I think there was an "important insight" that I didn't really capture until I was doing some in-depth calculations: the filter enhances κr by ~ QF compared to the same resonator/coupler without the filter. Now that I get it, it seems obvious, but to be honest, it didn't leap out at me for a long time. It's clear in your thesis equation (4.10), but perhaps this deserves a sentence emphasizing it.
The emphasis is on improving κr*T1, but if I haven't realized that κr and T1 both increase when you use the filter (and primarily κr increases!), a plot of T1 that includes filter vs. no filter is somewhat confusing. There's not much change to T1 vs. the factor of ~ 80 improvement in κr*T1 I'm looking for. I attached a plot of T1 and κr*T1 versus qubit frequency. The two lines have the same resonator/coupler, but one is just hanging on a transmission line, while the other is in a filter based on Kelly 2015. This is from the same calculation I was emailing you about.
Usually when I hear people talking about this filter, it's framed as "decreasing Purcell rate while keeping the same measurement time" (where measurement time is of order 10/κr). Of course, you can use the filter for that, but it seems like one of the big benefits is getting 1/κr down near 10 ns, which is difficult with a simple hanger measurement. That's all while keeping the Purcell rate reasonable. If I understand correctly, that's its main function going from Barends 2014 to Kelly 2015.
I think there was an "important insight" that I didn't really capture until I was doing some in-depth calculations: the filter enhances κr by ~ QF compared to the same resonator/coupler without the filter. Now that I get it, it seems obvious, but to be honest, it didn't leap out at me for a long time. It's clear in your thesis equation (4.10), but perhaps this deserves a sentence emphasizing it.
The emphasis is on improving κr*T1, but if I haven't realized that κr and T1 both increase when you use the filter (and primarily κr increases!), a plot of T1 that includes filter vs. no filter is somewhat confusing. There's not much change to T1 vs. the factor of ~ 80 improvement in κr*T1 I'm looking for. I attached a plot of T1 and κr*T1 versus qubit frequency. The two lines have the same resonator/coupler, but one is just hanging on a transmission line, while the other is in a filter based on Kelly 2015. This is from the same calculation I was emailing you about.
Usually when I hear people talking about this filter, it's framed as "decreasing Purcell rate while keeping the same measurement time" (where measurement time is of order 10/κr). Of course, you can use the filter for that, but it seems like one of the big benefits is getting 1/κr down near 10 ns, which is difficult with a simple hanger measurement. That's all while keeping the Purcell rate reasonable. If I understand correctly, that's its main function going from Barends 2014 to Kelly 2015.