- Joined
- Mar 20, 2013
- Posts
- 16,675
- Likes
- 12,487
Hi,
From our Thorsten on his tea break:
Interesting yes, but in more ways. If you do not mind our commenting (and this is just our comments nothing more nothing less):
1st set of graphs
The first set of graphs show the output from iDSD with the actual audio range compressed into the very first fraction of a millimetre and shows RF signals up to a frequency of 5MHz. So the graph shows the output at high and very high supersonic frequencies. These graphs do not reveal anything whatsoever about what happens in the audible range.
At the latest count not even the most golden ear’d of audiophiles have claimed to hear noise at several MHz directly. The extant literature fails to support such a feat of hearing for even Dracula himself in bat form and while one may conceive that the Almighty One or the Archangel Emanuel may be capable of such a feat we'd probably ask them to pass an ABX test on that first.
It is debatable to what degree the high and super high frequencies influence other replay equipment, but no matter what, the frequencies themselves are entirely inaudible, they can cause audible problems only by interaction with other gear and if the rest of the system relies on well-designed equipment, there should be no issue. And if there is, with Bitperfect or Minimum-Phase filters, simply use the standard filter instead.
2nd set of graphs
The second set of graphs is a bit more telling of in and near audio band noise. It is worth noting the different frequency scales for the bitperfect filter and the other filters and it would probably be better to see all graphs on the same scale to be able to draw fair conclusions. Mmm.
Looking only at the graphs that are directly comparable we find that any changes caused by up-sampling to 768kHz externally compared to the iDSD micro's build in filters are below -110dB and/or supersonic. Noisefloor is basically the same. If converting PCM to DSD we see a rising in-audio band noisefloor, though the different amplitude scales between DSD256 and DSD512 make it hard to see just how much is going on.
How much of the above happens to be artifact of the actual test gear is also open to question. All test gear has limits. Wisely the gentleman who posted these measurements has generally refrained from making any comments that attempt to relate the measured results to anything meaningfully audible, as such a thing would not be an easy thing, if defensible at all.
Yet any owner of the iDSD micro can make an experiment in this line.
To obtain 352.8kHz/384kHz upsampling simply engage the digital filter. As the graphs show, the differences between this and external 768kHz upsampling are quite small. If you prefer using the standard digital filter then you may prefer the external upsampler too. If you prefer minimum phase filtering (which allows more ultrasonic noise than standard) or G*d forbid the horribly noisy (above the audio range anyway - a consequence of not applying any digital filtering) Bitperfect filter, then any other upsampler will likely not work for you either.
In the same line, as long as the OS is windows and the playback software supports ASIO, it is easy to try out DSD256 or DSD512 conversion of PCM using ASIO-Proxy. Again, if you like the result better than straight PCM with any of the available filters, maybe you will like other conversion software even better. Then again maybe Bitperfect is preferred.
The nice thing with the iDSD micro is that you can try it all out with the iDSD micro for yourself: PCM up/oversampling, DSD conversion, you name it.
Take a listen for yourself and draw the conclusions for yourself, rather than listen to what anyone suggests you should hear.
Somewhat of a long way round but in the end, graphs don’t mean they are more authoritative just because they have lots of squiggles.
We just wanted to offer a more rounded perspective.
