adyc
100+ Head-Fier
- Joined
- May 8, 2016
- Posts
- 146
- Likes
- 73
Thanks. That is very useful. I'm looking forward to my Blu2.
Mr. Watts. I can't wait to receive my Blu II and hear the fruits of your years of hard work for myself.
If you had a moment, I had a simple question that has been nagging me for some time. I appreciate it is a bit of a circular argument, however, it deals with the colourations of electronics or certain headphones namely, if you have been adjusting the code and the electronics based on what you hear through a handful of headphones or speakers (and with a handful of music titles) how can you be certain that the improvement will be heard for other material or systems? Also, if prior ADCs were less capable of capturing the waveform than Davina will be, will the benefits of being optimized for the other ADCs translate to the better recordings with a more capable ADC?
As I have enjoyed reading about your design process and how you have approached the problem, and you seem to be a perfectionist, how can you be sure that the final product is ready and another tweak or two of code or electronics won't offer a better result still? I suppose that is the challenge every engineer must deal with, just wondering how you approach it really
Much appreciated,
Mike
This is actually a profoundly important question; take a look at my blog Watts up? and I discuss the problems of listening tests, and the search for getting to the truth.
The reality is that there is no absolute reference system; we have to deal with what we have. Of course, I have a reference target - that of live un-amplified acoustic music - and the goal for reproduced audio is to be able to get that kind of sound quality. For me at least, the M scaler represents a big leap forward to that goal.
Given that we do not objectively know the qualities of the whole chain, making subjective assessments based on a considerable amount of uncertainty is a dangerous game; and for sure I have made mistakes in the past where a slightly brighter sound was assessed as being more transparent when in fact it was not - it turned out to be a tiny amount of noise floor modulation. Because small changes are subject to error, one must constantly look back, and re-confirm original tests, and re-asses the technical understanding for the explanation of why something is better.
Having said all that, given we do not know the total audio chain for sure, we can have some certainty. My listening tests are often based on extracting a distortion or error, and then making that error smaller; and you can measure that error either thru direct measurement or simulation; and my job is too reduce the error so that one can no longer hear the change once it gets smaller. When you do a listening test and get a null result (can't hear a difference) then that's certainty; there is no question of interpretation. Then from an engineering POV you just make it say ten times smaller than this level and you are done. A null test means that the quality of the recording chain is not important in an absolute sense. Of course, there are still assumptions - the observer being a very sensitive listener for one (fortunately I have not met someone with better hearing than me but I guess old age beckons) and if the quality of the system was poor then it becomes more difficult to hear any kind of changes. So I repeat null tests, and so far I have not had a past result overturned.
There are also other factors that can give you some certainty; and that is variability. When one hears live un-amplified music I am struck by the extreme variability in the sound quality - timbre with bright instruments being very bright, and dark ones being very dark - sound-stage depth in real life is incredible, one can truly detect depth to fantastic precision - something that reproduced audio fails to do (oddly with the M scaler is very much deeper depth). So when doing tests, if the variability is better, then we know for sure that a fundamental improvement has been made.
Your question about ADC's is pertinent and I don't have the answer to that - it's one of the things I want to fine out. So far, I have found that a WTA change to the algorithm is best for all recordings - but some recordings show bigger changes to WTA tweaks than others, but when I made improvements all recordings benefited. The reality of course is that the better optimized it is, the closer we are getting to the original signal in the ADC before it was sampled.
As to your final question - no I can never be sure that the code or design is optimal. But I spend a lot of time optimizing before release; and I only release once I have hit brick walls, such as capacity of FPGA. But of course, I am always knowledge limited, but getting major insights into things takes years not weeks - I have been doing this for a long time.
Of course, Davina will provide a reproduction chain that I will have total control over and this I am sure will give benefits. Moreover, I will know for sure how many taps are needed to perfectly recover the signal from a subjective point of view; and that will be immensely valuable.
Also I have come up with a technical measurement, so I will be able to show exactly how much the decimation > interpolation process creates in terms of transient timing errors. This will give numbers for the transient timing error and may enable better optimization of the WTA algorithm. I hope also that it will also prove that what I have been prattling on about for the past 36 years is measurable directly.
Rob
Ordered Blu II (in U.K.) should be three to four weeks.
Hi. I believe that they are now shipping all preordered Blu2. My audio dealer placed a preorder for me the first week of March (if I remember well). Chord has just shipped it to him today.
Oscar
A query about Blu 2's input priorities and how they work. There are three inputs: CD, USB and BNC. I propose to use all three, with my streamer's digital out connected to BNC - simply to use Spotify. So far as I can tell from the manual, there's no way of switching between the three sources on either the unit or the remote, as one can on the DAVE, and they operate according to a system of priorities: CD always has priority, next USB if there's no CD input - presumably this means no CD playing - and finally BNC where there's no CD or USB input - again, I assume no CD playing and no audio files playing via my laptop. If I'm using Spotify via BNC and I want to play an audio file via USB, is it simply a matter of stopping or pausing playback on Spotify and starting USB playback? I appreciate that I won't know for sure until my Blu 2 arrives, but a heads-up would be helpful.
A query about Blu 2's input priorities and how they work. There are three inputs: CD, USB and BNC. I propose to use all three, with my streamer's digital out connected to BNC - simply to use Spotify. So far as I can tell from the manual, there's no way of switching between the three sources on either the unit or the remote, as one can on the DAVE, and they operate according to a system of priorities: CD always has priority, next USB if there's no CD input - presumably this means no CD playing - and finally BNC where there's no CD or USB input - again, I assume no CD playing and no audio files playing via my laptop. If I'm using Spotify via BNC and I want to play an audio file via USB, is it simply a matter of stopping or pausing playback on Spotify and starting USB playback? I appreciate that I won't know for sure until my Blu 2 arrives, but a heads-up would be helpful.