Spinning now!
Attenuation still at 22,50 kHz instead of 22,05 kHz or is this a typo?
Code:
22 IIR Anti-alias, 352.8 Ksps, 1/8 [B]22k50Hz 15th order[/B]
22 IIR Anti-alias, 384 Ksps, 1/8 22k50Hz 15th order
23 IIR Anti-alias, 352.8 Ksps, 2/8 22k50Hz 15th order
23 IIR Anti-alias, 384 Ksps, 2/8 22k50Hz 15th order
24 IIR Anti-alias, 352.8 Ksps, 3/8 22k50Hz 15th order
24 IIR Anti-alias, 384 Ksps, 3/8 22k50Hz 15th order
25 IIR Anti-alias, 352.8 Ksps, 4/8 22k50Hz 15th order
25 IIR Anti-alias, 384 Ksps, 4/8 22k50Hz 15th order
26 IIR Anti-alias, 352.8 Ksps, 5/8 22k50Hz 15th order
26 IIR Anti-alias, 384 Ksps, 5/8 22k50Hz 15th order
27 IIR Anti-alias, 352.8 Ksps, 6/8 22k50Hz 15th order
27 IIR Anti-alias, 384 Ksps, 6/8 22k50Hz 15th order
28 IIR Anti-alias, 352.8 Ksps, 7/8 22k50Hz 15th order
28 IIR Anti-alias, 384 Ksps, 7/8 22k50Hz 15th order
29 IIR Anti-alias, 352.8 Ksps, 8/8 22k50Hz 15th order
29 IIR Anti-alias, 384 Ksps, 8/8 22k50Hz 15th orderCould you comment on the technical differences between F5 and F7?
OK, so one biquad is a 15:th order filter and you run 8 of them "in series" for each freq hierarchy ?
//
//
Yes it was clipping at 0dB so I corrected but further check its needed.
Thanks. I never listen at 0dB.
No it's just different way of notation 22k50Hz so it's 22(k) thousands 50 Hz
F7 its NOS with one multiplication on first stage which is weaking gain but allow smooth upsampling, second stage is NewNOS with 8 multiplication by 1.
F5 its 8 coefficient upsampling filter which keep signal consistent as NOS but has more smooth and extended magnitude response. I need more time to test all properties of this filter.
No it's just different way of notation 22k50Hz so it's 22(k) thousands 50 Hz
A new one ;-) the "k" or "R" or... usually represent the comma. Tricky to use own style 😉
0R5ohm is never 0,05 ohm...
//
A new one ;-) the "k" or "R" or... usually represent the comma. Tricky to use own style 😉
0R5ohm is never 0,05 ohm...
//
After measurements of earlier filter packs it was recommended to drop those 2 dB to prevent clipping.
If your gain is anywhere near those then this could be a ballpark figure?
P.S. Have been listening to your new 12th order F7 filter this evening (note the description still says 15th). Major Parkinsons "Blackbox" is just awesome on this one... really dark (in a good way!) with lots of detail.
Last edited:
Well, so does 3,5 😉 I agree, I don't understand why but more knowledgeable persons says so. And apparently none of us are correct... "the peaks exceed the maximum and minimum codes by a factor of 1.414 (3.01 dB)" from Intersample Overs in CD Recordings - Benchmark Media Systems the yet some safety is added...
//
//
Yes just Googling for it is like opening a can of worms. Let’s just agree that some amount of headroom in the 3 dB range is a good armor in the loudness wars 🙂
@gumisb I was thinking of your filters as I was re-reading the following:
[1] Archimago's Musings: NOS vs. Digital Filtering DACs: Exploring filtering turned off, implications, fidelity and subjective audibility. (Recent BorderPatrol DAC chatter...)
[2] Archimago's Musings: MUSINGS: More fun with digital filters! Archimago's "Goldilocks" Intermediate Phase suggestion...
#1 pretty much proves why NOS only works well when you've got high res input material. Especially at high frequencies the magnitude and distortion are pretty bad with 44,1 kHz material. Which supports your thought that smooth upsampling is what's needed.
#2 then goes on showing the improvement of first doing upsampling, then feeding it into a NOS filter.
Then it struck me, isn't that exactly what you're trying to do here?
The filter response and impulse response look a lot like yours.
[1] Archimago's Musings: NOS vs. Digital Filtering DACs: Exploring filtering turned off, implications, fidelity and subjective audibility. (Recent BorderPatrol DAC chatter...)
[2] Archimago's Musings: MUSINGS: More fun with digital filters! Archimago's "Goldilocks" Intermediate Phase suggestion...
#1 pretty much proves why NOS only works well when you've got high res input material. Especially at high frequencies the magnitude and distortion are pretty bad with 44,1 kHz material. Which supports your thought that smooth upsampling is what's needed.
#2 then goes on showing the improvement of first doing upsampling, then feeding it into a NOS filter.
Then it struck me, isn't that exactly what you're trying to do here?
The filter response and impulse response look a lot like yours.
Did someone try (-> Sox) with -p45 (->Archimago)? Sorry I am only on ~ #500 with reading...
Yes . I'm aiming for aliasing/imaging free, time and phase coherent filter with flat response from 20 to 20kHz. Current DAC architecture is forcing upsampling so I'm trying to find way how to do it right. There are tons of other solutions for example, external upsamplers, perfect FIR sinc filter with millions of coefficients but somehow they doesn't work for me.
Your resampling parameter "...100:45": Is "45" like the Sox-equivalent on commandline for nearly Linear?