SMSL VMV D3 Review (R2R DAC)

[voodooless]

Maybe wrong thread for this but allow me a quick question:
The R2R/NOS issue with the wobbly high frequency sine near niquist - does it disappear when using let's say 384khz for an 44,1 signal? Or is 44,1 signal played at 44,1 sample rate with a filter still better even then?

You mean by using software upsampling? Yes that works by shifting the problem up in frequency where audibility is less of an issue. But one can hardly call that NOS.. an oversampling DAC does it for the exact same reasons

 

[gvl]

Maybe wrong thread for this but allow me a quick question:
The R2R/NOS issue with the wobbly high frequency sine near niquist - does it disappear when using let's say 384khz for an 44,1 signal? Or is 44,1 signal played at 44,1 sample rate with a filter still better even then?

Digital filter essentially upsamples the incoming signal to a higher rate, so these approaches are similar and will smooth out the waveform. Digital filter also suppresses frequencies above Nyquist, or should anyway but the same can/should be done when using external upsampling. In theory you can achieve better filtering quality externally, say on a PC, because of more processing power there than on a DAC chip, in practice however this makes little difference because gains are small.

 

[lewdish]

I think for R2R its far from the worst but given its price its rather overblown for what it is~

 

[amirm]

However, for the time when they were designed for 16 bit CD playback, doesn’t this meet reasonable standards?

This DAC came out in 1998. So not very old and certainly in the era of recording/playback in 24 bits. Otherwise it would have been a 16 bit DAC.

 

[aj625]

A simple test indicates why at all R2R can sound a bit fuller. Strip 16bits of a 24bit file, pad those 16bits with zero and now play this file which has only effectively 8bit information through sota DS dac. It will sound a bit fuller like R2R dac. reason is simple less information are easier to process by brain which can sound a bit fuller. So may be low linearity measurements of r2 are responsible for that typical r2r sound, if at all it has any sound signature.

 

[alpha_logic]

It will sound a bit fuller like R2R dac. reason is simple less information are easier to process by brain which can sound a bit fuller.

Not sold on that 'brain' statement - sources, studies? I think the obvious explanation is that most old R2R DAC chips resolve less treble, thus sounding 'fuller'. Maybe that's what you're trying to say, and I misunderstood.

 

[aj625]

Not sold on that 'brain' statement - sources, studies? I think the obvious explanation is that most old R2R DAC chips resolve less treble, thus sounding 'fuller'. Maybe that's what you're trying to say, and I misunderstood.

That's why i said it's a simple test. One can very easily do it on a free daw like audacity. And yes r2r dacs resolve less finer details "including in treble frequencies " due to low linearity. You see lower bits represent low level details and those lower bits are not converted accurately due to poor linearity by r2r dacs.

 

[ousi]

A simple test indicates why at all R2R can sound a bit fuller. Strip 16bits of a 24bit file, pad those 16bits with zero and now play this file which has only effectively 8bit information through sota DS dac. It will sound a bit fuller like R2R dac. reason is simple less information are easier to process by brain which can sound a bit fuller. So may be low linearity measurements of r2 are responsible for that typical r2r sound, if at all it has any sound signature.

Coincidentally, I have to work with WAV PCM files directly for my work recently. If one does what you said, you will simply get - either total silence or a chopped off waveform with anything over 8-bit signed integer (over 127 or smaller than -128) since RIFF WAV samples are usually in little endian. So I don't quite understand what you mean by padding 16bits with zero of a 24-bit file. If my guess is right, you want to do a conversion. In this case you'll need to convert 24-bit signed integer (−8,388,608 to 8,388,607) into 8-bit signed integer (-128 to 127) proportionally.

 

[ousi]

That's why i said it's a simple test. One can very easily do it on a free daw like audacity. And yes r2r dacs resolve less finer details "including in treble frequencies " due to low linearity. You see lower bits represent low level details and those lower bits are not converted accurately due to poor linearity by r2r dacs.

Sounding "fuller" is easy to do, you don't even need R2R DACs as anything will do. Turn up the mid-range, and mid-bass section, turn down the top end a bit. That's what "fuller" sound like according to reviewers.....

 

[ousi]

This DAC came out in 1998. So not very old and certainly in the era of recording/playback in 24 bits. Otherwise it would have been a 16 bit DAC.

Yea! Back then 24bit 96kHz was the buzz. The DCS Elgar was one of the first that does 24/96 in 1996 (and in 2000, they "upgraded" it to support 24/192).

