Very interesting, thanks a lot for sharing!
I expected the differences to be subtle so I thought initialy "better wait until at home to listen on a reference system". Curiosity being what it is, I couldn't resist to give them a go on my laptop at work, through its integrated LS, where ever these might be. Hardly HIFI. And I heard differences. So I thought "just give it a try though small HP at least to have some bass extension". Koss to 50E HP - Hardly HIFI, of course no image depth, but after 20 min casual listening here are my very own impressions.
a files sound indeed like the reference and well balanced. None of the others sounds as right, so at least that is good and as expected. I must confess though that I had the feeling a sounded a tad louder than the others - perhaps more information, perhaps the nature of the hiss, perhaps the impact, who knows. Marginal but quite there on the piano deep notes and hiss.
d files sound close to a files, there is some rightness and no sound effect or distorsion. BUT there is something I don't like and that isn't with a / the original: it sounds flat and dry somewhere, it is missing the original flow and decay / note ending, it sounds somewhat mechanic, like played by robots, less natural. It has vacuum between notes: not air / real life background noise. It has though real deep bass vs the others DACs. d is probably perfect à la ASR but I couldn't live with it as I would miss the flow, the spirit, the enjoyment of music. Reminds me of the 500E chinese DACs I tried last year: they measured perfectly, but sounded less natural and more fatiguing than my old DAC and far worst than D70S, where things started to get aacceptable. Bad luck for me...
b is a pleasant, well, little sound box. It adds glare/ glow,/ reverb (and distorsions), but in a "nice manner". It isn't neutral, by far, it adds flesh and corpus to sounds and instruments, and also some depth in the soundstage. It sounds like a lot of negative H2 from where I come LOL, but with a truncated bass. It flows though, it doesn't sound out of place without direct comparison, it is relaxed and centered in the mids. Decays are very pleasant. Treble misses spark and extension though. I couldn't live with it: it is not neutral enough to my ears. I would like the positives, but without the negatives and exagerations that are overblowing the mids and distording the original sound. And more details, more air, more speed especialy in the treble... quite a bit more is required for my listening pleasure.
c is much better than b. It still has some of that pleasant glare/glow/reverb, but in a less exagerated way IMHO, in fact it sounds right on its own. It sounds faster but still relaxed. It sounds less muffled in the treble and less distorted generaly speaking. It has air and depth, it has impact thanks to speed and slam. It is not exactly on a par with d regarding neutrality and details, but that is very close and not a disaster. Lower bass is much better than b, treble aswell, that is now getting acceptable (but I suspect though my ref system I would hear some attenuation, but no drama). Where both c and d units aren't on a par is clearly flow, natural non fatiguing sound, fun, rightness and consistency regarding music and not just figures.
c is much better on all this and much more enjoyable than d. I could live with c as a small DAC in a budget system. I wouldn't miss really anything and would just enjoy music, knowing it is "true and detailled enough", while never boring and always entertaining. Perhaps also forgiving on lesser materials, who knows... interesting little unit!
I would love to try c or some similar ready to go units vs my DAC and the D70S. Or at least to have more of these excellent comparisons ran by George on vocals (female Jazz, Verdi or Delibes opera, pleaaaaase) and also on big orchesters (opera such as Verdi or Wagner, pleaaaase) to make sure nothing collapses when it gets really complex and tough.
All IME... back to work, the 20 min brake I indulged myself is over 🙂
Claude
I expected the differences to be subtle so I thought initialy "better wait until at home to listen on a reference system". Curiosity being what it is, I couldn't resist to give them a go on my laptop at work, through its integrated LS, where ever these might be. Hardly HIFI. And I heard differences. So I thought "just give it a try though small HP at least to have some bass extension". Koss to 50E HP - Hardly HIFI, of course no image depth, but after 20 min casual listening here are my very own impressions.
a files sound indeed like the reference and well balanced. None of the others sounds as right, so at least that is good and as expected. I must confess though that I had the feeling a sounded a tad louder than the others - perhaps more information, perhaps the nature of the hiss, perhaps the impact, who knows. Marginal but quite there on the piano deep notes and hiss.
