Someone who can recognise this distinct sound can also tell us if he or she can hear the same sound in the emulated NOS signal.
If this persons says its 'the same' we can conclude that the distinct sound is somehow caused by the sample and hold, in this case emulated by the replication of 4 or 8 44.1KHz samples.
If not we have to look further...
But if this test succeeds, its quite a shortcut to the search that is now proposed.
You could greatly assist us in evaluating the resampler you've been suggesting, if you were to download the four 44.1 source files we've using for testing the PGBB resampler, from their Dropbox, and then to up/down 44.1 resample those four source files. We could then conduct one more round of 'transparency' testing utilizing those, with Hans Polak managing the test proceedure.
Is it possible for you to perform the up/dn 44.1 resampling of the source files? I'm asking you this because I presume that you are already in possession of, and familiar with using the resampling tool. Which would releave Hans of needing to acquire and learn to use it, just to create the resampled test files. If you are interested and willing to assist the thread, contact me via PM, and you and I can communicate further about it.
If that test indicates a subjective transparency similar to the PGGB's, I think we can then move to conduct a direct subjective comparison between it and the PGGB, both upsampling to 88.2.
I understand your motives and the limitations that are set, as well as the scope, reasoning and goals that are behind it and also think it's one of audio's more interesting phenomena and I am really curious where it will lead to, I also know there not an easy fix, so I hope I can at least assist in getting you the tools I find useful.
I'd be happy to assist any way I can but:
PGGB , High-end Audioconverter (and Izotope?) etc are off-line, stand alone converters. They convert file A to file B in X speed. HQPlayer pro, around several thousands of $ does this as well.
However, the most used consumer type ,the non-pro HQPlayer Desktop, is around €250, but does it real time. So upon file "play" it goes straight to the dac, while it is processing. You press play, initialise for a few seconds and from then on playback is real time. It doesn't allow to save the file to disk. This is the version I also use, which restricts any up- and downsampling test I could do for you.
The real issue with letting someone do the conversion for you, is that of all those 39 filter options, one would ideally make a pre selection which ones really are useful, exhibit some magic, some synergy and which ones can be discarded. Only after that one could toss around versions. It is a known issue with hqplayer: many options, steep learning curve. But they're there for a reason.
So I'd best be honest and say to everyone who is interested, to download a copy, read the manual and see where it leads you.
I won't be surprised if it doesn't beat PGGB, it can, but it will give more insight in what oversampling does by just clicking some boxes. You might as well find the solution.
HQPlayer is available at Signalyst.com and Jussi Laako is available almost any time by email, or on the forum of audiophilestyle, be it for Windows, Mac or Linux.
HQ Player - Software - Audiophile Style
Once again,, because I know I come across like the signalyst sales man, I could easily have made a case for any other piece of software, but this is what I use and it fits your description of wanting to have more to say about the parameters used for upsampling in order to get to a satisfiable result. It really is an engineers toolbox. Hell, maybe the one thing that will stand out is that we don't like a certain dither, for all I know.
Following the thread with interest
Marco
Indeed it is dependant on the interpolation filter used. In this case it was the upsampling filter of audacity. But it would give a good indication of the number of expected events in an actual DAC that oversamples and can not output above the 0dBFS output voltage. And these events are very rare in the files we used so far.
Last edited:
Not wanting to be rude, but I have no idea what you are talking about.
We are talking about sweet and sweeter, not about sweet and bitter.
Hans
Marco, does that mean that the HQplayer only makes it's resampled output stream available via an USB port on the PC?
With these limitations, the only was I can see that we could conduct a listening evaluation is if everyone individually downloads and installs the HQplayer software on their computer. I think that the inconvenience and complication of that is, unfortunately, probably a show stopping impediment. Perhaps, the group feels differently, or someone can suggest a practical alternative.
Customizing the upsamling process may be desirable for pleasure listening, but having a uniformly produced test of files is preferable for conducting a group test.
Once again,, because I know I come across like the signalyst sales man, I could easily have made a case for any other piece of software, but this is what I use...
Following the thread with interest
Marco
No worries, Marco. Anyhow, I still would be interested to read your anecdotal, subjective assessment of the general sonic character produced by your HQplayer, upsampling to 88.2KHz (utilizing whatever option settings you prefer) sounds to you, compared to the general sonic character of 44.1KHz NOS playback.
May I suggest the soxr resampler in VHQ (very high quality) setting?
That's an open source tool that
a) is available to anyone for free
b) can be used to "offline" process files and also is available in several software projects as a realtime resampling option (for example pulseaudio)
c) provides one of the cleanest resampled outputs in terms of measureable artefacts
That's an open source tool that
a) is available to anyone for free
b) can be used to "offline" process files and also is available in several software projects as a realtime resampling option (for example pulseaudio)
c) provides one of the cleanest resampled outputs in terms of measureable artefacts
Hi, Tfive,
Thanks for that suggestion. Our first round of digital filter testing utilized Audacity. Which, as I understand it, features the SoX resampling utility. Hans Polak prepared those first up/dn resampled test files using the highest quality setting offered in Audacity.