 

[aj625]

Coincidentally, I have to work with WAV PCM files directly for my work recently. If one does what you said, you will simply get - either total silence or a chopped off waveform with anything over 8-bit signed integer (over 127 or smaller than -128) since RIFF WAV samples are usually in little endian. So I don't quite understand what you mean by padding 16bits with zero of a 24-bit file. If my guess is right, you want to do a conversion. In this case you'll need to convert 24-bit signed integer (−8,388,608 to 8,388,607) into 8-bit signed integer (-128 to 127) proportionally.

not correct. by removing the last 16bits (without dither, as dither surprisingly still retains lot more low level information even if you remove last 16 bits ) and replacing with zero retains "most of the music content" and you lose only low level information. consequently in low music passages you could listen noise where there was some information in the original file. for doing that open 24 bit file in daw, go to bit depth reduction, select 8 bit without dither. after that go again to bit depth and select 24 bit. this will pad the last 16 bits with zeros. wav files of original and processed file will be of same size but flac of processed file be drastically reduced as last 16bit are all zero.

 

[ousi]

not correct. by removing the last 16bits (without dither, as dither surprisingly still retains lot more low level information even if you remove last 16 bits ) and replacing with zero retains "most of the music content" and you lose only low level information. consequently in low music passages you could listen noise where there was some information in the original file. for doing that open 24 bit file in daw, go to bit depth reduction, select 8 bit without dither. after that go again to bit depth and select 24 bit. this will pad the last 16 bits with zeros. wav files of original and processed file will be of same size but flac of processed file be drastically reduced as last 16bit are all zero.

I guess I know what you mean now by "zeros". The bit depth converter likely will just shift the bits 16 times to one side to drop the least significant bits and become 8-bit, and when the bit depth converter was asked to "restore" to 24-bit, it will move the 8-bit 16 times to the other side and filled the least significant bits with 0. This was what I meant by converting the 24-bit signed integer representation of the wave form into 8-bit. So we are talking about the same thing. The term "stripping 16-bit out of 24-bit" threw me off as I thought you are stripping the last 16-bit in a WAV sample which is usually encoded in little endian format, i.e. the last 16-bits are the significant bits.

FLAC is not ZIP so it doesn't actually look at those as "zeros". It's likely that because the waveform is now so uniform (only 8-bit resolution essentially), in the "Prediction" phase, it will fall onto the Constant model for longer time. Yea, you can say that it's "because of the zeros" but this is not exactly because those are 0-bits like how normal file compression look at the content.

WAVE file always store each sample with full bit-depth ignoring what's inside, hence it's the same size.

But I still don't get what you meant by "less information will make brain happier". How do you like telephone signal which are encoded at a much lower frequency and lower bit depth? That's definitely less information but I don't find that make my brain happier. It is the contrary which makes it much harder to recognize the sound precisely.

 

[aj625]

I guess I know what you mean now by "zeros". The bit depth converter likely will just shift the bits 16 times to one side to drop the least significant bits and become 8-bit, and when the bit depth converter was asked to "restore" to 24-bit, it will move the 8-bit 16 times to the other side and filled the least significant bits with 0. This was what I meant by converting the 24-bit signed integer representation of the wave form into 8-bit. So we are talking about the same thing.

FLAC is not ZIP so it doesn't actually look at those as "zeros". It's likely that because the waveform is now so uniform (only 8-bit resolution essentially), the "Prediction" models can describe it much more precisely and left with much less residual coding. WAVE file always store each sample with full bit-depth ignoring what's inside, hence it's the same size.

in a 24 bit word you cut last 16bits and fill again last 16 bits with all zeros. first 8 bit information remains intact. there is no change to wave form which you can see without zooming. i have already done this. flac encoder sees a repetitive pattern of zeros and thus reduces the size drastically.

 

[ousi]

in a 24 bit word you cut last 16bits and fill again last 16 bits with all zeros. first 8 bit information remain intact. there is no change to wave form which you can see without zooming. i have already done this. flac encoder sees a repetitive pattern of zeros and thus reduces the size drastically.

It depends on the byte ordering (also called endianness), the last 16 bits might be significant bytes (little endian, as in the default sample format of RIFF WAV files). Clearly you are not aware of how bits and bytes are stored in computer files or represented in memory.

Have you read how FLAC do the compression? https://xiph.org/flac/format.html#prediction It has nothing to do with pattern of zeros. FLAC is not a normal file compression algorithm unlike ZIP will tally up and count how many repeating patterns are there in a block.

