[Rob Watts]

The new prototypes get used - and modern recordings benefit as much as older ones - perhaps even more so with recent recordings.

 

[Reactcore]

perhaps even more so with recent recordings.

Is the WTA algorythm itself much changed vs the existing Mscaler aside from more taps?

Or is it mainly different tuning of the 5 variables

 

[thePhones]

Just got myself the Wiim Pro streamer. The nice thing about it is that it also has an optical in, so I can aply it’s eq to incoming signal from my tv. So I finally have the ability to eq video games which are sometimes still tuned for “small tv speaker” sound. The included eq has some detail loss but it was not too bad. Today we played an adventure/riddle game where sync did not matter that much and we activated the full million taps. The difference was huge, voices and music sounded so lovely. Is there still a way of also improving the short delay filter to get better timbre variation or is the technical limit reached because of the short delay?
With Nvidia Shield movies are covered, but with video games you mostly need the short “video filter”.

 

[Rob Watts]

Unfortunately, we can't escape the maths behind sampling theory - shorter latency will always degrade transient reconstruction. Only way around it is to have games output audio at 768k, and that's not going to happen! Low latency is also an issue with studios, so it's an issue I am giving some attention too.

 

[thePhones]

Unfortunately, we can't escape the maths behind sampling theory - shorter latency will always degrade transient reconstruction. Only way around it is to have games output audio at 768k, and that's not going to happen! Low latency is also an issue with studios, so it's an issue I am giving some attention too.

Nice to hear you are trying to get the best out of it. Forgive me if I am asking a stupid question but why isn’t it possible to just look a few more seconds into the past while the stream is coming in and thereby adjusting the present output signal. I know you often said how important pre-ringing is to get it mathematically right, but wouldn’t it be a compromise to at least “work out” the past? Do you always need the same amount of pre- and post- ringing regardless of filter type to get an audible improvement?

Btw, if the release of the new scaler should fall on the start of February I would like to take a look in Hamburg, here: http://www.hifitage.de/
Maybe it is a bit to small for Chord because it is “just” a very big Hotel but it has the advantage that only 1-3 exhibitors (depending on size requirement) have each separat Hotel rooms so it is a lot better for listening tests than a noisy open area. Also I would love to be the first listening to the tech behind it on one of your presentations

 

[Rob Watts]

Asymmetric FIR (that is short ringing on the "future" to give you short latency) and long post ringing certainly is better than just short pre and post. The video mode on the Hugo M scaler does this; I originally listened to symmetric vs asymmetric with video mode and the asymmetric sounded a lot better. With video mode you are getting the full M scaler taps for post ringing; its just the pre ringing that is being adjusted.

 

[audionewbi]

There were few discussions about potential of AI earlier on within this thread, it seems it is slowly getting some reality.

https://www.google.com/amp/s/www.ma...mpling-audio-to-incredible-48khz-quality/?amp

 

[thePhones]

because the ear/brain is adept at accommodating frequency response changes - the brain measures the environment acoustic and compensates for it. If you are walking around with a chatting friend, and go from a acoustically dead room, to a live room, to then the outside, you don't suddenly think your friend's voice has been degraded or sounds different - the brain just deals with radically different acoustics

When reading older posts I found this one and it reminded me of a funny story with a friend, I thought you guys might enjoy.

A few years ago, I had just treated my room with foam panels to the point where even an audio engineer claimed it was almost “studio quality dry”. So a friend came over for a chat and when going from the hallway to the living room she immediately startet shouting. She asked why we were laughing. When we told her that she was shouting, she said: ”I can’t hear myself talking anymore! What is happening?”. That was when I knew I had enough panels installed

So I guess brains are sometimes different when it comes to compensating for acoustic changes.

 

[Kentajalli]

When reading older posts I found this one and it reminded me of a funny story with a friend, I thought you guys might enjoy.

A few years ago, I had just treated my room with foam panels to the point where even an audio engineer claimed it was almost “studio quality dry”. So a friend came over for a chat and when going from the hallway to the living room she immediately startet shouting. She asked why we were laughing. When we told her that she was shouting, she said: ”I can’t hear myself talking anymore! What is happening?”. That was when I knew I had enough panels installed

So I guess brains are sometimes different when it comes to compensating for acoustic changes.

