My impressions of DSD64 vs DSD128

[freddie40]

My DAC was recently updated to play DSD128 and so I thought I would share my experience.

 

First of all:

DAC: Ayre QB9 DSD (recently updated to play DSD128).

Computer: 2009 Mac Mini (8GB RAM, 124GB Samsung SSD)

USB Cable: Maple Shade Clearlink Plus with 5v wire removed.

Software: JRiver 22

Music: Meghan Andrews Special Event 43. I have both the DSD64 and DSD128 versions.

 

 

Background: I saw Meghan Andrews play twice at 2016’s RMAF. The second time I was 8 feet from her and heard her play “Waiting Around to Die: and “Venus Rising” unamplified. I must admit that I am a big fan and have listened to this album many times.

 

So, how do the 2 versions compare:

Over the last few days I have listened to these 2 songs many different times so that I could honestly convey my opinion of the differences between the DSD64 and DSD128 versions. Today I sat down and listened to every song, individually comparing the 2 versions. Both are very good, but I tend to lean toward the DSD128 version. With every song I could hear a tiny amount of digital hash on the DSD64 version that I just do not hear on the DSD128 version. Not only that but with the DSD128 version I feel that there is more air around the instruments and vocals that give an impression of being a bit more realistic. In other words I really feel that Meghan is in the room with me when I listen to the DSD128 version. “Waiting Artound to Die” is stunning and I highly recommend it.

 

Please don’t get me wrong. Both versions of this album are incredible. However, given the choice I would buy the DSD128 version. In the future I will be buying all albums at the resolution that is closest to the original based upon provenance.

[tboooe]

Thank you for the insights. Curious, did you compare DSD64 upsampled to DSD128 against the DSD128 version you purchased?

 

Btw, great speakers. About 10 years ago I owned to 805, 804, then ultimately the 802D.

[One and a half]

Upsampling has its moments, and depends on the music. I find acoustic music, jazz trios and the like, minimal mixing, lends itself to upsampling. More complexity and the music is lost. I don't like to fiddle with these settings, so tend to play whatever the track sample rate is, with the exception of DSD. DSD128 upsampled in software or the DAC is OK, but any higher, music is lost again, it's mush. Not a great deal in it mind, but if the tools are there, may as well use them.

 

Garbage on the recording, garbage at the speaker. No garbage at Blue Coast originals

[freddie40]

Thank you for the insights. Curious, did you compare DSD64 upsampled to DSD128 against the DSD128 version you purchased?

 

Btw, great speakers. About 10 years ago I owned to 805, 804, then ultimately the 802D.

 

Sorry, I didn't compare up sampled DSD64 to DSD128. I am pretty sure that JRiver does not upsample DSD64 to DSD128 if you choose that option. I typically don't upsample anything.

 

Thanks for the compliment on my sneakers. I love them.

 

Dave

[JSWolf]

Upsampling is fooling you. It's no better than the original bits you started with.

[ted_b]

Upsampling is fooling you. It's no better than the original bits you started with.

 

? Explain please. I'd like to hear your take on upsampling, as you seem to want to steer the OP in a certain direction.

[edtsui]

+1. would like to know why upsampling is no bettr than the original. If so, why is it that most DAC would automaticially upsample musical files.

[Kal Rubinson]

+1. would like to know why upsampling is no bettr than the original. If so, why is it that most DAC would automaticially upsample musical files.

Marketing?

[firedog]

The rationale for upsampling is that it allows use of different and/or more gentle filters That has nothing to do with adding extra information to the "original bits". This is especially true with a program like HQP that allows choice of many filters and modulators.

 

In addition, most DACs work natively at a hi-res rate of some kind, so upsampling before the DAC may improve sound if the DAC is fed it's native sampling rate, with upsampling on the server end.

[JSWolf]

The rationale for upsampling is that it allows use of different and/or more gentle filters That has nothing to do with adding extra information to the "original bits". This is especially true with a program like HQP that allows choice of many filters and modulators.