Still you haven't really answer the counter-example I raised about your "less detail" theory. How do you explain the telephone case?

 

[welsh]

You also must be knowing that vinyl of modern digital recording is nothing but another form of listening native digital recording. But vinyl pressing of original analog recordings do sound more natural despite all the limitation of format. Reason is simple during the process of transferring from master tape to vinyl you did not chop the analog wave unlike digital where you save only some samples. Dac fill in those samples to recreate the wave back. Some dacs use more processing power to recreate that analog wave and use more and more samples for interpolating as close as possible as per Shannon theorem. One of my friend rips his old vinyls at 768khz and since with 768khz you get lot more points, the sound quality is very close to original vinyl. But if you rip to say 48khz the dac has to do lot more work to fill in those spaces and sound quality is not near to original vinyl. So imo people use vinyl mainly for original analog recordings. Vinyl of modern digital recording is nothing more than a fad. You would be lot better using a dac with that digital recording.

I have a decent turntable, because I inherited a huge vinyl collection from my father. In no way does it compare to my digital components. I listen mainly to what is loosely called ‘classical’ music, and hands down the worst medium (compact cassette excepted) for this is vinyl. Think back to the classic rock era and every album had a ballad next to the dead wax - because the information per second degrades as the stylus traverses the record. With a rock album, one can adjust the position of tracks, but ‘classical’ recordings often have no choice but to present climactic crescendoes at the very point at which the deeply flawed vinyl medium is most disadvantaged. I remember that my father switched to CD as soon as it became available. Scraping a rock around a plastic groove simply does not sound ‘more natural’ compared with digital.

 

[Veri]

Have you read how FLAC do the compression? https://xiph.org/flac/format.html#prediction It has nothing to do with pattern of zeros. FLAC is not a normal file compression algorithm unlike ZIP will tally up and count how many repeating patterns are there in a block.

It's just packing, when the .FLAC file is loaded it's just the same as the .WAV, both unpack to the same WAVE/PCM. It'd be absurd if a lossless codec would not, in fact.

 

[DSJR]

Back in the 80's, a valve output CD player (CAL Tempest - I think the donor chassis was a Philips from memory) was doing the rounds (those around at the time will remember it). I loved it, but got the feeling the higher noise on the output, although subliminal, may have been something. reducing stereo separation, especially at high frequencies also makes for a 'nicer' sound sometimes I found back then (and not just headphones either) and there was a 'Francinstein' (or similar name) add-on box which did just that. The Vinyl format does both of course and moreso as the hf is naturally compressed in addition to anything the mastering engineer added...

The other thing about very low level noise is how I remember getting a headache listening to the first Philips 14 bit machines, but recording said output onto casette (a then top Nakamichi three head with full calibration - a 682-ZX possibly), the noise wasn't directly audible but the headache disappeared...

I'm just suggesting that away from sota measurements, some 'tweaks' may just make for an easier-to-listen-to sound - possibly...

 

[welsh]

You seam very knowledgeable about this... i shuld better not waste more time arguing with you.


~17Bits dynamic is poor?
SO every CD(player) is also Poor?
If so every record player Must be Garbage.
Almost every Speaker is trash as well as almost every amp?
Why shuld we use different ratings for them?



But it has all the clarity you could ever want or need for the vast majority of music (CD 16bit 44,1k)
it is audible transparent for almost every listener/music. Its better then every CD player.
How is this Poor?


So why can't a person preferring R2R for style? They can't hear the difference, you can't hear the difference.

The difference is that you can get a better-performing DAC for a fraction of the price.

 

[welsh]

OMG that’s hilarious. €3k and it doesn’t even come in a case!
“the pcb is flat-press-mounted on a specially treated spruce-board. This construction is time-consuming during manufacture, but improves sound quality, as all components are mechanically coupled to a musical sound board.” I.e. he glued it to a piece of wood instead of paying for some metalwork.

I can absolutely believe this! After all, all my acoustic guitars have a spruce soundboard. Why would it be different for digital audio!

 

[welsh]

What Paul had to say last week should fit on the SMSL V3, too:

However, @Spocko an @jam among others are right when they remind us to strictly uphold the values and standards of this forum: People in demand for coloration should use filters, DSP, put a tube preamp in between, whatever they want.
For Amir and his audience here is simply no other choice than dismissing this product.
People might advocate for that item on forums like h**d-f*.org or elsewhere.

High-end DACs are expensive because of my pension requirements.