Hi, Hans, I'll bet that the Sun has you well bronzed by now.
I concur, with the above. It seems that accurate identification of the fundamental underlying mechanism responsible could require a series of well controlled listening experiments. Which I don't feel is practical for the thread at this time. Perhaps, it might be conducted within some future thread, or maybe added on the end of this thread at some uncertain future date.
Our experiments to investigate the NOS/OS subjective difference question have concluded. As you already know, only the PGGB upsampled 176.4KHz listening experiment remains, although it is not actually part of the original investigation question. Instead, it is part of an effort of 'turning over a last stone' in hopes of identifying a playback method which is subjectively superior to either 44.1KHz NOS or 88.2KHz PGGB upsampling.
Regarding the Day-O track which you recorded direct from LP, I think that the fact we unanimously preferred it (the only instance throughout all of the tests where that happened) in the up/dn test is worthy of further examination. So, what I propose, is that you prepare some new tracks, direct from LP as you originally did with preferred Day-O file, but after the PGGB 176.4 upsampling test has concluded. We could then offer a 'bonus' 😀 listening experiment, for participation by those who still have an interest in determining whether your digital recording set-up consistently produces 44.1 digital transfers which sound superior to high-performance upsampling of those same files.
After all of which, I will write and post a concluding investigation report. The thread will then 'officially' end in the purpose for which it was begun. Of course, it will continue to exist, but will be free for others to extend, or to repurpose in some related direction should they wish.
If it is arguably accepted that NOS reproduction is of advantage to OS reproduction wouldn't it be of advantage to first test NOS reproduction at 44.1KHz and at perhaps 88.2KHz whereupon the recording is done at the same frequencies. Is there already a conclusion that NOS 88.1KHz is superior to NOS 44.1KHz?
All of our experiments have indeed utilized 44.1KHz NOS as the base testing format. The initial objective of the thread was to identify what was causing 44.1KHz NOS and 44.1KHz 'typical' OS to sound characteristically different from each other. Beyond that, to hopefully discover how to improve upon the sound of either of those with native 44.1KHz tracks.
Certainly, 88.2KHz NOS would be interesting to evaluate, except that it properly requires the recorded native sample rate be 88.2KHz, and not be upsampled from some lower rate. I don't know how common native 88.2KHz recording are, but I have the impression they are not that common. For audio frequencies, 88.2 is not superior to 44.1. What it does offer, however, is the opportunity to utilize anti-alias and anti-image filters featuring a more gentle entry in to their transistion bands.
NOS Sound compared with HQ Player
I will keep it short, as reactions were zero last time, but may be still interesting for the conclusions from Ken – if not just say so, I will back off 😀
I listened to original 44.1 kHz tracks and compared on the fly (Roon and HQPlayer allows for that) to upsampled /filtered versions FSx4 (176.4kHz) – by doing this the PCM1794 modulator runs 4x faster
Hardware: DDDAC1794 in no filter mode – OPT output – FiFoPi with Accusilicon XO clocks
Start: No filtering, no upsampling
First: Roon with normal FIR Filter
Second: HQP with FIR Filter
Third: HQP with Polynominal-1 Filter (See old post – VERY close to sample and hold filter)
Listening compared to Start situation:
First: uninspiring is the word. It just misses the freshness and directness of the start version
Second: Now HQP does the FIR better. The uninspiring level is much less as in First
Third: Now we are talking. It sounds almost the same in terms of typical NOS sound. Almost, very close… probably not able to hear this “the next day without knowing what plays”
BUT… On a positive, for all voices and air instruments sound just a bit cleaner – same effect when mains power is cleaned up – This is for a sure a small PRO
All having said so…. The differences are small between start and third and I will need to listen further if I would like to have this now standard in my reproduction chain or not. Both Start and Third has pro and cons (so far in my ears)
I will keep it short, as reactions were zero last time, but may be still interesting for the conclusions from Ken – if not just say so, I will back off 😀
I listened to original 44.1 kHz tracks and compared on the fly (Roon and HQPlayer allows for that) to upsampled /filtered versions FSx4 (176.4kHz) – by doing this the PCM1794 modulator runs 4x faster
Hardware: DDDAC1794 in no filter mode – OPT output – FiFoPi with Accusilicon XO clocks
Start: No filtering, no upsampling
First: Roon with normal FIR Filter
Second: HQP with FIR Filter
Third: HQP with Polynominal-1 Filter (See old post – VERY close to sample and hold filter)
Listening compared to Start situation:
First: uninspiring is the word. It just misses the freshness and directness of the start version
Second: Now HQP does the FIR better. The uninspiring level is much less as in First
Third: Now we are talking. It sounds almost the same in terms of typical NOS sound. Almost, very close… probably not able to hear this “the next day without knowing what plays”
BUT… On a positive, for all voices and air instruments sound just a bit cleaner – same effect when mains power is cleaned up – This is for a sure a small PRO
All having said so…. The differences are small between start and third and I will need to listen further if I would like to have this now standard in my reproduction chain or not. Both Start and Third has pro and cons (so far in my ears)
Attachments
How you can say NOS is uninspiring? It is the most involving, live, natural and inspiring in my experience. Missing some smoothness of OS, but downloaded hi-res files carry the same smothness at no expense of losing the above attributes (which happens when it is resampled with Foobar/SoX).