In addition, most DACs work natively at a hi-res rate of some kind, so upsampling before the DAC may improve sound if the DAC is fed it's native sampling rate, with upsampling on the server end.

 

So your saying that 16/44.1 is not effective because the hardware used does not do it properly and we'd be better off with hi-res audio because that does work better. So basically, upsampling is just to fix a problem with the hardware. Sound a lot silly.

[mansr]

+1. would like to know why upsampling is no bettr than the original. If so, why is it that most DAC would automaticially upsample musical files.

Grated cheese isn't better than a solid lump, but it's a good first step to making a pizza.

[ted_b]

JSWolf, since you are calling these ideas silly, please tell us what your approach is.

[audiventory]

Upsampling is fooling you. It's no better than the original bits you started with.

 

+1. would like to know why upsampling is no bettr than the original. If so, why is it that most DAC would automaticially upsample musical files.

 

 

Upsampling don't improve sound quality as itself. For ideal DAC we can't see and hear difference.

 

It only give ablity use the best DAC's mode (sample rate/bit depth/modulation) between available in the one.

 

Using upper sample rates + steeper (than analog) digital filter give lesser aliases and, hence, lesser intermodulation-distortion products in audible range.

 

Of course, all dpend on implementation. Using higher sample rate guarantee nothing.

[firedog]

So your saying that 16/44.1 is not effective because the hardware used does not do it properly and we'd be better off with hi-res audio because that does work better. So basically, upsampling is just to fix a problem with the hardware. Sound a lot silly.

 

No, that has nothing to do with what I wrote. Try to understand instead of just trying to argue.

 

Did I say 16/44.1 "is not effective"? No, and I don't even know what that means. But if you're asking, I think 16/44.1 can sound wonderful. In some setups, using upsampling and changing filters can make it sound even better.

 

The only silly thing is your approach, and your refusal to try and understand how digital audio actually works.

 

Try to read up on filtering and why there are different kinds of filters, and why upsampling and filtering might improve the sound of what you hear. Until you realize you have something to learn, you are only arguing with points that the rest of us aren't making, but you imagine that we are.

[JSWolf]

Upsampling don't improve sound quality as itself. For ideal DAC we can't see and hear difference.

It only give ablity use the best DAC's mode (sample rate/bit depth/modulation) between available in the one.

Using upper sample rates + steeper (than analog) digital filter give lesser aliases and, hence, lesser intermodulation-distortion products in audible range.

Of course, all dpend on implementation. Using higher sample rate guarantee nothing.

 

So then we could very well be better off with hi-res audio because according to you, the hardware handles it better than CD audio.

[audiventory]

So then we could very well be better off with hi-res audio because according to you, the hardware handles it better than CD audio.

 

It depend on:

 

- how recorded/mixed/mastered,

 

- how resampled/non-resampled,

 

- how implemented analog filter of DAC,

 

- frequency and non-linear distortions of DAC.

 

Sometimes pro-audio card at 48 kHz may sound better than home one at 192 kHz.

 

But high sample rates give more abilities for achieving better result in audible range.

[wisnon]

So then we could very well be better off with hi-res audio because according to you, the hardware handles it better than CD audio.

 

The hirez version may be more poorly mixed than the RBCD version, hence you would then be better off up sampling the RBCD version than buying the hirez version.

[Roy Boy]

I have an opinion on DSD64 vs DSD128 from a different viewpoint. I bought a Korg DS-DAC-10R to experiment with ripping vinyl.

I have a Lumin A1 for streaming. My analog front end is a rebuilt Techinics SP10-MkII with Basis Vector arm and Soundsmith Zephyr MIMC.

My objective was to see if my vinyl rips could get sonically as close to playing the vinyl that I could be satisfied streaming the ripped vinyl vs playing the record. So I ripped at 24/96, 24/192, DSD64 and DSD128. One caveat, since the Lumin A1 does not support DSD128 I used Minimserver to transcode to 24/384. This transcoding was applied to both DSD64 and DSD128.

