search

OxfordIonTrapGroup / stabilizer_current_sense

Low-noise current sense mezzanine for Stabilizer
Other
4 stars 2 forks source link

DAC-Filter-Capacitor causes large temp-co #10

Closed pathfinder49 closed 5 years ago

pathfinder49 commented 5 years ago

C33 is causing ~50 mV fluctuations in V_err. 67689050-ac229600-f992-11e9-9db8-70882b72895f This measurement is when lying on a desk, undisturbed.

Covering the capacitor with tissue removes the fluctuations. De-soldering the capacitor gives a steady V_err with noise consistent with the unfiltered DAC.

For the de-soldered capacitor: RMS noise = 8.4 mV 200 kHz pole in V_offset addition 2000x gain to v_err => 9 nV/sqrt(Hz) noise From DAC data-sheet: 11.8 nV/sqrt(Hz)

Screenshot from 2019-10-28 18-51-24

cjbe commented 5 years ago

Good find! What is the plan to fix this - replace it with an electrolytic?

pathfinder49 commented 5 years ago

The Previous measurement was with the board:

TS 270/19 002

Using another board:

TS 270/19 001

The behavior is NOT observed. The output is stable with 6 mV RMS noise. Touching the capacitor still causes causes large voltage spikes of over 1 V.

pathfinder49 commented 5 years ago

Good find! What is the plan to fix this - replace it with an electrolytic?

Seems to be the case of a broken capacitor?

hartytp commented 5 years ago

I don’t know what to do here. Let’s gather statistics with the other boards before making a decision. Maybe it was a broken cap (I don’t understand the physics behind such a large thermal emf otherwise)?

With electros I’d worry about the leakage temp co as the dac has a high z output. Could consider reducing the cap size and going for a large plastic?

dnadlinger commented 5 years ago

We could always add a second buffer stage and use the leakage-free parallel RC technique (i.e. keep zero volts across one capacitor up to R matching, which can then be leaky).

dnadlinger commented 5 years ago

It really feels like this shouldn't be that big an effect, though.

hartytp commented 5 years ago

Let's assume it was a broken cap. We'll gather more statistics for looking at other boards, and only worry if the issue surfaces on a new board...