Ahh, nice to hear from an old friend! ... I've listened so many times to these tracks over the years, they're like members of the family ... 
Group B is the one where the low level detail is better reproduced, in Group A the breakup, distortion of the sound is quite distinct - I'm interested that the hiss level is so high, it's masking the fact that these tracks are undithered, which normally comes through very clearly in the -60dB version.
Group B is the one where the low level detail is better reproduced, in Group A the breakup, distortion of the sound is quite distinct - I'm interested that the hiss level is so high, it's masking the fact that these tracks are undithered, which normally comes through very clearly in the -60dB version.
Wrapping it all up. Results.
You've nailed it Frank
The player was an early Philips CD150 that uses twin TDA1540 DAC's. The files here are comparing the player in Oversampling mode and Nonoversampling mode.
The TDA1540 DAC's were 1st generation 14 bit devices but Philips introduced a very clever 'noise shaping' and 4x oversampling circuit ahead of these DAC's that, it was claimed, allowed 16 bit resolution. As always in audiophile circles, there were claims and counter claims, one of which being that putting the chip set back to 14 bit mode was in fact audibly superior. Having an old player at my disposal seemed like a good way of investigating these modifications.
The modification itself is straightforward and consists of little more than removing the TDA(SAA)7030 and putting the TDA(SAA)7000 back to 14 bit mode by changing a logic level on one of its ports (there are also four links to fit in place of the TDA7030)
So far so good. The first thing I did was measure the output voltage of the player in NOS vs OS mode and this showed a big change. Could this alone account for reports of the modification being 'better'.
Player output voltage NOS vs OS. Note how the original set up meets the 'standard specification' of 2 vrms output at 0db.
Left channel.
0.000db 2.5420 vrms vs 2.0410 vrms. 1.906db difference.
-1.00db 2.2660 vrms vs 1.8200 vrms. 1.904db difference.
-3.00db 1.7990 vrms vs 1.4450 vrms. 1.903db difference.
-6.00db 1.2730 vrms vs 1.0230 vrms. 1.900db difference.
-10.0db 0.8020 vrms vs 0.6440 vrms. 1.905db difference.
-20.0db 0.2615 vrms vs 0.2100 vrms. 1.905db difference.
-30.0db 0.0825 vrms vs 0.0662 vrms. 1.912db difference.
-40.0db 0.0259 vrms vs 0.0208 vrms. 1.904db difference.
-50.0db 0.0080 vrms vs 0.0065 vrms. 1.803db difference.
-60.0db 0.0024 vrms vs 0.0020 vrms. 1.500db difference.
-70.0db 0.0007 vrms vs 0.0006 vrms. Rounding errors stop accurate result.
Right channel.
0.000db 2.5300 vrms vs 2.0320 vrms. 1.904db difference.
-1.00db 2.2550 vrms vs 1.8110 vrms. 1.904db difference.
-3.00db 1.7900 vrms vs 1.4380 vrms. 1.902db difference.
-6.00db 1.2670 vrms vs 1.0180 vrms. 1.900db difference.
-10.0db 0.7990 vrms vs 0.6410 vrms. 1.914db difference.
-20.0db 0.2600 vrms vs 0.2090 vrms. 1.896db difference.
-30.0db 0.0821 vrms vs 0.0660 vrms. 1.896db difference.
-40.0db 0.0258 vrms vs 0.0207 vrms. 1.913db difference.
-50.0db 0.0080 vrms vs 0.0064 vrms. 1.938db difference.
-60.0db 0.0024 vrms vs 0.0020 vrms. 1.584db difference.
-70.0db 0.0007 vrms vs 0.0005 vrms. Rounding errors stop accurate result.
Scale factor correction of 1.9db applied to OS files to equalise levels.
So on to the music files in post #1
Frank, you found B, C and E to be preferable which as you can see are the standard player in '16 bit' mode. In the Group A vs Group B selection Group A was 'Non Oversampled' and Group B was 'Oversampled' Group B which you leaned toward was also the player in standard form.
Jay had B,C and F as preferred choices. Two picks from the standard player.
rayma leaned towards B, D and F. Two votes for the modified and one for unmodified and in the low level files perhap leaned to group B which was Oversampled, the player being in standard form.
Statistically there aren't really enough votes and participants to draw definite conclusions but it was certainly interesting.
As always in these tests, a big thank you to all that took part
Files.
File A = NOS (modified player)
File B = OS scaled file. (Standard player)
File C = OS scaled file. (Standard player)
File D = NOS (modified player)
File E = OS scaled file. (Standard player)
File F = NOS (modified player)
Group A was Non Oversampled (modified form).
Group B was Oversampled (standard form).
