daniel.sigg@LIGO.ORG - posted 14:41, Monday 06 February 2023 - last comment - 16:30, Monday 06 February 2023(67245)
Front-end decimation filters
Looking at the hard coded 4x decimation filter in the front end I noticed that it has zeros in the right half plane.
The attached plot shows the transfer function of this filter compared to one where the zeros have been flipped into the left hand plane. The difference in phase is about 1° at 1kHz.
Non-image files attached to this report
Comments related to this report
Tagging CAL. Note -- this filter, sometimes referred to as the "IOP downsampling filter," has been around since 2016 and is used on every 65 kHz to 16 kHz channel (which is every one except for the new 524 kHz OMC DCPD channels). Peter designed an update to this filter for less phase loss and less ripple -- see T1600066. You can see, even in that note, the two complex pair poles are designed to have phase of 127 and 95 deg. The phases should be between 0 and 90 deg for them to remain in the left-half-plane. Huh! (Note for any other sample rate, which, in our case is really only 4096 Hz for the SEI models and 2048 for the ASC models, they use filters from a different design, so don't suffer from this same flaw).