Are those measurements from the Metrum or the ULN-2?
Metrum Acoustics Octave - Page 13
Gear mentioned in this thread:
Originally Posted by Rob McCance
If that much ringing is already encoded into every 44/16 then decoding would be an impossible nightmare, and that's not the case.
Miska is correct on that one.
All 44.1kHz recordings have been subjected to anti-aliasing filtering during production, be it in the ADC or in the DAW (when coming from a higher sample rate).
There are two main AA filter categories: minimum phase and linear phase.
The former can exist as a high-order analogue filter in front of the sampling stage, or as an IIR digital filter operating on a stream at higher sample rate than the target.
The latter can only exist as a FIR digital filter.
The former adds phase distortion and post-ringing at the transition frequence. The latter adds pre- and post-ringing.
Analogue minimum phase AA filters were the norm in the early years of digital music recording, with convertors like the Sony PCM1630 in use, perhaps well into the nineties.
From the early nineties on delta-sigma ADCs pushed other architectures away, and these invariably used FIR linear phase filtering (copying exactly what the reconstruction/oversampling filters of delta-sigma DACs were doing).
Sample rate conversion software dominantly is FIR/linear phase, although since the past six years or so IIR, or FIR/minimum phase is available, too.
All of these imprint their impulse response, with its ringing, on the output data.
And a NOS DAC reproduces this ringing. It has to.
As Miska already suggested, the only way of getting rid of the innate Fs/2 ringing in a 44.1kHz signal would be to suppress it by filtering with a shallower filter (i.e. less ringy, but still ringy) starting much earlier in the audio band (e.g. at 18kHz) and reaching significant attenuation well before 22kHz.
That's what I think, usually popular upsampling rate is 8x, but don't ask me what software is best to do this as I'm not familiar.
NOS dacs are well known to roll off early in the HF, I dont know why its a surprise tbh. minimum phase apodising filters are IMO the best currently available. this is how the oversampling filter in puremusic works, as well as the analogue filters in some wolfsons dacs having it as an optional filter type.
there is ringing already in the signal from the Deltasigma ADC stage that exists in ALL studio mastering equipment, i'm unaware of anything but DS ADCs available in the high end or otherwise. if you then add the wrong filter in the DAC analogue filter you can get it adding together and causing aliasing.
the metrum is good in this aspect, very good in fact, reproducing test signals its square wave response is phenomenal, but reproducing actual music recordings the ringing is still there from the ADC stage, they have the advantage of not adding their own phase error in this regard, but its a tradeoff, nos has other issues like the rolloff and also out of band noise depending on the design, treating your ears as the low pass filter. I dont like this last one, as in my experience when using wideband amps like I do, this HF junk can be folded down into the signal through room modes. you can of course get rid of that with an passive filter, but my gear is mostly DC coupled for a reason and I prefer not to use more caps than needed.
i'm looking forward to hearing the metrum at the meet next weekend, but there are many reasons it could never be part of my system.
I prefer deltasigma when done well and using apodising OS filters prior to sending the signal to the dac.
you are mistaking upsampling for oversampling Redbull, how could you upsample 44.1 x 8 (352.x kHz) and then send to the metrum when its inputs are limited to 96khz? apparently 192 works with some gear, but last I read the inputs were only rated at 96.
Edited by qusp - 8/11/12 at 12:56am
Its all correct what Miska is saying (presumably this is gleaned from CA) but for one thing. The ringing of the filter is only going to be present if its excited. A bell only rings when its struck, so its a tad misleading to say that the ringing is 'innate' in a 44k1 signal. If there's no content near the ringing frequency then it won't appear.
ok perhaps ringing was a badly chosen term on my part, I followed suit, what i'm talking about is the inherent aliasing from the SD convertors.
haha I just did a search to see if I could find some scope shots to illustrate and came upon the thread you are talking about at CA just now, yes erm gleaned is the word. for that matter my post reads like it could even be from that thread, many of the same issues are talked about there by the looks of it.
Edited by qusp - 8/11/12 at 9:53am
Oops, my bad, never consider the max allowed input of Metrum.
Yes I still confuse between upsampling and oversampling. What's the difference?
So then the best possible upsampling is 2x?
- 2,261 Posts. Joined 12/2009
- Location: Poland
- Select All Posts By This User
If you have 44.1 files than no problem with 4x upsampling as Metrum works well with 176.4 and works iwth 192 for some (for me it works with AP2).
- 59 Posts. Joined 1/2011
- Location: Australia
- Select All Posts By This User
In the manual for my octave it says there's a low pass filter in it.
Low pass filter: -3 db first order 65khz.
- 13,956 Posts. Joined 1/2008
- Location: An Australian living in Fukuoka, Japan.
- Select All Posts By This User
I was measuring the Octave using the ULN-2 by the way. The mic inputs supposedly have a -130 dB noise floor, however either RMAA is screwed or the ULN-2 has other issues as some of the measurements give weird results half the time.
Up-sampling has a target frequency. Oversampling is simply a multiple. They are essentially the same thing.
Some of the music players available for Mac OS X have iZotope or their own proprietary up-sampling software built in. Using the 2x/4x setting and/or setting the target at 176k works with the Octave via coax.
Something I might have to try is switching my Ref 7.1 to NOS mode and seeing if it can produce a square wave like the Octave.
So upsampling can result to any expected sampling rate, like 44.1 => 96 KHz while Oversampling can only multiply to it's own sampling 'family' like 44.1 => 88.2, 176.4?
I would love to see Ref. 7.1 square wave on NOS mode Curra.
How do you upsample your files? Is Foobar upsample good enough?