You've nailed it Frank
The player was an early Philips CD150 that uses twin TDA1540 DAC's. The files here are comparing the player in Oversampling mode and Nonoversampling mode.
The TDA1540 DAC's were 1st generation 14 bit devices but Philips introduced a very clever 'noise shaping' and 4x oversampling circuit ahead of these DAC's that, it was claimed, allowed 16 bit resolution. As always in audiophile circles, there were claims and counter claims, one of which being that putting the chip set back to 14 bit mode was in fact audibly superior. Having an old player at my disposal seemed like a good way of investigating these modifications.
The modification itself is straightforward and consists of little more than removing the TDA(SAA)7030 and putting the TDA(SAA)7000 back to 14 bit mode by changing a logic level on one of its ports (there are also four links to fit in place of the TDA7030)
So far so good. The first thing I did was measure the output voltage of the player in NOS vs OS mode and this showed a big change. Could this alone account for reports of the modification being 'better'.
Player output voltage NOS vs OS. Note how the original set up meets the 'standard specification' of 2 vrms output at 0db.
Left channel.
0.000db 2.5420 vrms vs 2.0410 vrms. 1.906db difference.
-1.00db 2.2660 vrms vs 1.8200 vrms. 1.904db difference.
-3.00db 1.7990 vrms vs 1.4450 vrms. 1.903db difference.
-6.00db 1.2730 vrms vs 1.0230 vrms. 1.900db difference.
-10.0db 0.8020 vrms vs 0.6440 vrms. 1.905db difference.
-20.0db 0.2615 vrms vs 0.2100 vrms. 1.905db difference.
-30.0db 0.0825 vrms vs 0.0662 vrms. 1.912db difference.
-40.0db 0.0259 vrms vs 0.0208 vrms. 1.904db difference.
-50.0db 0.0080 vrms vs 0.0065 vrms. 1.803db difference.
-60.0db 0.0024 vrms vs 0.0020 vrms. 1.500db difference.
-70.0db 0.0007 vrms vs 0.0006 vrms. Rounding errors stop accurate result.
Right channel.
0.000db 2.5300 vrms vs 2.0320 vrms. 1.904db difference.
-1.00db 2.2550 vrms vs 1.8110 vrms. 1.904db difference.
-3.00db 1.7900 vrms vs 1.4380 vrms. 1.902db difference.
-6.00db 1.2670 vrms vs 1.0180 vrms. 1.900db difference.
-10.0db 0.7990 vrms vs 0.6410 vrms. 1.914db difference.
-20.0db 0.2600 vrms vs 0.2090 vrms. 1.896db difference.
-30.0db 0.0821 vrms vs 0.0660 vrms. 1.896db difference.
-40.0db 0.0258 vrms vs 0.0207 vrms. 1.913db difference.
-50.0db 0.0080 vrms vs 0.0064 vrms. 1.938db difference.
-60.0db 0.0024 vrms vs 0.0020 vrms. 1.584db difference.
-70.0db 0.0007 vrms vs 0.0005 vrms. Rounding errors stop accurate result.
Scale factor correction of 1.9db applied to OS files to equalise levels.
So on to the music files in post #1
Frank, you found B, C and E to be preferable which as you can see are the standard player in '16 bit' mode. In the Group A vs Group B selection Group A was 'Non Oversampled' and Group B was 'Oversampled' Group B which you leaned toward was also the player in standard form.
Jay had B,C and F as preferred choices. Two picks from the standard player.
rayma leaned towards B, D and F. Two votes for the modified and one for unmodified and in the low level files perhap leaned to group B which was Oversampled, the player being in standard form.
Statistically there aren't really enough votes and participants to draw definite conclusions but it was certainly interesting.
As always in these tests, a big thank you to all that took part
Files.
File A = NOS (modified player)
File B = OS scaled file. (Standard player)
File C = OS scaled file. (Standard player)
File D = NOS (modified player)
File E = OS scaled file. (Standard player)
File F = NOS (modified player)
Group A was Non Oversampled (modified form).
Group B was Oversampled (standard form).
Logitech desktop speaker again... But the difference is so obvious, I just have difficulties in picking one to prefer... It would be easier if I'm familiar with piano, but I'm not.
With -40dB files, background noise is louder with B. Piano is more snappy with A. Similarly, violin (or is it cello I'm also not familiar) sounds like violin with A but more like woodwind with B. From this I prefer A.
With -60dB files, at 0:16 A has louder music/instrument where in B it is pushed behind the noise. This is a plus for A of course, because I might say that A has higher resolution. But from the "snappy" characteristics on the piano, B here is similar to A in -40dB files. B has a whistling sound in the background noise. I think this whistling noise that makes B more snappy and the pitch is higher than A. If only I were familiar with piano I would be able to tell which one is wrong in pitch. But this is not about right or wrong, isn't it?