No I agree, but i2s, Right Justified, Left Justified, 44k1 & 48k, families, 16, 18, 20, 24 and 32 bits also aren't one standard, nor do they say anything about exact DNR or SNR per se.
A modern (PCM ADC )chip needs internal glue logic to communicate to the outside world and regarding SDM converters: there are no multibit SDM chips in the audio world, if that's what you mean by missing standards for SDM. it's all single bit outside and multi bit for PCM. The inner workings of a chip only matter when you specify something like NOS and want to find what causing the change in character, but maybe you look at that differently. I'd like to know why.
NOS, or Native is using the original format in a DAC chip without alterations. A SDM FIR dac is a good example of a native SDM dac, the newer solid state R2R PCM DACs for PCM.
By native in this context I mean staying in the original format, from AD converter to DA converter, and in this specific case without further processing to alter bitrate, bitdepth, as that is what is the subject here, more or less, right?
One can choose to completely omit the recording part, and it's used filters, but that sure makes me scratch my head.
Iirc one of the promises made by the introduction of SACD/dsd was that the filters needed for PCM recording and playback could be left out and that this would lead to a better, more transparent recording & playback chain. That's around 30 years ago and we're here discussing if specific filters used in the recording chain, can make for a different outcome when mixed with varying filters in the playback chain.
Imho this isn't debatable, you either accept that at both sides they matter or they don't. If they do, the type of filter at playback could accentuate or obfuscate the pros and cons of the filter used at the recording side, and vice versa. Either way they interact and that's the thing: using different recordings yet 1 type of upsampling proces might lead to confusing results, since the filters at the recording side can/will be different for the varying test files, they will be more or less compatible with the chosen upsampling software at the playback side.
They should, or at least could is the point.
It's just a matter of finding the best upsampling setting for a given song, those settings are not likely to be best for all songs.
Mterbekke,
The purpose of this thread is by no means to find the best upsampler for a certain recording.
Not even is the purpose to declare NOS and OS the winner.
The question is if and why a difference in sound is perceived between the two.
Out of the many possible reasons that Ken has identified, the upsampling process is one of them.
Using Audacity gave a clear indication that it damaged the sound, that’s why Ken found ZB generously willing to support our search with their excellent PGGB software, several orders of magnitude more accurate than Audacity.
So PGGB became the vehicle of choice to help us finding out whether using an average or excellent upsampler makes a difference in the perceived sound and whether this can be one of the reasons or even the main reason of the perceived sound difference between NOS and OS.
So discussions that other upsamplers are easier to use, almost as good or a better match with the original recording are not in any way part of this process.
Hans
The purpose of this thread is by no means to find the best upsampler for a certain recording.
Not even is the purpose to declare NOS and OS the winner.
The question is if and why a difference in sound is perceived between the two.
Out of the many possible reasons that Ken has identified, the upsampling process is one of them.
Using Audacity gave a clear indication that it damaged the sound, that’s why Ken found ZB generously willing to support our search with their excellent PGGB software, several orders of magnitude more accurate than Audacity.
So PGGB became the vehicle of choice to help us finding out whether using an average or excellent upsampler makes a difference in the perceived sound and whether this can be one of the reasons or even the main reason of the perceived sound difference between NOS and OS.
So discussions that other upsamplers are easier to use, almost as good or a better match with the original recording are not in any way part of this process.
Hans
I have not followed all this thread, but what's wrong with Izotope upsampler? It offers no interpolation option. It just doubles / quadruples the sample rate without any filter. I've been using this iZtope SRC included in Audirvana for more than a few years in no interpolation mode.
PS: and it does real time conversion.
PS: and it does real time conversion.
Last edited:
For the PCM179x chips running in NOS (No digital filter) this could be a very nice option to speed up the modulator and shift noise to the right...
I just posted a feature request in Roon Forum
Upsampling with no interpolation - Feature Requests - Roon Labs Community
Any one who has an idea for a software I can try this ? iZtope costs 120$ and could not find a free trial
?
Thanks for the recap Hans, it was hard for me to understand what has been written before.
Sarc off
I can see you disagree, as none of this has anything to do with the point I was trying to get across.
So let me recap, shorter is better ain't it Hans?!
Do you admit the process used in recordings have an effect on the sq of the files processed, at playback, or not?
If that's a "No" then I wish you all the best in theorizing this endeavor further.
If that's a "yes", then what do you think is the best way to find out?
Try a multitude of processes at playback or try 2 kind of processes?
You tried a bad one, Sox I believe, and nobody was convinced, you tried PGGB and way more were nodding their heads. This is without using PGGB to its fullest.
