Hugo 2 without mscaler is great. With mscaler which i have the music is breathtaking unlike anything i have ever heard. The million taps are needed to recover the analogue signal in the ADC to better than 16 bit accuracy. Normal dac chips have excessive noise floor modulation harming the end sound.
Interesting. What do you mean by modulation of the noise floor?
I did some digging and in fact Rob is referring to the filter accuracy compared to an ideal infinite filter. Rob Watts claims you need 16 bit accuracy in the Sinc filter compared to an infinite Sinc filter in order to accurately create the original analog waveform at the D to A of a 16 bit digital file.
I donât make the leap that Rob makes. Typical filters used have a passband ripple of 0.002 dB and are therefore 99.977% accurate for in band audio. 16 bit audio has a resolution of (1 - 1/65536) x100 = 99.9998 % but this does not guarentee accuracy to that level - in fact dithering is added so the individual data point at the LSB has a 0.5 bit random noise added. I think that amplitude errors of up to 0.023% arenât likely to be audible - considering typical speaker or headphone linearity.
I think resolution and accuracy are being used interchangeably but they have different meaning and importance. Mathematical accuracy to 99.9998% is achievable in a very long filter but not likely to make a difference over typical 99.977% accuracy because the analog world just isnât that perfectly accurate.
Anyway just curious if you could explain why such small improvements in accuracy might be audible and how it is affecting the noise floor modulation. You obviously hear something and like it!
I am not technically trained and understand only the basics behind chord fpga dac design. What i do know for certain and as someone who played classical violin for 10 years and that is i have auditioned literally dozens of dacs from all variety of manufacturers. Chord dacs win hands down everytime, musicality, timing , timbre , etc and the new mscaler quite literally takes my hugo2 to a whole new level. It uses 740dsp cores in conjunction with the WTA filter. I am totally unbiased just what i hear. No wonder they won best dac in every category what hifi and eisa awards.
I have a DAVE and am eagerly the M-scaler which I pre-ordered. I was not aware they were shipping yet. Do you have a pre-release version?
no they are shipping in the uk for some weeks now. i have a silver mscaler and hope to upgrade hugo2 to tt2!!
Iâm very happy with my current setup with Chord DAVE, but Iâm also very curious about the M-scaler. Itâs a new year with new possibilities and now Iâm considering upgrading or at least seriously demoing the unit.
Did you get the M-scaler and how did it affect the overall performance of your system?
And you have had the M-scaler for quite some time now. How do you perceive the difference between having it and not? (Also, it seems really easy to A/B test with just a push of a button. Have you tried that?)
First, let me say that I admire Chordâs work and Iâm very happy with my Hugo 2.
But I saw that RMAF lecture and I wish Rob Watts wouldnât talk about 300 dB.
I remember from my university days, a bunch of engineering students, the conversation was flagging, until somebody said â300 dB!â and we all fell down laughing and had another beer and the party continued. Geek humor. But why is 300 dB a good joke? Because it is too much. There arenât 300 dB.
If you count all the stars in the galaxy, exactly, thatâs only 110 dB.
If you count all the stars in the observable universe, thatâs only 220 dB.
If you can estimate the mass of the planet to the nearest kg, thatâs 240 dB.
If you have a typical current in electronics, one milliamp, and count the individual electrons flowing through and their speed, ok that would be 350 dB, but thatâs impossible under Heisenbergâs uncertainty principle â physical reality does not have 350 dB precision.
So 300 dB is a good joke. And that ruined the rest of Wattsâs lecture for me.
Ok, if you listen carefully, he doesnât really claim that his devices measure at 300 dB SNR. I think he means that when he builds a design for which the math shows a 300 dB precision, he can hear the difference from a 240 dB design. And that is not a crazy statement. If indeed that is what he meant. But I wish he would say so clearly.
We donât need geek jokes.
2 Likes
See below link.
Correct, heâs never claimed that anywhere.
