The sound quality of DSD?

[AnalogSteph]

Can you replace AD745 opamp now (NRND)? Without paralleling new, more noisy IC.
"ULTRALOW NOISE PERFORMANCE
2.9 nV/Hz at 10 kHz
0.38 uV p-p, 0.1 Hz to 10 Hz
6.9 fA/Hz Current Noise at 1 kHz"

(Expanded the specs there for ya.)

That really depends on the circuit.

The AD745 is a bit of an odd beast, being about the lowest-noise JFET input opamp I know of. (OPA1656 gets close but, being MOSFET input, has more 1/f noise.) For very specific applications that benefit from its low voltage and current noise and where little to no common-mode voltage is present, it's a great part no doubt. MM phono preamps or the transimpedance amplifiers for hydrophones mentioned in the datasheet would still be good applications. Not sure I would use one in a DAC I/V without at least some prior filtering any more, slew rate is only 12.5 V/µs and that's not too hot for a FET input part.

The reason why the part isn't a lot more common comes in the INPUT IMPEDANCE section. Input capacitance is 18 pf common-mode / 20 pF differential! (Your average JFET input part has about 4 pF.) And that, I am sure, includes plenty of common-mode nonlinearity thereof. If you were to test this part Samuel Groner style, input impedance linearity and common-mode linearity would probably look really bad. In a lot of common audio applications involving lowish-gain non-inverting circuits at highish levels, it would probably rather stink. And I can't imagine it ever was that cheap.

 

[mansr]

Ni!

 

[Esotechnik]

That one is spec'd at -108 dB THD+N, wonder what they used in terms of ADCs? AK5390, CS5390? The CS5394 wouldn't even have been out until late '96, I think, neither would the AK5391.

AK5391 is 24 bit, not 20.
CS5390 has too slow anti-aliasing filter for precision analog capture.
Maybe chips is selected and trimmed dual AD1879 in mono-mode. This is the latest theoretically correct delta sigma audio-ADC to date.
PCM1760+DF1760 has lower THD+N (-90...-92dB).

"The AD1878/AD1879 consists of two integrated circuits in a single ceramic 28-pin DIP package. The modulators and reference are fabricated in a BiCMOS process; the decimator and output port, in a 1.0 µm CMOS process. Separating these functions reduces digital crosstalk to the analog circuitry. Analog and digital supply connections are separated to further isolate the analog circuitry from the digital supplies."

"They provide linear phase and a narrow transition band that permits the digitization of 20 kHz signals while preventing aliasing into the passband
even when using a 44.1 kHz sampling frequency. Passband ripple is less the 0.001 dB, and stopband attenuation exceeds 115 dB."

"Due to the linear phase properties of FIR filters, the group delay variation, or differences in group delay at different frequencies is zero."

Expanded the specs there for ya

square root symbols have not been copied

 

[watchnerd]

Today I was comparing the same file using DoP and straight PCM.

At first I thought the DoP had a ton more air, but then I realized it was just playing about 10 dB louder due to fixed volume.

 

[Esotechnik]

iii) totally outmoded technologically

PCM 44/16 is too?

Due to EMI problems (RF noise) at same 44 kHz samplerate "new", hi-res compatible digital filter SM5847 with DAC AD1862NJ subjectively sounds worse than old SM5842 with the same algorithms, but with a different technology.

SM5847, 3-5V supply voltage: Maximum frequency 37 MHz (5V), "hard switching" logic.
SM5842: maximum frequency 21.2 MHz (at maximum fs = 55.2 kHz), 5.0±0.25 V supply, "soft switching".

If you don't have a broadband spectrum analyzer like R&S FSC3, you'll be blindly wondering where the significant bits have gone.

For example, see spectrum graphs of voltage on power bus of USB to I2S board ("Flamenco"), with software Aplayer and 352 kHz upsampling:
http://forum.vegalab.ru/attachment.php?attachmentid=373389

Direct play, w/o upsampling:
http://forum.vegalab.ru/attachment.php?attachmentid=373391

 

[Pluto]

Can you replace AD745 opamp now (NRND)? etc. etc. etc.

I wouldn't know. These are aspects I ask my maintenance engineer to deal with!

Ask me about sound quality or editing – I can cope with those!

 

[Pluto]

PCM 44/16 is too?

I really wish I understood what you were on about, and the point you are trying to make.

 

[Thomas savage]

Umm, purgatory for many of You is going to look a lot like this thread.

Ouch lol

 

[Jimbob54]

He's back! And he brought a shrubbery!

Ball is in your court. *runs for popcorn*

 

[BDWoody]

Due to EMI problems (RF noise) at same 44 kHz samplerate "new", hi-res compatible digital filter SM5847 with DAC AD1862NJ subjectively sounds worse than old SM5842...

Can you point me to this research please?

I'm assuming the subjective tests you are relying on were controlled? Pretty meaningless otherwise, as I'm sure you know.

 

[bennetng]

How is 0dBFS for DSD determined?

No-brainer for PCM.

But what sets the limit (full scale) for DSD?

An experiment, could be encoder-dependent:

The clipping limit seems frequency dependent. In the range of low noise floor (purple to blue) DSD doesn't clip at 0dBFS, at higher frequencies the ~60kHz noise cluster interacts with the sweep, and caused clipping even at -6dBFS PCM level.