With -60dB it is harder to pick preference. Because even the method I use to pick my preference will be different here, because I cannot listen for musicality when both files are not musical I also don't know which background noise is more disturbing to me, the one A has or the one B has.
But if I have to be consistent with the -40dB files, I should pick B, because they have this "snappy" characteristics.
But the fact that at 0:16 A has the instrument sound more "above" the noise and that the piano pitch seems too high (as affected by the whistling noise), I could have been preferring the wrong file from the beginning hahaha
With -40dB files, background noise is louder with B. Piano is more snappy with A. Similarly, violin (or is it cello I'm also not familiar) sounds like violin with A but more like woodwind with B. From this I prefer A.
With -60dB files, at 0:16 A has louder music/instrument where in B it is pushed behind the noise. This is a plus for A of course, because I might say that A has higher resolution. But from the "snappy" characteristics on the piano, B here is similar to A in -40dB files. B has a whistling sound in the background noise. I think this whistling noise that makes B more snappy and the pitch is higher than A. If only I were familiar with piano I would be able to tell which one is wrong in pitch. But this is not about right or wrong, isn't it?
With -60dB it is harder to pick preference. Because even the method I use to pick my preference will be different here, because I cannot listen for musicality when both files are not musical
I also don't know which background noise is more disturbing to me, the one A has or the one B has.But if I have to be consistent with the -40dB files, I should pick B, because they have this "snappy" characteristics.
But the fact that at 0:16 A has the instrument sound more "above" the noise and that the piano pitch seems too high (as affected by the whistling noise), I could have been preferring the wrong file from the beginning hahaha
I have been waiting for you to say that this is about OS and NOS, because from the beginning I have drawn one useful conclusion...
I prefer BCF or OS-OS-NOS.
As I have said before, C is obvious for me (more preferable than D). B is next (more preferable than A), but between E and F I have difficulty to pick my preference but I decided to pick F.
Here is the key point: NOS is not good with highly dynamical music, but is good with vocal and slow music (such as E/F). Exactly the trade-off when you pick the classic NOS DAC TDA1543. It is a cheap DAC but many audiophiles still keep it just like I do, for a reason.
Effect of LPF
Okay, so we are comparing OS and NOS, to find out which one is "better". This is a digital modification but unfortunately it has relationship with analog LPF. Meaning that, for each file to sound at its best, the LPF should also be different. Any comment?
Surprisingly, the OS files (especially E) has more HF contents (sorry, I mean soundwise, because I don't have sound analysis software). What is going on? Isn't it impossible I can hear the sampling frequency? So, why? Is it a side effect of doing the oversampling?
Okay, so we are comparing OS and NOS, to find out which one is "better". This is a digital modification but unfortunately it has relationship with analog LPF. Meaning that, for each file to sound at its best, the LPF should also be different. Any comment?
Surprisingly, the OS files (especially E) has more HF contents (sorry, I mean soundwise, because I don't have sound analysis software). What is going on? Isn't it impossible I can hear the sampling frequency? So, why? Is it a side effect of doing the oversampling?
OS has higher resolution??
I didn't follow, which one is supposed to be higher in resolution? NOS=14 bit, OS is equal with 16 bit?
Listen to this -60dB files at 0:16. A (NOS) has louder instrument sound while in B (OS) it is almost pushed to the background. So, which one is higher in resolution???
I didn't follow, which one is supposed to be higher in resolution? NOS=14 bit, OS is equal with 16 bit?
Listen to this -60dB files at 0:16. A (NOS) has louder instrument sound while in B (OS) it is almost pushed to the background. So, which one is higher in resolution???
But the fact that at 0:16 A has the instrument sound more "above" the noise and that the piano pitch seems too high (as affected by the whistling noise), I could have been preferring the wrong file from the beginning hahaha
Typo. What I meant here is that
(1) A (NOS) has more "resolution".
(2) B (OS) piano pitch seems too high.
(3) B (OS) has whistling noise that might cause snappiness and high pitch.
(4) I picked B as preference, against "objectivist" criteria.
To be objective, may be A (NOS) is the correct sound???
Or may be too many things affecting this, in other words, uncontrolled test?
Or may be there is no right or wrong, it all depends on implementation?
The Big Conclusion
From listening to the files, and then knowing what they are, I have this new finding that mostly reinforce what I have believed about NOS (Non-Over Sampling) DAC.
As has been well known (by their fans of course), NOS sound so natural, and I agree with that. But I also realized that NOS is less musical and brings less enjoyment than OS DACs. I didn't know what caused this, but mainly I blame the not so good specification of the DAC IC in its dynamic ability (DNR or SNR).