If you purchase some titles of good labels, you can download both CD and a high resolution version. By example ECM titles 96kHz sound better than the same album in CD format without losing NOS attributes. 88.2/96kHz format sounds the best on the budget devices. Jumping to 192kHz require installing low jitter clocks.
There are samples in number of formats professionally resampled from hi-res recording, free for download: 2L High Resolution Music .:. free TEST BENCH
If you purchase some titles of good labels, you can download both CD and a high resolution version. By example ECM titles 96kHz sound better than the same album in CD format without losing NOS attributes. 88.2/96kHz format sounds the best on the budget devices. Jumping to 192kHz require installing low jitter clocks.
There are samples in number of formats professionally resampled from hi-res recording, free for download: 2L High Resolution Music .:. free TEST BENCH
He didn't. He called Roon with normal FIR filter uninspiring if you read more carefully... 🙄
I would be interested in a benchmark of a few high-end resampling solutions against NOS & standard filtering.
E.g. an experiment could be to take a few very high quality 44.1 kHz recordings and upsample them with state of the art solutions to both 88.2k and 176.4k (perhaps even 352.8k and 768k, though not everybody can play this). E.g. with PGGB, M scaler, Grimm MU1 etc. NOS users could try whether the higher sampled versions sound better then 44.1 kHz NOS and OS users could compare whether the 88.2 and 176.k sound better than the standard filtering and whether it gets progressively better with higher rates...
E.g. an experiment could be to take a few very high quality 44.1 kHz recordings and upsample them with state of the art solutions to both 88.2k and 176.4k (perhaps even 352.8k and 768k, though not everybody can play this). E.g. with PGGB, M scaler, Grimm MU1 etc. NOS users could try whether the higher sampled versions sound better then 44.1 kHz NOS and OS users could compare whether the 88.2 and 176.k sound better than the standard filtering and whether it gets progressively better with higher rates...
Hi, Doede, good work. 🙂
If I correctly understand what you found; options one and two essentially sounded like typical OS, and were decidedly inferior sounding to NOS. The preferred third HQP option utilizing a slow transition filter sounds better than options one and two, yet still sounds inferior to NOS. Do I understand this correctly?
A question: The far right photo appears to show the frequency-response of the filter options. Why do they all appear to be not flat below 20KHz?
Last edited:
We have conducted just such an experiment for PGGB 88.2KHz upsampling, and will soon conduct one for PGGB 176.4KHz upsampling as well.
While we have been utilizing the PGGB upsampling software since it's the best performing of which we are aware, I was hoping to get an subjective evaluation of the Chord M scaler versus the PGGB from one of our thread contributors who owns the Chord, Lampie519. The road-block to obtaining such a comparison is that his digital relay chain only supports SPDIF, so he cannot directly play our test files from his PC without subsequent conversion to SPDIF.
Newer mind... Indeed, sorry guys.
Listening now to Angel's Dance: Airelle Besson - Try! (2021)
Airelle Besson - Try!: lyrics and songs | Deezer
Album Try!, Airelle Besson | Qobuz: download and streaming in high quality
BTW, S/PDIF do not require converting, it will play bit-perfect on most DAC's.
Listening now to Angel's Dance: Airelle Besson - Try! (2021)
Airelle Besson - Try!: lyrics and songs | Deezer
Album Try!, Airelle Besson | Qobuz: download and streaming in high quality
BTW, S/PDIF do not require converting, it will play bit-perfect on most DAC's.