Good one!
When I talk, my intention is for others to hear me.
Oh well, it takes all sorts .

 

[GuiltyRocker]

When reading older posts I found this one and it reminded me of a funny story with a friend, I thought you guys might enjoy.

A few years ago, I had just treated my room with foam panels to the point where even an audio engineer claimed it was almost “studio quality dry”. So a friend came over for a chat and when going from the hallway to the living room she immediately startet shouting. She asked why we were laughing. When we told her that she was shouting, she said: ”I can’t hear myself talking anymore! What is happening?”. That was when I knew I had enough panels installed

So I guess brains are sometimes different when it comes to compensating for acoustic changes.

Very interesting.

 

[LisaShayBrown]

When reading older posts I found this one and it reminded me of a funny story with a friend, I thought you guys might enjoy.

A few years ago, I had just treated my room with foam panels to the point where even an audio engineer claimed it was almost “studio quality dry”. So a friend came over for a chat and when going from the hallway to the living room she immediately startet shouting. She asked why we were laughing. When we told her that she was shouting, she said: ”I can’t hear myself talking anymore! What is happening?”. That was when I knew I had enough panels installed

So I guess brains are sometimes different when it comes to compensating for acoustic changes.

Be careful of over-foaming, unless you really know what you're doing. You'll suck up all the mids/highs, and end up with an excess of low mids and bass.

 

[Jawed]

@Rob Watts What do you think about using square waves as test signals for upsampling?

Because a band-limited square wave is defined as the sum of a finite set of fundamental + harmonics, the upsampled square wave can be directly compared with the native square wave at the higher sample rate.

The timing of transients error should be easy to measure in this situation, since the reference square wave is perfect, subject to quantisation noise or precision (64-bit or better).

 

[Rob Watts]

Except that square waves do not have transients, since they are composed entirely from fixed steady state sine waves. Putting a digital square wave into a perfect DAC will give the results of an ideal bandwidth limited square wave - which all of my DACs do with ease - at least within the constraints set by the WTA filter's >FS/2 stop band attenuation. The result of bandwidth limiting a square wave is reduced rise time plus Gibbs phenomena. You can read about it here.

Since it's a steady state signal, it won't suffer from reconstruction timing errors, as there are no transients in-between adjacent samples on the digital input.

 

[tlcocks]

Hi Rob Watts,
I have recently entered into a discussion about DEQX 64 bit floating point vs your proprietary 104 bit algorithm on the Mojo2. I would respectfully request your commentary on his paragraph pasted below. If this topic has been addressed earlier by you on head fi, please point me there.
Thanks Rob! Happy Mojo 2 owner

“Let's assume 104 bits is correct. The DEQX has a 64 bit floating point processor. Regular processors like the one that is in your Mojo can only work in integers. Floating point processors can calculate down to infinitely small fractions. This gives them much higher accuracy and with DSP a much higher dynamic range. The higher dynamic range is essential in DSP to maintain decent resolution at low volumes and to prevent boosting filters from clipping. They are also a lot more expensive and run hot, too hot for a small unit like the Mojo. Against any floating point processor the additional bits in the Mojo do not mean much as the formats it is working with are either 32, 24 or 16 bits as it can only generate a fixed number of values, whereas the floating point processor can generate an infinite number of values. ”

 

[Rob Watts]

The problem with 64 bit floating point is that as the signal gets smaller, the resolution changes, and this creates noise floor modulation. OK so we are talking about over 300dB of innate resolution, but this is subjectively significant noise floor modulation and is audible. Moreover, as a signal disappears into the noise floor, it will be treated differently whether there are larger signals present or not. Now you may argue that these errors are very small but at the end of the day, it's about sound quality, and to me these errors are very significant subjectively. And you only need to look at the Mojo 2 thread to see the very positive comments about the EQ compared to traditional 64b FP EQ.

Additionally, a very important feature is the EQ running at 705/768 or 16FS with all the internal nodes being noise shaped. Without the noise shaping, I would need much more bit depth than 104 bits. This could turn into a significant design problem for me when doing EQ for pro audio. If they process at 192k or greater, I can use noise shaping; but doing EQ completely transparently at 44.1k could be a major headache. Incidently, the benefit of noise shaping the internal truncation errors is that the filter still functions for signals well below the bit depth of the system, as errors are never lost - just re-cycled.