d files sound close to a files, there is some rightness and no sound effect or distorsion. BUT there is something I don't like and that isn't with a / the original: it sounds flat and dry somewhere, it is missing the original flow and decay / note ending, it sounds somewhat mechanic, like played by robots, less natural. It has vacuum between notes: not air / real life background noise. It has though real deep bass vs the others DACs. d is probably perfect à la ASR but I couldn't live with it as I would miss the flow, the spirit, the enjoyment of music. Reminds me of the 500E chinese DACs I tried last year: they measured perfectly, but sounded less natural and more fatiguing than my old DAC and far worst than D70S, where things started to get aacceptable. Bad luck for me...
b is a pleasant, well, little sound box. It adds glare/ glow,/ reverb (and distorsions), but in a "nice manner". It isn't neutral, by far, it adds flesh and corpus to sounds and instruments, and also some depth in the soundstage. It sounds like a lot of negative H2 from where I come LOL, but with a truncated bass. It flows though, it doesn't sound out of place without direct comparison, it is relaxed and centered in the mids. Decays are very pleasant. Treble misses spark and extension though. I couldn't live with it: it is not neutral enough to my ears. I would like the positives, but without the negatives and exagerations that are overblowing the mids and distording the original sound. And more details, more air, more speed especialy in the treble... quite a bit more is required for my listening pleasure.
c is much better than b. It still has some of that pleasant glare/glow/reverb, but in a less exagerated way IMHO, in fact it sounds right on its own. It sounds faster but still relaxed. It sounds less muffled in the treble and less distorted generaly speaking. It has air and depth, it has impact thanks to speed and slam. It is not exactly on a par with d regarding neutrality and details, but that is very close and not a disaster. Lower bass is much better than b, treble aswell, that is now getting acceptable (but I suspect though my ref system I would hear some attenuation, but no drama). Where both c and d units aren't on a par is clearly flow, natural non fatiguing sound, fun, rightness and consistency regarding music and not just figures.
c is much better on all this and much more enjoyable than d. I could live with c as a small DAC in a budget system. I wouldn't miss really anything and would just enjoy music, knowing it is "true and detailled enough", while never boring and always entertaining. Perhaps also forgiving on lesser materials, who knows... interesting little unit!
I would love to try c or some similar ready to go units vs my DAC and the D70S. Or at least to have more of these excellent comparisons ran by George on vocals (female Jazz, Verdi or Delibes opera, pleaaaaase) and also on big orchesters (opera such as Verdi or Wagner, pleaaaase) to make sure nothing collapses when it gets really complex and tough.
All IME... back to work, the 20 min brake I indulged myself is over 🙂
Claude
Thanks to you again, George.
TBH, given the straightforward findings, I do not expect much more on my LS, apart of course from the 3D definition which works best with my magnetostatic LS (I am not good with HP and head imaging). And the deep bass of course.
It would be nice to read what other hear or prefer, especialy given your went the amazing trouble to provide us with such revelating files - that say much more than my 1000 words. That is so nice of you. And I am sure some comments could also be helpful for development - sometimes another pair of ears helps redefining the line, perhaps - my case when I am head down for too long LOL.
Enjoy all these units and thanks again
Claude
TBH, given the straightforward findings, I do not expect much more on my LS, apart of course from the 3D definition which works best with my magnetostatic LS (I am not good with HP and head imaging). And the deep bass of course.
It would be nice to read what other hear or prefer, especialy given your went the amazing trouble to provide us with such revelating files - that say much more than my 1000 words. That is so nice of you. And I am sure some comments could also be helpful for development - sometimes another pair of ears helps redefining the line, perhaps - my case when I am head down for too long LOL.
Enjoy all these units and thanks again
Claude
Claude, there were those 10 seconds at the end of the piano piece that made me jump up from the couch. "Wait a minute!" (that pedal action).
That, through the speakers
George
That, through the speakers
George
Last edited:
Ah... so I have to give them a listen! I am not sure I listened to the end of each piece TBH, more concentrating on comparisons between bits
What were your findings / what made you jump like this?
Fascinating,...
What were your findings / what made you jump like this?
Fascinating,...
Thanks for the files! But... changing level digitally degrades the recording considerabely, so unless all files are changed in exactly the same way they will show different levels of degradation. Better to match analog record level prior to ADC. Or at least record with much higher resolution, level match, and downsample to the final format. But in my experience, best is to record directly in final resolution and apply absolutely no changes to the files....