All of a sudden a messenger pops up that explains why it ain't a 100% score, says what else you can do about it and you try to shoot the messenger.
Somehow you must know they interact, or can be more important than you dare to admit.
It seems somehow I'm a problem in your quest, but that reasoning has no bearing on the truth.
Good luck to you all.
Last edited:
Marco, it's unfortunate that you took Hans' response in that particular way. No one is attempting to shoot you. All thread participants are welcome and appreciated, including you. We are only trying to keep the investigation moving productively forward in an orderly manner. One of the ways we do that is to constrain it's scope from expanding much. Hans and I both desire the experiments to be thorough so that they can be conclusive, but practical concerns tend to often work against that.
While you make some interesting observations, we can only entertain a few of them, such as I did in my willingness to test the HQ player you've suggested. Until I realized that doing so would probably be too inconvenient in a group test. Meaning, the group participation in conducting the test would probably be low, to zero.
Added to the above, when talking about NOS and OS, the whole system is meant and not just a chip.
It’s important to be aware that many Dac systems are using Sample Rate Converters in front of the Dac chip (multibit or SDM) diffusing the effect of externally upsampled content.
An SRC is nothing but a another (huge) set of FIR filters to interpolate between incoming samples.
So NOS without SRC is the purest thing to evaluate the effect of upsampling and to notice the difference between upsamlers.
I will put the 88.2 test outcome in one overview, and will add to that the info to my best knowledge what kind of Dac system was used.
Hans
It’s important to be aware that many Dac systems are using Sample Rate Converters in front of the Dac chip (multibit or SDM) diffusing the effect of externally upsampled content.
An SRC is nothing but a another (huge) set of FIR filters to interpolate between incoming samples.
So NOS without SRC is the purest thing to evaluate the effect of upsampling and to notice the difference between upsamlers.
I will put the 88.2 test outcome in one overview, and will add to that the info to my best knowledge what kind of Dac system was used.
Hans
I use Audirvana exclusively, though with the suspicion that the iZtope SRC (64 bit) is not active given that Tidal Hi-Rez is 44.1KHz, and that my DAC doesn't support higher than 48KHz. On an aside, the other reason for using Audirvana is that it shuts down other programs on my MAC that could otherwise inject noise. On one of my flawed pre-amps that was attached to the DAC it was loudly picking up noise of the data transfer of Tidal files during the filling up of the internal buffer in the MAC. Once the buffer was filled the noise stopped. Although this noise isn't normally audible in any "well" designed line stage, it must be noted that such noise exists as needing be dealt with, and that in that process can nevertheless impact on audible experiences.
It doesn't appear the overall intentions of this thread are to find some best SRC, AS or NOS mechanism, rather that to understand the nature and extent of issues involving all manner of sampling that in turn focuses direction of further investigation upon that understanding. There is no guarantee that iZtope SRC is as good as it gets in the absence of hearing anything better. Is there something better, cheaper, etc.?
If the DAC is not immune to noise coming via the USB lines that's a whole other story and has nothing to do with the topic at hand. I know, it's a sad reality that a lot of DACs are not immune to EMI garbage coming over the lines, but that must be fixed in Hardware, not by any means of (not) oversampling.
Are you suggesting that limiting or shutting down programs in a computer can't have any audible impact to the results of the topic at hand. Further that those engaged in subjectively evaluating files respecting the topic at hand are all using DACs immune to EMI garbage coming over the lines? Simply because EMI has nothing to do with the intent of the topic at hand does not mean it can't corrupt the results or alter their subjective impressions. Of course it is relevant concern.
I have tried to summarize to the best of my knowledge the responses on the four 88.2Khz test in one sheet .
Although Lampi did not report on the 88.2Khz file, I have added him to the list because he repeatedly mentioned that a very good upsampler leaves the content intact on his NOS Dac.
It's also remarkable that the other two NOS Dac users noticed no difference between original 44.1 and upsampled 88.2 version, where the OS users had mixed results.
Since an OS Dac is mostly preceded by an SRC, this SRC will get a different job in interpolating/upsampling when starting with an incoming frequency twice as high, whereas the Dac chip may still be running in both cases at the same frequency. So perceived sound differences in that case will be caused by the SRC.
Hans
.
Although Lampi did not report on the 88.2Khz file, I have added him to the list because he repeatedly mentioned that a very good upsampler leaves the content intact on his NOS Dac.
It's also remarkable that the other two NOS Dac users noticed no difference between original 44.1 and upsampled 88.2 version, where the OS users had mixed results.
Since an OS Dac is mostly preceded by an SRC, this SRC will get a different job in interpolating/upsampling when starting with an incoming frequency twice as high, whereas the Dac chip may still be running in both cases at the same frequency. So perceived sound differences in that case will be caused by the SRC.
Hans
.
