On a 2'nd thought looks like I paralleled the additional caps on 47uF caps (CE5, CE6, CE16, CE18) instead of 100uF caps (CE3, CE4, CE14, CE19). The 2x47uF caps for each DAC are for VcomL and VcomR reference voltage from inside PCM1795 DAC and there's no need to adjust these 2 caps.
So, I'm redoing this job now by paralleling existing 4 x 100uF/35V caps (2 per PCM1795) with additional 4 x 100uF/35V Nichicon KA and 4 x 10uF/25V tantalums.
As you can see, the +5V ripple+noise measured directly on the very small SMD capacitor (2mm nearby pin 23 of PCM1795) decreased from 1.4mV p-p (see my last post) down to about 1mV p-p (most of the noise is actually below 0.9mV).
I don't think there's something else I can do to this +5V signal to make it look better on my scope, unless installing 470uF caps (like M2Tech Young or Vaughan) in parallel with existing 100uF caps (CE3, CE4, CE14, CE19) and check if noise improves. Most likely ripple+noise will decrease a bit, perhaps somewhere around 0.8-0.9mV, but unfortunately I don't have right now at home decent low ESR 470uF caps, but I'll probably give it a try sometime.
BTW, I just measured ripple+noise of my ODAC, directly on ESS9023 +3.5V pin and it's around 4mV p-p, no matter what USB port I'm using on my Thinkpad laptop. So, I believe getting about 1mV for my Essence One PCM1795 chip might be OK after all...I don't know.