Jump to content


  • Content Count

  • Joined

  • Last visited

1 Follower

About opus101

  • Rank
    Sophomoric Member

Personal Information

  • Location
    Hangzhou, PRC

Recent Profile Visitors

2542 profile views
  1. The review contains a few typos, mainly in numbering of various ICs. The most misleading one in my view is in saying it uses 'OPA86 opamp in its I/V'. A quick check on the manufacturer's website shows that it uses 'OPA861' which most certainly isn't an opamp, rather a 'super transistor' or 'OTA' designed to operate open-loop.
  2. Dead easy to localize on many classical recordings I have. That's from the hit though which extends way beyond 150Hz, not from its fundamental.
  3. Its clear to me his interpretations of the FFTs he posts are sometimes too simplistic. For example when a stimulus with more than one tone is used there are often many spurious tones arising from non-linearity in a device under test. He takes no care to take them all into account when (in my view, casually) stating that 'the dynamic range is X' or 'Y equivalent number of bits'. He simply takes the level of the peak spurious tone and quotes the distance in dB between that and the stimulus tone(s). Someone who had a deeper understanding of FFTs than he does would make an attempt to sum all the spurious tones (a power summation would be a good start) in order to estimate the overall noise power in the given bandwidth.
  4. That indicates that he does know about the existence of 'negative frequencies' at the DAC's output but doesn't make the connection that they upset the level meter used by the AP to gauge the frequency response. Perhaps he makes the tacit assumption that the AP can tell the original stimulus apart from the images? He generally also publishes the FFT of a white noise stimulus which does show the true FR and so far I've noticed no remark on why the two measurement methods disagree.
  5. In general yes I would agree. But there is one 'edge case' I've noticed where he appears not to understand what he's actually measuring. That case being an unfiltered NOS DAC - its invalid to make a direct swept frequency response measurement using an AP as the upper octave level gets severely corrupted by the unfiltered images. He seems blissfully unaware of this.
  6. Not really that odd - good numbers on a DAC attracts customers who tend towards the view that a DAC with outstanding measurements is essentially transparent. Hence nothing to talk about.
  7. Personally I prefer a slightly more nuanced version - 'everything matters but to differing degrees, including infinitesimally small'. I would agree that many 'subjectivist' mindsets don't seem to have the concept 'low-hanging fruit'. I also prefer 'show' to 'tell' but for me, ASR leans too far in the direction of the latter. The corollary of 'show' to me is observe and observations have quality. Numbers don't. Incidentally @JoshM I think the reason they keep on measuring DACs is because its popular. Going back a couple of years, ASR didn't seem to me to be adding many new members. After Amir got his brand new AP it really picked up and now has become a honeypot for numbers guys.
  8. If you ask them for a falsifiable definition of 'properly engineered' you'll never get one. Sometimes 'competently' is used instead of 'properly' but the intent is the same. A more nuanced mantra is 'All competently engineered DACs sound identical under level matched, double blind conditions'.
  9. Looks once again that you've missed the point. See if you're any wiser after reading this article on Goodhart's Law : https://adexchanger.com/data-driven-thinking/what-goodharts-law-can-teach-you-about-performance-data/
  10. Looks like dodgy reasoning to me. When people target a particular distortion metric, that particular metric ceases to be one guaranteeing a good outcome for customers.
  11. Summed up in Goodhart's Law : When a measure becomes a target, it ceases to be a good measure.
  12. Even the very latest AKM DAC (AK4499) has only -100dB stop-band rejection. Also note it uses an equi-ripple filter. These are known to exhibit pre-echoes - see a paper on this by Julian Dunn. http://www.nanophon.com/audio/antialia.pdf
  13. Thanks - I should have added that there's no speaker protection on these so I would recommend either using only cheap speakers or find a protection module to strap on the output.
  14. I don't know if classD amp cards come within the remit of this thread (delete this if not) - if so I'd like to give a shout-out for these highly affordable ($11) mono amp modules. https://item.taobao.com/item.htm?spm=a230r. They don't come with input buffers (Zin ~3.3kohm) and you do need to supply a balanced power supply. Its a very low-cost route to respectable classD sound - you can tweak the card fairly simply to get what I consider to be high-end sound.
  15. Since the topic of input filtering has already come up on this thread, here's something about aliasing from OOB input signals : https://www.diyaudio.com/forums/class-d/321632-hypex-ncore-nc400-input-anti-alias-filter.html
  • Create New...