But from this experience in this thread conducted by Mooly, and this thread:
http://www.diyaudio.com/forums/full...-comparison-3in-5in-drivers-round-2-a-21.html
I found out that naturalness, honesty, truth, is not always preferable. I knew that ScanSpeak 10F was the best driver from listening, but to be honest, I like to listen to cheap Aluminum cone Peerless (playing in the background) when I browse the internet. This Peerless has big overshoot and long ringing in the impulse response, but mainly it has elevated HF response. And we all know how some people deliberately add non-musical HF content to "modify" the sound so is subjectively become better.
In this test from Mooly I could feel from the beginning, how the "zinger" (or HF "content") from OS files improve musicality dramatically. That's why in A/B, piano is so dynamics and musical in OS file.
Strange for me is between C/D, the one with more HF "content" (which is C or OS file) sounds more "emotional". Somehow, this HF content gives an effect that makes me feel the emotion, not only the vocal but the music.
Between E/F, because I cannot hear big difference in musicality and emotion, I picked based on what I think is more "correct".
With all of these, assuming that Mooly's test is adequate or representative enough (well, nothing is perfect) then I believe that NOS has a tendency to be more accurate? What a surprise!!
But like Mooly said, this is low quality DAC where Over Sampling implementation could be inferior to today standard with better DAC ICs.
From listening to the files, and then knowing what they are, I have this new finding that mostly reinforce what I have believed about NOS (Non-Over Sampling) DAC.
As has been well known (by their fans of course), NOS sound so natural, and I agree with that. But I also realized that NOS is less musical and brings less enjoyment than OS DACs. I didn't know what caused this, but mainly I blame the not so good specification of the DAC IC in its dynamic ability (DNR or SNR).
But from this experience in this thread conducted by Mooly, and this thread:
http://www.diyaudio.com/forums/full...-comparison-3in-5in-drivers-round-2-a-21.html
I found out that naturalness, honesty, truth, is not always preferable. I knew that ScanSpeak 10F was the best driver from listening, but to be honest, I like to listen to cheap Aluminum cone Peerless (playing in the background) when I browse the internet. This Peerless has big overshoot and long ringing in the impulse response, but mainly it has elevated HF response. And we all know how some people deliberately add non-musical HF content to "modify" the sound so is subjectively become better.
In this test from Mooly I could feel from the beginning, how the "zinger" (or HF "content") from OS files improve musicality dramatically. That's why in A/B, piano is so dynamics and musical in OS file.
Strange for me is between C/D, the one with more HF "content" (which is C or OS file) sounds more "emotional". Somehow, this HF content gives an effect that makes me feel the emotion, not only the vocal but the music.
Between E/F, because I cannot hear big difference in musicality and emotion, I picked based on what I think is more "correct".
With all of these, assuming that Mooly's test is adequate or representative enough (well, nothing is perfect) then I believe that NOS has a tendency to be more accurate?
What a surprise!! But like Mooly said, this is low quality DAC where Over Sampling implementation could be inferior to today standard with better DAC ICs.
There is no right and wrong, both are valid set ups in the audio world. Which you prefer is a subjective choice.
The LPF is a crucial part of any replay chain, and of course this modification does not carry that problem forward to include a redesigned filter. I wonder whether many of the reports of the NOS mod performing better are being based on the significantly higher level the NOS version puts out. A typical listener not understanding or realising might well feel it is more dynamic (louder) because that is how it would appear for a given volume setting.
The LPF is a crucial part of any replay chain, and of course this modification does not carry that problem forward to include a redesigned filter. I wonder whether many of the reports of the NOS mod performing better are being based on the significantly higher level the NOS version puts out. A typical listener not understanding or realising might well feel it is more dynamic (louder) because that is how it would appear for a given volume setting.
But I found here in this test, and also in real experience that NOS is not more dynamic. It is less "defined" than OS. But it is "different", and most of us agree that this difference is in naturalness.
Please note carefully that, when people have opinion about sounds, that opinion is built based on long time listening with different songs and of course various volume settings. And many subjectivists use this fact as evidence of "uncontrolled" test...
But note again, that the opinion is not based on ABX or something like it, so a difference in volume level is really irrelevant to how the opinion has been built.
What you call "naturalness", Jay, is actually the absence of unpleasantness - digital not quite working right can add an edge of uncomfortableness to the listening experience, and even though I have not explored any of this NOS type of playback, that seems to be one thing that it gets right - less "uncomfortable" distortion injected.
However, the real point is that 'correct' playback is both dynamic, and natural; it's never about one, or the other; the two qualities should co-exist. Personally, I take the "dynamic" first, and then tweak until it's also "natural" - without losing anything on the way !!
Agree.
I took both ways, to see how far each method can go. Unfortunately "dynamic" always win. Because it is the main ingredient for musical enjoyment (without fatiguing distortion of course).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.