 

[bennetng]

The first and final SBMD test done by a lay person:

• Convert CD FLAC files (comprising of sea waves, cicadas, popular music, classical music) to readable WAV files (use Audacity or sth) and then upconvert to 2.8MHz DSF files using VAIO DSD Direct upsampler v 2.0.02 of 2008 from VAIO VGN-TT2 series.
• Downconvert using Sony SBM Direct on a VAIO VGN-FE4 PC using both 24 and 16 bit options
• Analyse originals and resulting wav files using wavspectra (efu) https://efu.jp.net/soft/ws/ws.html

Conclusions made:

As the spectrums are identical, DAADDA using DSD and SBMD makes no difference except at the highest frequencies which I did not compare in detail.

Subjective listening shows little discernible difference – unless a highly trained ear in professional environments is present, doing the DAADDA is worthless?

Perhaps recording DSD from analog may introduce the DSD Recorder's sound character into the sound, and it is highly likely that Sony's ADA-7000R in Tokyo is used for that reason. Unfortunately recording it as DSD at 1 times speed is simply not logical.

Sadly, some things are only true in the Sony Music studio but not a lay person's house full of Sony VAIO PCs.

Noise spectrum of Sony Super Bit Mapping:
http://audio.rightmark.org/lukin/dither/dither.htm

So about 6-10dB noise reduction below 14kHz, when compared to flat dithering. Can't say it is better or worse than others but seems a reasonable choice for matching the noise floor of studio equipment in that era.

 

[Esotechnik]

Can you point me to this research please?

Only in Russian, uncontrolled listening without measurements. Filters DF1704/1706, SM5842/5847 were compared in 2001.
That expensive DAC contained IC SM5842, EPM7064SLC44-10, 74ABT574, ISO150, AD1862, AD811, AD843:
https://sc33595ac73a18eec.jimconten...089387/module/8395308893/name/DAC_Lynx7V2.pdf

The same author (not me) in 2005 use the MAX7000 series FPGA (EPM7128) with 15ns switching speed instead of 10ns, because with the slow grade IC power supply noise level in the 10GHz band was 2-3dB less.
http://sc33595ac73a18eec.jimcontent...6206019/module/8395700193/name/DAC_Lynx24.pdf

Non-audio, general proof link:
https://www.intel.com/content/dam/www/programmable/us/en/pdfs/literature/ds/archives/m7000.pdf

"The output buffer for each MAX 7000E and MAX 7000S I/O pin has an adjustable output slew rate that can be configured for low-noise or high-speed performance. A faster slew rate provides high-speed transitions for high-performance systems. However, these fast transitions may introduce noise transients into the system. A slow slew rate reduces system noise, but adds a nominal delay of 4 to 5 ns."

Speed Grades 5/6/7/10/12/15/20ns are available.
15ns - better for audio, than 10.

EPM7128STC100-15 "Stock: Not Available" (mouser)

 

[BDWoody]

I'm assuming the subjective tests you are relying on were controlled? Pretty meaningless otherwise, as I'm sure you know.

Only in Russian, uncontrolled listening without measurements

Ahhh...so, in other words no.

 

[mcdonalk]

Through my system, and to me, the attached productions sound superior to any other of my classical recording from my collection of CD's and PCM downloads:

They all happen to be DSD256 files purchased and downloaded from nativedsd.com. This web site claims that their products are all natively recorded in DSD format (hence the web site title).

I also purchased the CD version of the Prokofiev production for comparison. I am not going to state my observations on how these DSD256 recordings sound superior, or how the Prokofiev CD compares to the DSD256 version, because I would not be believed here or taken seriously. Furthermore, some on this scientific forum would probably ridicule me outright, and I am not interested in that. I am here just to recommend these recordings for those who are curious.

 

[danadam]

They all happen to be DSD256 files purchased and downloaded from nativedsd.com. This web site claims that their products are all natively recorded in DSD format (hence the web site title).

"Original Recording Format" in "Additional Information" tab says DXD (so PCM, not DSD) for Franck and Stravinsky and DSD64 for Haydn and Prokofiev.

 

[watchnerd]

"Original Recording Format" in "Additional Information" tab says DXD (so PCM, not DSD) for Franck and Stravinsky and DSD64 for Haydn and Prokofiev.

So a Native DSD website carries recordings originally made in DXD?

That seems contradictory to their name...

 

[mansr]

So a Native DSD website carries recordings originally made in DXD?

That seems contradictory to their name...

That's what I've been saying for years. I guess there was so little true DSD material available that they had to fudge the definition to keep the business viable.

BTW, converting from one DSD rate to another necessarily involves an intermediate PCM stage.

 

[chorus]

Mcd,

Thanks for sharing the four selections you found so enjoyable to listen to!

 

[watchnerd]

That's what I've been saying for years. I guess there was so little true DSD material available that they had to fudge the definition to keep the business viable.

BTW, converting from one DSD rate to another necessarily involves an intermediate PCM stage.

369 DXD original recording format / 1881 titles = 20% originated as high bit rate PCM.

That's a fair chunk.