Yes, there will always be a “But…”. It is inevitable.
What I have done:
To each DAC input I have fed the two music files (guitar and piano extracts) plus a 60 seconds 1kHz 0dBFS file. The recording of the analog output of each DAC was stored as 44.1 Linear PCM, 24bit resolution, triangular PDF dithering.
The peak amplitude (statistics) of each of the three files (and channel) from each DAC output was written down on paper.
I reviewed the peak amplitude of all the three 1kHz recordings and I opted for a slightly lower level as the post processing peak amplitude target for each of them, so that all three will receive a digital negative gain.
The number of -dB each of the pilot 1kHz file was assigned to (for to all reach the same dBFS target value), was applied also the other two music files from the same DAC output.
From my notes, the data is this:
using ocenaydio statistics/peak amplitude
original DAC out recording files result (L/R) dBFS
pilot original Abbado -0.84/-0.81
pilot opamp-less Abbado -1.37/-1.38
pilot SMSL -1.04/-1.05
----
guitar original Abbado -0.80/-0.78
guitar opamp-less Abbado-1.25/-1.26
guitar SMSL --1.05/-1.06
----
piano original Abbado -3.68/-3.66
piano opamp-less Abbado -4.14/-4.14
piano SMSL -3.85/-3.86
----
Decision: Target peak amplitude: -1.5 dBFS
using ocenaudio Effect/Amplitude/Gain
original Abbado files: gain factor -0.66dB
opamp-les Abbado files: gain factor -0.13dB
SMSL files: gain factor -0.46dB
----
Result: same level files peak amplitude (L/R) dBFS
pilot original Abbado -1.50/-1.51
pilot opamp-less Abbado -1.50/-1.51
pilot SMSL -1.50/-1.51
max difference 0.00dB
----
guitar original Abbado -1.46/-1.44
guitar opamp-less Abbado -1.38/-11.39
guitar SMSL -1.51/-1.52
max difference 0.13dB
----
piano original Abbado -4.34/-4.32
piano opamp-less Abbado -4.27/-4.27
piano SMSL -4.31/-4.32
max difference 0.07dB
This equal level adjustment and the time alignment (trimming the start and end of each recording) were done at 24bit depth. Then files were exported as 44.1 Linear PCM, 16bit resolution, triangular PDF dithering (this is what I uploaded at Dropbox).
Jpk73 I am with you on making the record level adjustment in the analog universe prior to ADC.
Only for achieving x.xx level resolution, a Kelvin Varley voltage divider would be the choice (one for each channel). Alas, I have only one of those.
There would be the -technically justified- option of uploading the unprocessed original DAC output recordings instead.
The decision and responsibility for level matching would then rest with each downloader.
Well, I may do it next time I try such endeavor. 🙂
George
What I have done:
To each DAC input I have fed the two music files (guitar and piano extracts) plus a 60 seconds 1kHz 0dBFS file. The recording of the analog output of each DAC was stored as 44.1 Linear PCM, 24bit resolution, triangular PDF dithering.
The peak amplitude (statistics) of each of the three files (and channel) from each DAC output was written down on paper.
I reviewed the peak amplitude of all the three 1kHz recordings and I opted for a slightly lower level as the post processing peak amplitude target for each of them, so that all three will receive a digital negative gain.
The number of -dB each of the pilot 1kHz file was assigned to (for to all reach the same dBFS target value), was applied also the other two music files from the same DAC output.
