REW's built-in high-pass adds phase shift, is there any way to remove dc but keep the phase?

Hi!
I'm rtying to measure emitted acoustic energy by a subwoofer (vs frequency, of course) by measuring the cone velocity (with an accelerometer), pressure (with a microphone), and scalar-multiplying them to obtain the energy. I expect one to be pretty close to 90deg phase-shifted relative to the other, so a small error in phase measurement can result in a large error in the result. So i'm now going super-paranoid on getting rid of as many dc-blocking high-pass filters as possible.

Now, my soundcard happens to have a pretty high dc offset. This seems to cause some troubles in the measured spectrum (ripples). REW conveniently offfers a high-pass filter to fix that, but it adds a bit of phase shift in the frequencies of interest (5.5deg as 20Hz). I wonder if there is a better way to fix that.

For example, expanding the impulse response window helps quite a bit. But sometimes not sufficiently, it seems.

Here are some pics. This is just loopback, the ripples are not terrible. Real measurements are sometimes more concerning. Looking at the step response, it begs to have the slope subtracted (i.e., offset in impulse response)

hm, how did i not think of this myself! thanks, i'll try!

according to rew, it's just 1.4 over 32767. I think that the full-scale input is about 1 Vpk, so the offset is 40 microvolts, which is way less than i'd expect from even just the opamps i've added to have a balanced input with phantom power. I just didn't expect that such a tiny offset would be so noticeable during measurements.

awesome, thanks!