Last edited:
Almost Ken. thanks for your summary. I would not even call it inferior… The third in my ears is very close to typical NOS, with the extra pro (likely only as I use a pcm1794 in NO Filter mode with BCK driven Modulator) that the sound is a bit cleaner, (thanks the 4x higher BCK) hence I wrote that I am bit torn apart, what I would like more eventually….
On your question, this is not a frequency response, it is the FFT of a square wave… hence the falling curve. I wanted to have a lot of harmonics and their resulting noise effects
A "good" square wave contains a not ending equal level of HF. Is that what you want out of a DAC?
The PCM theorem don't work like this. It's ignorance to think that a good looking square wave is something good - its not, in context of PCM, its a sign of a broken, faulty reproduction mechanism.
Your lack of understanding really shows when you associate a "perfect" square wave with "NOS sound".... sound?
//
The PCM theorem don't work like this. It's ignorance to think that a good looking square wave is something good - its not, in context of PCM, its a sign of a broken, faulty reproduction mechanism.
Your lack of understanding really shows when you associate a "perfect" square wave with "NOS sound".... sound?
//
Why not do an experiment like:
PCM, Fs: 2ksps, 16bit, stereo.
a) Replay: 950Hz FIR filter, down -120dB at 999 Hz, (64ggr OS).
b) Replay: No filter (NOS!!) for a perfect square wave 😉
Record this with 24/96 with a good ADC and post the files.
Listen...
//
PCM, Fs: 2ksps, 16bit, stereo.
a) Replay: 950Hz FIR filter, down -120dB at 999 Hz, (64ggr OS).
b) Replay: No filter (NOS!!) for a perfect square wave 😉
Record this with 24/96 with a good ADC and post the files.
Listen...
//
In stead of trying to show your superiority and suggestions of my lack of understanding, you could have understood, that a square wave shows what the filters are doing. I have tried to find correlations with the typical NOS sound and share my perceptions.
Why is it some people try to push down on others? I know perfectly well the backgrounds. And of course in music there are no square waves.
Still this whole thread is about sharing experience and that is what I just did.
There is no need at all to start the canceling routine and show off how great you are. I am getting a bit fed up with this.
Sorry, just needed to let this out. I will ignore further allegations
Why is it some people try to push down on others? I know perfectly well the backgrounds. And of course in music there are no square waves.
Still this whole thread is about sharing experience and that is what I just did.
There is no need at all to start the canceling routine and show off how great you are. I am getting a bit fed up with this.
Sorry, just needed to let this out. I will ignore further allegations
Hi TNT,
Maybe you can explain what you don’t like when NOS Dac produces a analoque square wave from a digital version.
A digital square wave can never be recorded that way and only serves to prove that it is indeed a NOS Dac.
Hans
Maybe you can explain what you don’t like when NOS Dac produces a analoque square wave from a digital version.
A digital square wave can never be recorded that way and only serves to prove that it is indeed a NOS Dac.
Hans
Would you judge option three as sounding similar to the PCM1794A with it's internal OS interpolation-filter set in slow roll-off mode?
Good point Ken !
I did this already long ago when I was working on the DDDAC prototypes and recently tried again, based on this thread...
I would say it is on par or even below Roon with FIR.... even in slow roll off
Again, it is all not so BIG, but clearly audible.
The hardware solution probably has its limitations.
It measures very nicely though (But that is not what we are after, right ?)
I did this already long ago when I was working on the DDDAC prototypes and recently tried again, based on this thread...
I would say it is on par or even below Roon with FIR.... even in slow roll off
Again, it is all not so BIG, but clearly audible.
The hardware solution probably has its limitations.
It measures very nicely though (But that is not what we are after, right ?)
Sure the theorem is correct, yet from my understanding there must exist some mechanism that converts the HF spectrum back down into the audio spectrum for that to subjectively relevant. Furthermore, whatever mechanism is used in the digital domain to manipulate numbers the outcome is still a new set of numbers containing an infinite spectrum you suggest. Hence taking your perspective there isn't anything digital that isn't broken. Finally, if filtering of a square wave is perfectly done in the analog domain how is it variant from its digital counterpart?
TNT, dddac is just measuring the step responses of various filters to reverse-engineer what they do.
In principle a zero-order hold is almost the worst filter you can think of, with their roll-off in the audio band and lots of barely suppressed images, which is why it is peculiar that many people like their sound anyway. Hence this whole thread to figure out why.
In principle a zero-order hold is almost the worst filter you can think of, with their roll-off in the audio band and lots of barely suppressed images, which is why it is peculiar that many people like their sound anyway. Hence this whole thread to figure out why.