From my notes, the data is this:
using ocenaydio statistics/peak amplitude
original DAC out recording files result (L/R) dBFS
pilot original Abbado -0.84/-0.81
pilot opamp-less Abbado -1.37/-1.38
pilot SMSL -1.04/-1.05
----
guitar original Abbado -0.80/-0.78
guitar opamp-less Abbado-1.25/-1.26
guitar SMSL --1.05/-1.06
----
piano original Abbado -3.68/-3.66
piano opamp-less Abbado -4.14/-4.14
piano SMSL -3.85/-3.86
----
Decision: Target peak amplitude: -1.5 dBFS
using ocenaudio Effect/Amplitude/Gain
original Abbado files: gain factor -0.66dB
opamp-les Abbado files: gain factor -0.13dB
SMSL files: gain factor -0.46dB
----
Result: same level files peak amplitude (L/R) dBFS
pilot original Abbado -1.50/-1.51
pilot opamp-less Abbado -1.50/-1.51
pilot SMSL -1.50/-1.51
max difference 0.00dB
----
guitar original Abbado -1.46/-1.44
guitar opamp-less Abbado -1.38/-11.39
guitar SMSL -1.51/-1.52
max difference 0.13dB
----
piano original Abbado -4.34/-4.32
piano opamp-less Abbado -4.27/-4.27
piano SMSL -4.31/-4.32
max difference 0.07dB
This equal level adjustment and the time alignment (trimming the start and end of each recording) were done at 24bit depth. Then files were exported as 44.1 Linear PCM, 16bit resolution, triangular PDF dithering (this is what I uploaded at Dropbox).
Jpk73 I am with you on making the record level adjustment in the analog universe prior to ADC.
Only for achieving x.xx level resolution, a Kelvin Varley voltage divider would be the choice (one for each channel). Alas, I have only one of those.
There would be the -technically justified- option of uploading the unprocessed original DAC output recordings instead.
The decision and responsibility for level matching would then rest with each downloader.
Well, I may do it next time I try such endeavor. 🙂
George
Jpk73 you are welcome.
Your concerns well dully noted.
The management decided I will do the customer service again.
Now the parent file (guitar + piano)
https://www.dropbox.com/s/ezx1j6b12k0u6hi/guitar + piano Maurice Ohana parent file.wav?dl=0
The 24 bit unprocessed DAC out recordings:
Original Abbado
https://www.dropbox.com/s/jmo6vsu6thqq8q4/guitar+piano original Abbado DAC out.wav?dl=0
Opamp-less Abbado
https://www.dropbox.com/s/ihmi8dol0u4ehsj/guitar+piano opamp-less Abbado DAC out.wav?dl=0
SMLS
https://www.dropbox.com/s/xa792gaue0nx6ll/guitar+piano SMSL DAC out.wav?dl=0
The lowest peak amplitude file is the opamp-less Abbado.
For to bring the other files to the same peak amplitude level, you ‘ll have to apply the following attenuation to them:
Parent file: 1.30dB
Original Abbado: 0.49dB
SMSL: 0.25dB
I will need the storage space at Dropbox soon, therefore those interested, please download any of the new and the older files by tomorrow.
George
Your concerns well dully noted.
The management decided I will do the customer service again.
Now the parent file (guitar + piano)
https://www.dropbox.com/s/ezx1j6b12k0u6hi/guitar + piano Maurice Ohana parent file.wav?dl=0
The 24 bit unprocessed DAC out recordings:
Original Abbado
https://www.dropbox.com/s/jmo6vsu6thqq8q4/guitar+piano original Abbado DAC out.wav?dl=0
Opamp-less Abbado
https://www.dropbox.com/s/ihmi8dol0u4ehsj/guitar+piano opamp-less Abbado DAC out.wav?dl=0
SMLS
https://www.dropbox.com/s/xa792gaue0nx6ll/guitar+piano SMSL DAC out.wav?dl=0
The lowest peak amplitude file is the opamp-less Abbado.
For to bring the other files to the same peak amplitude level, you ‘ll have to apply the following attenuation to them:
Parent file: 1.30dB
Original Abbado: 0.49dB
SMSL: 0.25dB
I will need the storage space at Dropbox soon, therefore those interested, please download any of the new and the older files by tomorrow.
George
Changing scene.
More complex music plus testing the emotional engagement element (for opera funs only)
The parent file (0:0 Die Walkuere, 3:31 Boris Godunov, 8:45 Don Carlo, 19:11 Rigoletto, 22:51 Norma)
https://www.dropbox.com/s/iddvpc5v4siel4r/voices parent file.wav?dl=0
The 24 bit unprocessed DAC out recordings:
Original Abbado
https://www.dropbox.com/s/56vldhahg8jcar0/voices original Abbado DAC out.wav?dl=0
Opamp-less Abbado
https://www.dropbox.com/s/aqjb9zsre8quapc/voices opamp-less Abbado DAC out.wav?dl=0
SMLS
https://www.dropbox.com/s/07to17mrjny4x4l/voices SMSL DAC out.wav?dl=0
As previously, the lowest peak amplitude file is the opamp-less Abbado.
