driving. Presently, I can't hit my timing/accuracies with the a) multi-millisecond step response to get the first 90% and b) the approx 100mS to have that last 0.04% settle out.
Can you please provide any insights? Are we using the DAC incorrectly? Even though unbuffered, I would think the 8832 output would settle within a few microseconds. I realize the 8832 data sheet states 1uS into a 10pF load. I also realize my VM-10 has an input impedance of around 120pF, but still, that doesn't even get me anywhere near multiple milliseconds that I'm seeing.
Regarding layout, the test point I'm using is right near the output of the chip:
We've also ensured that the thermal pad is soldered to our analog ground 2.
Thanks for any insights you can provide.
