Great thread. Just startet "reading" up on the subject
Good information in the first post and more information here
Filters and You: FIR Filters - Tuts+ Music & Audio Tutorial
BTW: Is "rephase" THE tool to use?
Can I use for instance audacity to see the impulse result from "rephase" simulation?
For the record..
Totally newbie in this.. But hopefully a quick learner..
Good information in the first post and more information here
Filters and You: FIR Filters - Tuts+ Music & Audio Tutorial
BTW: Is "rephase" THE tool to use?
Can I use for instance audacity to see the impulse result from "rephase" simulation?
For the record..
Totally newbie in this.. But hopefully a quick learner..
Last edited:
You need something that will output filter coefficients in .txt format. rePhase is probably the easiest most accessible tool that does that.
TNT's friend looks to be using something like MATLAB or Octave. MATLAB has a filter design plugin which unfortunately hasn't been ported to Octave, but Octave can be used if you know the parameters required.
http://vdl2-ase.wikispaces.com/HowTo+-+FIR+filter+design+in+GNU+Octave
There are commercial tools like ScopeFIR if you want to outlay a couple of hundred $US:
ScopeFIR: FIR Filter Design Software for Windows | Iowegian International
cheers
Paul
"Nasty" is such an emotive term.
It's worth having a read of Doug Rife's paper on Oversampling Theory (link in the first post). There is an interesting section "Upsampling ameriolates differential non-linearity".
Rife concludes:
For context, in the preceding section Rife notes that
So it could well be that the predilection for sharp roll-off filters contributes in part to the harshness that some have reported. It would be interesting to explore Rife's theory and see how it can be applied in practice.
His observations about the filtering effect of air is worth taking into account, as there will be differences between headphones, close monitoring, and sitting 3 metres from the speakers. It's worthwhile having a play with the calculator at http://www.sengpielaudio.com/calculator-air.htm
At 25°C and 50% humidity, the attenuation per metre is:
10kHz : 0.147dB/m
15kHz : 0.306dB/m
20kHz : 0.502dB/m
25kHz : 0.717dB/m
30kHz : 0.937dB/m
With headphones the attenuation will essentially be 0dB across the spectrum, near-field monitoring (say 1 metre) a slight roll off, but by 3 metres the effect becomes far more significant with a -2.8dB roll-off at 30kHz.
cheers
Paul
Last edited:
Any one as any idea what kind of filter is dCS using in their famous R-2R DAC?
See this link http://audiofast.pl/main.asp?idm=1&ids=240&idsu=635&wersja=1
There is twonice White Papers
-''A Suggested Explanation For (Some Of The) Audible Differences Between High Sample Rate and Conventional Sample Rate Audio Material''
-''Effects in High Sample Rate Audio Material''
where they discuss filter types and their perceive sound. May be useful as well...
See this link http://audiofast.pl/main.asp?idm=1&ids=240&idsu=635&wersja=1
There is twonice White Papers
-''A Suggested Explanation For (Some Of The) Audible Differences Between High Sample Rate and Conventional Sample Rate Audio Material''
-''Effects in High Sample Rate Audio Material''
where they discuss filter types and their perceive sound. May be useful as well...
Last edited:
Page 19: http://www.dcsltd.co.uk/wp-content/uploads/2014/03/Vivaldi-DAC-Manual-v1_0x.pdf
They recommend filter types per sample rate in a quite unexpected way (for me). If correct, it may invalidate our assumption of developing a decent 44.1kHz filter and porting it to higher sample rates.
"Effects" paper (http://audiofast.pl/pdfs/effects.pdf) may be related to this...
I stepped back from doing greater than 44.1kHz versions of the filters because the requirements are significantly different. With 44.1kHz and 48kHz the narrow band between 20kHz and fs/2 or Nyquist (22050Hz and 24000Hz respectively) make filter design a matter of balancing trade-offs and compromises. Once you move to 96kHz, Nyquist is at 48000Hz so you have up to 28kHz of bandwidth in which to attenuate. This makes filter pre and post ringing much less of an issue.
I'd be a bit cautious about reading too much into the effects paper. The article looks at the complete chain of ADC and DAC using analogue tape as a source, and the authors were manipulating filters on both the ADC and DAC in the tests.
I'd also note the dcs paper is from 1997. Much of the move towards slow roll-off filters for 44.1kHz has occurred since the mid-2000's, so keep in mind that Mike Story's work predates these later developments.
Last edited:
OK, I have the NOS "filtering bypass" installed. Too be honest it's not anywhere as bad as the spectrum would suggest.
The sound is thicker, and possibly little more ponderous than the other filters. There is no sense of the shoutiness on the Lydon vocals on Album. There are perhaps hints of a slightly metallic haze on the voice in the sections that were previously shouty.
On Miles' "So What", it actually sounds impressively natural. There is a nice level of detail - you can really get a sense of the drummer using brushes for example.
Despite what I've said previously I wouldn't dismiss this NOS bypass without a decent listen. If nothing else it's a salutary lesson in what the DAM1021 can do without any filtering
cheers
Paul
"Nasty" is such an emotive term.
It's worth having a read of Doug Rife's paper on Oversampling Theory (link in the first post). There is an interesting section "Upsampling ameriolates differential non-linearity".
These statements are highly objectionable.
Mirror image = uncorrelated spectrum? Far from the truth...
Dither is random and has completely uncorrelated spectrum. The folded image is not.
Doug Rife makes wrong assumptions that foded images have de-correlating effects (like dithering which decorrelates quantization errors).
Upsampling has many benefits. The folded images of slow roll-off filters are none of them... The "nasty" artifacts indeed.
Good ! I believe because the room and time delays with walls interactions : NOS is not the same when listen to with headphones vs Floorstanders. The room have to help a little in the highs with NOS acting like a pad of interface.
Certainly a mixed way to find between NOS and FIR filter like TotalDac made ? This is not to said than upsampling can be bad but it seems to involve a huge work to find something which sounds pleasant !
Paul, did you find a better dynamic with NOS ? Especially in the low end and medium ? Cant it rock now ? (bass string, percussions, drums ?)
The "brickwall" filter I posted has excellent phase coherence. The signal spend more time in the filter but low and high frequencies come out with the same time relation that they got in to it. You have to wait a few extra milliseconds before you can hear the song after you hit play - thats the drawback in the time department
//