4

I have developed a DAC-based signal generator board that I want to use to generate sine waves of different amplitudes and frequencies. The board was working but unfortunately, the digital-to-analog glitch of my initially chosen DAC, the DAC8830 from TI, seriously degraded the signal quality.

Fortunately there is a pin-compatible DAC from Analog Devices, the AD5541 that has similar specs but a way lower digital-to-analog glitch (1.1nV-s instead of 35nV-s.)

I ordered this chip, desoldered the DAC8830 and inserted the AD5541 instead. The glitch is now barely visible but unfortunately my precision voltage reference, the REF6250 from TI, has become unstable and I have absolutely no clue why.

Here's the schematic of my DAC front-end:

The schematic of my front-end

Here's the schematic of the voltage reference:

The schematic of my voltage reference

Here's the oscilloscope screenshot of the unstable reference:

The unstable voltage reference

Has anybody an idea what I am missing here?

The reference was stable before and I only changed the DAC and nothing else. The "reference input" specifications of both chips is almost identical.

EDIT: PCB-Layout
My PCB has 4-Layers with the Stackup (Top to Bottom): Signal - Ground - Power - Signal(Ground)

Here's a detail view of the DAC and the reference on the top layer: PCB layout top layer Please note, that the 68pF capacitor C52 was included afterwards across the terminals of R12.

Mau5
  • 140
  • 7
  • 2
    47uF on a V-Ref output seems a lot. Maybe your DAC adds extra capacitance and the V-Ref starts ringing. For a quick test your could replace it with 1u for example. – ElectronicsStudent Jan 17 '23 at 08:51
  • Measure the current going into your 20k and into your DAC ref pin. Are they what you expect? – Designalog Jan 17 '23 at 09:00
  • Check your 5.5 volt rail for too much ripple. Maybe the AD device draws a few more mA from the 5.5 volts and it's gone unstable/ripply/noisy. – Andy aka Jan 17 '23 at 09:08
  • @Andyaka the 5.5V rail is very clean and stable – Mau5 Jan 17 '23 at 13:04
  • 2
    @ElectronicsStudent i know but the REF6250's datasheet recommends capacitances from 10uF to 47uF. TI even has a reference design board (REF6025EVM-PDK) where they use a 47uF capacitor (i took the exact same part). However, it seems that you were right. I put in a 22uF capacitor and now the output is stable. Thanks :) – Mau5 Jan 17 '23 at 13:10
  • @Mau5 you could insert a 1 ohm resistor to feed the 47 uF and this should work and, take note that this is the same as what Rohat says in his answer. – Andy aka Jan 17 '23 at 13:13
  • @Andyaka unfortunately not. The ESR of the capacitor should be between 5mOhm and 100mOhm according to the REF6250 datasheet. Increasing the ESR to 70mOhm has not shown an improvement. However, reducing the capacitance as ElectronicStudent mentioned resolved the problem. – Mau5 Jan 17 '23 at 13:23
  • @ElectronicsStudent please post your comment as an answer, such that i can mark it as a solution and close the question :) – Mau5 Jan 17 '23 at 13:27
  • @Mau5 Done - thank you! – ElectronicsStudent Jan 17 '23 at 13:32

3 Answers3

4

I suggest reducing the value of C45 as 47uF seems too high.

This, in combination with extra capacitance added by the DAC input could cause the ringing.

Reducing the capacitance - while staying in the recommended range (C and ESR) - can, in my opinion, solve the problem.

I would recommend 1uF as a starter.

EDIT 1:

As the VREF-IC requires 10<->47uF, the "person who asked the question" used a 22uF for C45.

JRE
  • 67,678
  • 8
  • 104
  • 179
ElectronicsStudent
  • 2,746
  • 5
  • 18
  • 2
    Hi, Just to be clear, what's a "TO" in "the TO used a 22uF"? Is that Topic Originator or similar? On Stack Exchange, the jargon used for that is OP (Original Poster - can also refer to the Original Post, rather than the person who posted it). Can you please clarify? Thanks. – SamGibson Jan 17 '23 at 14:03
  • TO="Thread-Opener" like in reddit. – ElectronicsStudent Jan 17 '23 at 14:08
  • 1
    ElectronicsStudent - "*TO="Thread-Opener" like in reddit*" Thanks for explaining. I don't use reddit :) FYI that is not the jargon used here. I recommend you use the Stack Exchange "OP" jargon for that in future, to avoid confusion. Thanks. – SamGibson Jan 17 '23 at 14:38
  • Reddit people say "OP", too - Original Post-er. – user253751 Jan 17 '23 at 17:09
3

The problem is the output capacitance, C45. The ESR of the output cap is needed for stability, so using a ceramic cap with zero ESR is calling for a trouble.

Replace it with an aluminium electrolytic cap, or simply replace R10 with something between 20 mR and 50 mR. Up to 0R1 should be alright.

Rohat Kılıç
  • 26,954
  • 3
  • 25
  • 67
  • You are right. However, the total series resistance from the C45 bottom terminal to ground is approx. 20mOhm, which is in the stability range according to the datasheet. I replaced R10 with an 47mOhm resistor. Unfortunately there was absolutely no change in the reference output. – Mau5 Jan 17 '23 at 13:17
2

In my experience with this part, OUT_F and OUT_S must be connected together at the capacitor terminals; same goes for GND_F and GND_S. This connection scheme should also be visualized on the schematic. The star ground at the capacitor terminal should then connect to the ground plane. And there definitely should be a ground plane. It also helps with stability to have the capacitor-reference unit close to the DAC - right next to it, even.

precision voltage reference [...] has become unstable and I have absolutely no clue why

It probably was barely stable to begin with. It helps a lot with these kinds of questions to have a readable, clear PCB layout images of all relevant layers that cover the reference, the DAC, and the decoupling capacitors.

The reference may also be destabilized by the impedance on the input. I've faced that very problem once.

Kuba hasn't forgotten Monica
  • 32,734
  • 1
  • 38
  • 103
  • You are right this information is missing but would be very important. I have edited my original post with an image of the top layer. The OUT_F and OUT_S pins are connected as you suggested. However GND_F and GND_S are both directly tied to ground. I thought that this is not very harmful, since I have a solid GND plane on the next layer which is of very low impedance and thus should not affect the operation of the reference. The input impedance is the same for both DAC's (9kOhm), however the input capacitance of the AD5541 is lower (26pF vs 75pF) and the input BW is higher (2.2MHz vs. 1.2MHz). – Mau5 Jan 17 '23 at 21:00
  • Just for kicks, I’ve hooked a network analyzer to a few DAC reference inputs on eval boards through an AC coupling network and they all seem unique past 100MHz or so. Some are very capacitive at lower frequencies, one is inductive, some are resistive. If such small changes get the reference unstable, it wasn’t really stable before. You just had more luck :) The input BW doesn’t always translate cleanly to a transfer function. Some old DACs that have quite low multiplier -3dB frequency are quite simple on a Smith chart and you’d think they go to hundreds of MHz but they don’t. YMMV. – Kuba hasn't forgotten Monica Jan 19 '23 at 20:24
  • That's very interesting. Yeah it seems that my reference was not really stable before. I designed three different boards with the exact same reference IC and surrounding parts. All of them were stable except of this one after changing the DAC. There is even a reference design for the REF6250, which uses the same component values for the surrounding parts. C45 is even the exact same part. Additionally, I don't think that my PCB layout is extremely bad. So i really don't get what happened here. Maybe it's some temperature related thing due to the manual de- and resoldering of the DAC. – Mau5 Jan 20 '23 at 08:00