Probably the search for "detectable difference" would be a better first step to see if there indeed is one. This is also less stressing as to leave out what is better/correct...
//
//
It's difficult to completely avoid linear-phase filters with passband ripple.
If it is an LP from the 1970's or beyond, digital tape recorders may have been used during production, although I suppose digital tape recorders from the 1970's used minimum-phase analogue filters. Regarding new records, a small subset of the records made by Record Industry in Haarlem is completely analogue.
The ADC that you use probably has a linear-phase decimation filter or filter chain. The same holds for the ADC I used for recording that amateur chorus.
None of this prevents NOS DAC users from preferring NOS DACs, though.
Hans, aside from the possible application to the echo test, as you'll recall, this subject is an unresolved mystery from our first experiment. The mystery of why your own digitization of the Day-O track direct from LP was unanimously preferred over it's PGGB up/dn resampled twin made from that source file. None of the commercially produced source tracks utilized in the experiment were unanimously preferred over it's resampled twin. Speaking only for myself, this remains a very intriguing unresolved mystery.
I didn't want to burden you with additional effort, or the thread participants with yet another experiment, so I let the mystery drop. However, I'm curious enough to participate in an experiment myself to solve this mystery, should you be willing to digitize more LP sourced test files. I'm uncertain as to whether such an experiment should be against PGGB 88.2, or rather 176.4. Contact me via PM to discuss if you prefer not to do so within the thread.
Actually Hans Polak has a good idea. I also agree for using PGGB for downsampling a file called 'original'. The main issue is transparency. We were told that one file was original, but actually it was downsampled (and still unknown quality). I did sound inferior to my Keith Jarret original copy and a downsampled realtime in SoX). I suspected a manipulation and decided to not submit my result. You would have one R2R NOS user, but you have none.
The hint for listening was completely misleading. I spent two days listening in repeat mode waiting for a fatigue (or listening most pleasure, or the least displeasure), but it didn't give a clue. So I switched to the analytical mode and made a choice based on the NOS sound properties (which were already degradated due to a poor downsampling). However my analysis was right. If you read my post #1075, and use a true NOS DAC, you will see it was correct: What do you think makes NOS sound different?
These tracks must be selected based on the natural sound characteristic. Delta-Sigma users have no chances, as smoothing destroy natural properties. A decision based on the pleasing factor can be wrong, it is what this test had proven.
You can't be successful with this test without R2R NOS users. It seems there is a hidden agenda for this test that a well made upsampling should justify use Delta-Sigma converters. It is wrong, as there are other factors like noise shaping, probably the real contributor for a fatigue in my brain.
If you want another round, announce a test on other forums in the 'appreciation' threads for specific brands specialising in NOS DACs. Based on the response decide a place for the venue, it doesn't has to be on diyaudio. The best is NOS-only DAC to avoid cheating. My Audio GD R2R-11 has no hardware capable of DSP processing, but higher priced models also offer true NOS as an option. Avoid brands with questionable NOS like Denafrips or Holo Audio.
EDIT: I would keep 88.2kHz for tests, but a double rate for processing files. It is to reduce jitter influence.
The hint for listening was completely misleading. I spent two days listening in repeat mode waiting for a fatigue (or listening most pleasure, or the least displeasure), but it didn't give a clue. So I switched to the analytical mode and made a choice based on the NOS sound properties (which were already degradated due to a poor downsampling). However my analysis was right. If you read my post #1075, and use a true NOS DAC, you will see it was correct: What do you think makes NOS sound different?
These tracks must be selected based on the natural sound characteristic. Delta-Sigma users have no chances, as smoothing destroy natural properties. A decision based on the pleasing factor can be wrong, it is what this test had proven.
You can't be successful with this test without R2R NOS users. It seems there is a hidden agenda for this test that a well made upsampling should justify use Delta-Sigma converters. It is wrong, as there are other factors like noise shaping, probably the real contributor for a fatigue in my brain.
If you want another round, announce a test on other forums in the 'appreciation' threads for specific brands specialising in NOS DACs. Based on the response decide a place for the venue, it doesn't has to be on diyaudio. The best is NOS-only DAC to avoid cheating. My Audio GD R2R-11 has no hardware capable of DSP processing, but higher priced models also offer true NOS as an option. Avoid brands with questionable NOS like Denafrips or Holo Audio.
EDIT: I would keep 88.2kHz for tests, but a double rate for processing files. It is to reduce jitter influence.
Last edited:
Another idea would be to start with unprocessed DXD files in 352.8/24 format.
Not much processing is to be expected in these files.
