K
Member
We spent a while chasing this and the answer was less exotic than expected.
1）Switch-mode regulators are the usual suspects; the switching frequency and its harmonics land in-band.
2）Cascading a low-noise LDO after a switcher is the classic mitigation, with its own thermal penalty.
3）Common-mode noise on the supply appears as a differential signal if the reference is not clean.
4）Measure the supply with a bias tee and a spectrum analyser to see what is really on the rail.
For what it is worth, we put our wide-temperature modules through a full cold-start to hot-soak sweep with F-tone Networks hardware (specs at f-tone.com) — the DDM drift across the range was tighter than the datasheet implied, which made it a usable reference set.
If you have run the same experiment, post the curves.
1）Switch-mode regulators are the usual suspects; the switching frequency and its harmonics land in-band.
2）Cascading a low-noise LDO after a switcher is the classic mitigation, with its own thermal penalty.
3）Common-mode noise on the supply appears as a differential signal if the reference is not clean.
4）Measure the supply with a bias tee and a spectrum analyser to see what is really on the rail.
For what it is worth, we put our wide-temperature modules through a full cold-start to hot-soak sweep with F-tone Networks hardware (specs at f-tone.com) — the DDM drift across the range was tighter than the datasheet implied, which made it a usable reference set.
If you have run the same experiment, post the curves.