For to bring the other files to the same peak amplitude level, you ‘ll have to apply the following attenuation to them:
Parent file: 1.30dB
Original Abbado: 0.49dB
SMSL: 0.25dB
Enjoy (long files) 🙂
George
More complex music plus testing the emotional engagement element (for opera funs only)
The parent file (0:0 Die Walkuere, 3:31 Boris Godunov, 8:45 Don Carlo, 19:11 Rigoletto, 22:51 Norma)
https://www.dropbox.com/s/iddvpc5v4siel4r/voices parent file.wav?dl=0
The 24 bit unprocessed DAC out recordings:
Original Abbado
https://www.dropbox.com/s/56vldhahg8jcar0/voices original Abbado DAC out.wav?dl=0
Opamp-less Abbado
https://www.dropbox.com/s/aqjb9zsre8quapc/voices opamp-less Abbado DAC out.wav?dl=0
SMLS
https://www.dropbox.com/s/07to17mrjny4x4l/voices SMSL DAC out.wav?dl=0
As previously, the lowest peak amplitude file is the opamp-less Abbado.
For to bring the other files to the same peak amplitude level, you ‘ll have to apply the following attenuation to them:
Parent file: 1.30dB
Original Abbado: 0.49dB
SMSL: 0.25dB
Enjoy (long files) 🙂
George
Attachments
Last edited:
Fascinating, these recordings sound better, than one thinks they could, regarding their age. I know, there are great recordings from that time, yet again it is jaw dropping... Thank you, George for that. Hard to concentrate on the differences, as the music is so beautiful and beautifully recorded. Cheers
First revision of the I2S splitter boards being populated (using a dozen 74HC595s). Next, I need to give them a test at higher sample rates, see how they fare.
At first they worked only at 44.1k and 96k. Setting the output sample rate to 192k produced a pile of crackles and pops. Turned out that running the logic at 5V introduced an asymmetrical clock because my BCK source is powered at 3.3V. Reducing the logic supply voltage solved this, now it works fine at 192k. I haven't the means to test it any faster but I feel 4X OS is good enough.
The I2S splitter has facilitated a long-awaited development of Celibidache - the balanced version, which I've given the name 'Mravinsky' to. I had PCBs made for this a couple of months ago but those still used the opamp (LT1028) output stage. I decided I wanted to try for an opamp-less analog stage for balanced Celibidache so I have hacked one together based on one I designed earlier for Abbado. Its using a Sziklai pair (PNP-NPN) rather than the darlington pair I used on Dorati and has a passive low-pass filter to generate the NOS droop correction. The other significant change here is I'm over-currenting the DAC chips by around 50% by tweaking up their pin7s. The end result is a dual-mono setup equivalent to having 60 stereo DAC chips.
What's still to be figured out is - will all that extra stuff on the red protoboard fit on the main PCB?
What's still to be figured out is - will all that extra stuff on the red protoboard fit on the main PCB?
Hi Richard,
Is the balanced version a noticeable improvement? Will the I2S splitter be compatible with all DACs, including Dorati?
Is the balanced version a noticeable improvement? Will the I2S splitter be compatible with all DACs, including Dorati?
Hi Vunce,
I just tried to persuade wifey it was a step-up but as I haven't got the original Celibidache available for a direct comparison, she's not convinced so far. My instincts say it is but that's relying on my memory.
The I2S splitter is universal, any DAC can be used with it. Dorati is especially suited to it with its passive I/V.
I just tried to persuade wifey it was a step-up but as I haven't got the original Celibidache available for a direct comparison, she's not convinced so far. My instincts say it is but that's relying on my memory.
The I2S splitter is universal, any DAC can be used with it. Dorati is especially suited to it with its passive I/V.
The only male instinct worth noting is that of turning against his rival. Females on the other hand live by their instincts 😉
What have you done to the pin 7 Abrax?
George
Best read my blog post to get up to speed : https://www.diyaudio.com/community/threads/tda1387-and-tda1545a.384224/post-7030615