Time scale for the impulse response?
Hi @jussi_laako, regarding on the impulse response, does it means with less ring(pre and post), the peak of the impulse will go up? I know MQA filter tends to be more noiser due to āleakyā design, which allow ultrasonic noise get reflected back into audio range. In your measurements, how much noise will you anticipated? I know HQPlayer have a MQA like filter but with lower noise?
Modern digital filters has kept aliasing out of bay since over-sampling digital filters came into use, I donāt understand why MQA has to break the fundamental rule of allowing aliasing to exist again and this has certainly spark another round of debate.
Is that one of the filters reverse engineered from the Dragonfly?
It is for 8x filter, in samples (37 taps), so for 705.6/768 kHz output rate (assuming 88.2/96k unfold).
There are two factors to this. One is the anti-aliasing decimation filter they use for the encoding side, to reduce the input rate to 88.2/96k (plus there could be leaks on the folding band-splitting). This leak is notable for some cases like the one Stereophile published where modulator noise is aliased back to audio band (among possibly other stuff).
However, this likely different filter than they use for upsampling at the playback side, one of which is presented above. This will leak ultrasonic images and some of that may alias back to audio band due to intermodulation in analog electronics.
So thereās no simple answer to the question, because there are multiple factors in play.
Yes, it is very short, slow roll-off one, available as linear- and minimum-phase variants. But is is still quite a bit more aggressive than the ones by MQA:
But much less attenuation and not nearly as aggressive as the other filters.
1 Like
What I meant was, what is the scale of that diagram? The X axis goes to 40, the main pulse is 10 wide. 10 what?
Samples/taps. In this case for 8x conversion ratio.
So impulse for the above 705.6k output case it is 1 / 705600 * 37 seconds long, which is 52.4 Āµs. When source is non-unfoldable like the 2L piano album at 48 kHz, it is 1 / 384000 * 37 seconds wide, which is 96.4 Āµs.
So the range for length of that filter in time, depending on situation is 104.9 Āµs - 48.2 Āµs.
To me MQA titles sound so much better than non-MQA titles.
Iām using Tidal (Desktop).
1 Like
So when will Roon start doing the MQA unfolding?
No release date has been announced. All we know is that it has a high priority, commercial agreement has been reached and development is underway. Testing is an iterative process; Build, test, Build, test etc. It takes as long as it takes to do it right.
2 Likes
Excellent because I really want to give 'em my money for a subscription.
A paper from Ayre about digital filtersā¦
https://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf
Coincidentally, Peter Craven is also mentioned here, he helps to develop MQA. Though this concept of so called ādeblurringā is not new, it has been used in Ayre products for many years.
1 Like
Indeedāit has occurred to me that from the audiophile perspective (that is, what we see an hear), thatās been the trend in digital audio for the last 10, 15 years: from focusing on frequency perfection toward more awareness of the time domain.
Yes ā I first came across this in a Meridian product, with a totally unexpected effect. At the time, I had a Meridian G-Series box that pre-dated the invention of apodizing filters (in this context). The G didnāt do HDMI switching, so I bought Meridianās external HDMI switch box, the HD621. In addition to switching HDMI, it also stripped off the audio and fed that to the G box. I had my Oppo disc player connected to the G with SPDIF for CD, and now connected it through HDMI for DVD. But in addition to stripping off the audio, the device also upsampled the CD sound with an apodizing filter. Both the G and the Oppo could upsample with regular filters.
To my very great surprise, this box that was installed for convenience when playing video significantly improved the sound from CDs.
I had been aware of the importance of time-domain stuff since forever, but this effect was so significant that it stunned me.
Today, most of my zones do apodizing, or at least Minimum Phase.
3 Likes
I have played back local MQA files on Audirvana and seen 192
It only does the first unfold up to max 88.2/96kHz. Unless your DAC has built-in a MQA renderer, then you will see higher.
It might display 192k as the āoriginal sample rateā. But I donāt have Audirvana.
If it display at the DAC side is 192k, then it may be fully unfold. I never tried Audirvana but it should behave like Tidal desktop app. In Tidal desktop app, I never get beyond 88.2/96kHz on my DAC even the album is originally mastered say 192kHz or higher.
True, but AFAIK itās only the software or device which does the first unfold which can authenticate the file/stream and display the original sample rate.
The display of the original sample rate you are referring to is on the software level; something like Roon are doing right now, kind of detection. As far as external DAC is concerned, I donāt see Tidal app outputting anything beyond 88.2/96kHz even its authenticity is true. This authentication really puzzled meā¦ If it is not doing anything but showing the original sample rate, I guessed it is quite misleading. Put it another way, If the embedded information inside the file is corrupted basically no decoding is possible.