For me, the results were that the DSD128 was best and equal to (and potentially better) than playing vinyl. Before I figured out the transcoding part, I was comparing DSD64 to PCM 24/192...I was leaning towards the 24/192...The DSD64 seemed a little softened compared to playng the vinyl. I did not spend a long time evaluating the DSD64 vs DSD128 since by then I was really deciding it is

was worth the extra disk space to rip at DSD128...Since disk space is very cheap in terms of audio file sizes, it did not take much to convince me to rip at DSD128.

One last comment/observation. The area where I think the DSD128 rip is possibly superior to playing the vinyl is in the area of bass

reproduction. When I was rippng, I was going from phono pre-amp directly into the Korg. So no feedback path from speaker to turntable.

I have two Seaton Subs and so it is not too difficult to create a feedback situation with bass heavy tracks when playing vinyl.

[audiventory]

I have an opinion on DSD64 vs DSD128 from a different viewpoint. I bought a Korg DS-DAC-10R to experiment with ripping vinyl.

I have a Lumin A1 for streaming. My analog front end is a rebuilt Techinics SP10-MkII with Basis Vector arm and Soundsmith Zephyr MIMC.

My objective was to see if my vinyl rips could get sonically as close to playing the vinyl that I could be satisfied streaming the ripped vinyl vs playing the record. So I ripped at 24/96, 24/192, DSD64 and DSD128. One caveat, since the Lumin A1 does not support DSD128 I used Minimserver to transcode to 24/384. This transcoding was applied to both DSD64 and DSD128.

For me, the results were that the DSD128 was best and equal to (and potentially better) than playing vinyl. Before I figured out the transcoding part, I was comparing DSD64 to PCM 24/192...I was leaning towards the 24/192...The DSD64 seemed a little softened compared to playng the vinyl. I did not spend a long time evaluating the DSD64 vs DSD128 since by then I was really deciding it is

was worth the extra disk space to rip at DSD128...Since disk space is very cheap in terms of audio file sizes, it did not take much to convince me to rip at DSD128.

One last comment/observation. The area where I think the DSD128 rip is possibly superior to playing the vinyl is in the area of bass

reproduction. When I was rippng, I was going from phono pre-amp directly into the Korg. So no feedback path from speaker to turntable.

I have two Seaton Subs and so it is not too difficult to create a feedback situation with bass heavy tracks when playing vinyl.

 

By noise level of D64 in audible range from «better than CD» to «about 24 bit PCM».

 

D128 is about 32-bit float.

 

D256 and more have wider, than D128, band with same noise level.

 

Noise level defined by sigma-delta modulator mainly.

 

D64 is most complicated format for achieving quality (lower noise level).

[GUTB]

I admit I don't understand all the technical reasons behind it -- but those DSD128 recordings from NativeDSD sound astoundingly good through my Sabre DAC via I2S. There's no comparison with PCM. I don't know if that's because of bypassing the SRC stage or some combination of factors but I hear what I hear...

 

 

Sent from my iPad using Computer Audiophile

[mansr]

By noise level of D64 in audible range from «better than CD» to «about 24 bit PCM».

 

D128 is about 32-bit float.

 

24-bit integer and 32-bit float have the same precision.

[audiventory]

24-bit integer and 32-bit float have the same precision.

 

There about 40...60 dB difference of noise floor.

[mansr]

There about 40...60 dB difference of noise floor.

 

Well, for very small values you might consider the minimum exponent as providing another 7 bits, or roughly 40 dB. As the signal gets stronger, this advantage goes away. When the peaks are near ±1, the precision is the same as for 24-bit integer.

[audiventory]

Well, for very small values you might consider the minimum exponent as providing another 7 bits, or roughly 40 dB. As the signal gets stronger, this advantage goes away. When the peaks are near ±1, the precision is the same as for 24-bit integer.

 

However, in audio we consider errors of quantization, i.e. noise floor.

 

Let take as refer signal -80 dB.

 

For 24 bit it has 64 dB (-80-144) signal/noise difference (it is not SNR).

 

For 32 bit float it has 90 dB (-80-170).