Some additional info/discussion/background:
https://www.head-fi.org/threads/watts-up.800264/page-73#post-14507341
Yikes. Not sure where all these numbers are coming from.
Benchmark:
https://audiophilestyle.com/forums/topic/46501-chords-new-m-scaler/?do=findComment&comment=883447
https://audiophilestyle.com/forums/topic/46501-chords-new-m-scaler/?do=findComment&comment=883465
https://audiophilestyle.com/forums/topic/46501-chords-new-m-scaler/?do=findComment&comment=883466
https://audiophilestyle.com/forums/topic/46501-chords-new-m-scaler/?do=findComment&comment=883511
See the conceptual upsampling section:
https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing
Chord DACs (current production) definitely up-sample beyond PCM768kHz:
Yes it is conceptual upsampling. They downsample to 211KHz for the converter chip. It allows very small timing adjustments to be done. They keep adjustments to below 1 Hz so as not to create audible jitter from timing adjustments.
Exactly. It doesnât do much to his credibility. To be fair, the maker of HQPlayer does the same thing. Extolling HQplayer filters that have 200+ dB of stop band attenuation. The claim is that HQPlayer filtering is better than any DAC because of this. The claim is that all DACs are inadequate because they donât have enough computer power. Rather dubious claims over levels of extreme digital accuracy that just arenât needed or achievable or applicable in the real analog electronics world.
Lots of angels dancing on the head of a pin these daysâŚ
One of the best DACs I had long ago was a Chord DAC64. Used with its long tapping filter, it was far more realistic than others of the same vintage (2001-2002).
I happen to have happily switched to the âClosed Formâ of HQ-player to drive my Neutron Star la Rosita DAC (384 kHz/32 bits float input on USB). It provides better ambiance retrieval and retains more natural transients.
I prefer it over the previously used polysinc-mp. With all indulgence for the long discussion I had on this community about linear phase versus minimum phase filters.
However I have a question for HQplayer software guys. How many taps does it implement and how does it mitigate with compute power (like I use a MacBook Air, to stay in battery supply and with 30 cm USB, which does have an impact on sound - I assume this is about jitter).
Letâs put some physics in terms of power on this 300 dB number.
Absolute dBs are relative to 10-12 Watts per square meters - 0 dB
120 dB is 10 power 120/10 x this = 1 Watt/ m2
300 dB is 10 power 300/10 x 10-12= 10^18 Watts/m2 = 1 Billion Watts/m2.
The sun radiates as a whole about 3.8610^26 Watts (Source Wikipedia solar core).
So 300 dB is the acoustic level equivalent to the energy of 3.86^10^(26-18) = 3.8610^8 parts of the whole sun⌠One billionth of the sun itself - 0.34% of the Earth volume of pure nuclear fusion, with its whole radiation focused only over a square m to produce a gigantic sound waveâŚif any air would still exist at this stage⌠Not in anyoneâs wildest nightmare.
This is a good illustration of how away from physical reality can be some numbers issued from digital processing theory. Keep feet on Earth, geeks!
2 Likes
I found another fun physics example:
What is the quietest room possible?
Well, itâs got air in it, and air consists of molecules that move around due to the temperature. And a lot of them hit the ear drum all the time. So there is the constant hiss, like rain. This is the âself noiseâ of air.
I found somebody on the internet who had determined that the self noise of air is about -24 dB SPL, 24 dB below the quietest sound we can hear.
The loudest sound that makes sense in audio is the maximum rock concert SPL, about 120-130 dB, before you get deaf.
So the total dynamic range between the hiss of molecules hitting the ear drum and the loudest sound is 150 dB.
300 dB is 150 dB more, or 10^15 times more.
The ear canal is about one cubic cm, and air contains 10^16 molecules per cubic cm. So imagine we remove all but ten of them. The self noise gets 150 dB quieter. Now we have a dynamic range of 300 dB. Except this is about the best ultra high vacuum achievable in the lab. And you couldnât hear anything.
1 Like