Hans
sajunky, one of the files is the originally purchased 44.1 kHz file, multiplied by 0.7071 if needed, dithered and requantized. I haven't a clue if any of them are decimated 96 kHz files or if your 96 kHz file is an interpolated version of the 44.1 kHz file.
Most digital recordings are made with sigma-delta ADCs, so most of them will be mixes of decimated sigma-delta modulates.
Most digital recordings are made with sigma-delta ADCs, so most of them will be mixes of decimated sigma-delta modulates.
It's a decimated sigma-delta modulate, so it would lead to another rant from sajunky and depending on how the decimation is done, it may have pre- and post-echoes. It will be less processed than most files, though.
sajunky, your opinions about SDM, and R2R are based on your own experiences. Because they are true for you, doesn't necessarily mean they are true for all others. We all walk on thin ice when we speak in absolutes regarding our own experiences. Just something to keep in mind.
While your contributions have been welcome, this isn't the first time that you've referred to some "hidden agenda". Either you need to be very specific regarding the details of such an accusation, or you need to please stop.
It's hard to determine exactly what you are proposing in the above. For example, what do specific brands of DAC have to do with Hans' suggestion?
Last edited:
Unknown, but I wonder that too. Also, and has already been mentioned, the degree of studio, and actual format release processing is unknown, and likely to be a factor.
Please not be ridiculous. Delta Sigma ADC is not the same. In a DAC digital processing cannot be avoided, but look at different ADC topologies; analog filters can be used. Even there is a FIR filter, there are different settings for noise shaping (my biggest enemy).
mvs0, thank you for this suggestion. No doubt, emphasized CDs played without de-emphasis produce an audible frequency-response error. Just the same, I think that we can safely discount emphasis/de-emphasis as being the cause of the persistent characteristic sound difference between OS and NOS, for the following reasons.
1. If it were due to emphasis/de-emphasis, then the NOS/OS character difference would only intermittently occur, depending on whether a given CD featured emphasis or not. The phenomena which we have been investigating is prevalent via non-emphasized CDs.
2. The PGGB 88.2 experiment showed that the character NOS/OS character difference could be essentially nullified via very high-performance upsampling/oversampling. If it had been due to emphasis/de-emphasis, then the performance of the interpolation-filter would not have nullified the subjective difference.
Perhaps, I've misunderstood your point regarding de-emphasis, so please let me know if I have.
Weren't you referring to Hans' suggestion to possibly utilize a set of specially prepared files to conduct a more determinative echo experiment? If you were not, then it's not clear to me what you were referring to in your post.
Not. Hans suggestion is covered in different section. If you don't understand, put it aside and read again on the end of the next failed round.
Last edited:
I didn't realize you are fine with sigma-deltas as long as the loop filter is analogue. Unless it's a DSD recording made without 352.8 kHz noise shaped PCM as an intermediate step (DXD), there will be a digital decimation chain in the path.
I can understand frustration sajunky, however this inclusion suggests of Shakespeare "to be or not to be", that by your statement implies that you are not of a person ridiculous in the comparison. Yet not to be a "ridiculous person" suggests you are a person of greater knowledge and authority to those you seem personally to conclude ridiculous, and whereupon you possess knowledge of some greater or unequivocal truth as not requiring any defence by argument to those you conclude as ridiculous. It doesn't seem you understand fallacies of argument or perhaps consider yourself of such innate wisdom that the rest of us are all ridiculous in the comparison.
Of course. When you don't understand, refer to the place you don't understand. Your question however was completely unrelated to the text you were quoting. It loooks like a diversion from the text contained in this paraghraph. Either subconsciously or intentionally driven (for what?). It is not my intention to guess which one it was. In such situation I am usually trying to give the person a time to process the content. There was a failure and you need to deal with like a man, not clown.
Ken,
Just let me think loud about what to do.
Wasting others people time with a zillion different experiments is no fun for nobody, on the other hand we had a clear sign that up and downsampling on a relatively pure file from an analoque recorded LP was detected by all participants.
One swallow doesn’t make a summer, but this is yet another sign that sound perception may be extremely sensitive to some form of digital processing.
In the light of this thread, where sound differences between NOS and OS are the subject of investigation, we already noticed that no filtering is better than using just some average filtering like Audacity.
And although it would be nice to know in more detail what part of processing is most critical, as tried in Marcel’s sophisticated test, but to test things in isolation doesn’t seem to be that easy.
But even when getting a more profound understanding on things, what would it bring other than intellectual satisfaction, because we have to live with the content offered by the market with no means of influencing it.
So to conclude, let Ken decide what seems to make sense and I’ll do whatever is in my power.
Hans