To DSD or not to DSD?

[sandyk]

Sounds very much like a phase/timing domain problem to me, which ties in to what I was mentioning about a couple of possibilities - that there are phase problems with DSD I am able to listen past and you are more sensitive to;

 

Hi Jud

I think that Peter is likely to be quite sensitive to phase differences, as I am, due to hearing in both ears being a bit different.

 

Regards

Alex

[Hiro]

The downside of the commonly used filters for Oversampling DACs is that they ring as hell when the filtering job is done as described.

 

What about oversampling SDM based ADCs?

 

Are both PCM ADCs and DACs must be non-oversampling units to get the benefit of ringing free playback?

[PeterSt]

What about oversampling SDM based ADCs?

 

Are both PCM ADCs and DACs must be non-oversampling units to get the benefit of ringing free playback?

 

This is a most difficult subject and I don't think I can answer it well. And, I am sure Miska will do a better job on this. What I (perhaps falsely) think ?

 

With the notice that the ringing as such emerges during playback at least and if all is right for Redbook only (to some degree, depending on the sampling rate of "hires" it is compared with) ... I think the "decimation" in order to take out the beyond fs/2 frequencies (like Redbook is not allowed to contain beyond 22049 Hz) doesn't ring. It may create anomalies depending on the decimation filter, but whether that implies ringing ? I don't think so.

[so here Miska will jump in :-)

 

The different subject is sheer PCM recording vs. SDM recording. So, genuine PCM recording sure existed - in 14 bits and through (video) tape machines. Next there's Pacific Microsonics of which it is believed that this recorded through PCM and without filtering (for the higher res not-yet-decimated stuff). Also, all what is recorded through that device sounds the best through my ears (and not only mine) through the PCM based NOS1. Coincidence ? So to give an example, Reference Recordings sound the best. Mind you, Redbook. Hires recorded through the PMI/PMII sound good just the same, but relatively less is around of that. Redbook ? spades (start with all HDCD).

 

Anyway, like DSD-DSD there's merit in PCM-PCM, although it is harder to see whether the recording has been real PCM because it's all much of a secret how the PM(II) worked on the ADC side. But (FWIW !) anticipated my next post for the PMII it works out the same or similarly as my next post will end with : while the ADC of the PMII is regarded the best, the DAC side of it is not at all (and not because I say so - customers do ... although I heard it myself).

 

ALL IS TOO D*MN DIFFICULT TO COMPARE

[PeterSt]

SDM based DACs ring and once you heard the difference with non-ringing you'd know what this is all about.

Although the principle of DSD is "SDM based" (for the recording part), funnily enough this does NOT ring (or hardly).

 

Peter, haven't you recently admitted that when the SDM signal is kept SDM end-to-end it helps with the ringing problem?

 

Hiro, I realize that this is all too difficult to be kept in its proper context, but in the case I talked about, this was foremost about uneven piano notes. You know, that one or a couple may jump out for level (especially the higher key ones). This wasn't about ringing. Also, I found the sound odd all over. You know about cymbals with nails in them at the sides (these formally exist) ? Here the cymbals (all of them !) sounded like stuffed with nails. Well, it could be called ringing of course and actually I don't know how it emerges, but at least they sounded real long (decaying). But correct ? no way. Same for the bass. Very poor.

And in between the lines : More than one CA member was there too and I'm not making up things for any reason. They should recognizing what I am saying here, but with the notice that I could find no-one dislikig the sound (and I asked many explicitly). Still it should be so that "we" all recognize that not ALL cymbals have nails in them. Like not all cymbals are Chine (see earlier post).

 

The biggest fun for me was that the host there was founder and executive for the best recording label I recognize : (late) Turtle Records. Now, the very same engineer I talked to did not recognize one single thing I said about the "negatives", or he must have went into sheer defense mode. Still we listened to his own mixing and mangling of their own recordings, and all he could say was something like "basses can never be captured well". This, while I own their recordings and they are the best of all (taken by a first 96KHz digital recording device which to me looks like a PCM thing). My conclusion of that was that in my listening room their own recordings sound better than through their own gear while they performed those recordings themselves. So, odd.

A last thing on this (might it be interesting) : XXHighEnd contains a sort of self protection for byte order anomalies named "Crack Detection". This contains a threshold about not too many steep transients (like the 1V on one go I talked about earlier), or otherwise I have "programmed" something wrong. So when it trips something should be wrong. But that threshold should be able to let through realistically occurring transients. And you know what ? I later had to adjust the threshold upwards for this same Turtle records. They just are the most dynamic existing and sound superb for it (this is all Redbook btw).

 

So what am I saying ? that all again says nothing much, except for maybe me being an ignorant prick to at least those guys who -thus says me- don't even know themselves how good they are. Next they too liked the DSD stuff for the better but I guess they forgot to compare to their own old digital recordings (90's stuff).

 

I *have* to be the one who is wrong, which is why I keep on seeking. But what to do if I can't do it myself (till now).

 

Regards,

Peter

[Miska]

Of course the especially the decimation filter (from SDM to PCM, or from higher rate PCM to lower rate PCM) is the one that rings. Plus it cannot be leaky either because otherwise it would let frequencies fold straight into the audible band too.

 

You can compare large bunch of software converters here:

SRC Comparisons

in the "Impulse" test result you can see how the filter rings and in "Sweep" you can check how much it leaks back to < 20 kHz after the sweep bypasses 22.05 kHz at around 5.2 seconds mark. Most software here are better than the filters in ADC chips that typically have max -100 dBFS to -120 dBFS stop band attenuation.

 

With suitable design, it is possible to replace the ringing with other kind of ringing in upsampling filter before DAC. That is where I've put my emphasis on (because I cannot change the source end).

 

One reason I like to use DSD DACs is that it allows me to turn modern SDM DAC into "NOS" mode and replace both the internal digital interpolation filter plus the modulator with my own running in software.

[Hiro]

Hiro, I realize that this is all too difficult to be kept in its proper context, but in the case I talked about, this was foremost about uneven piano notes. You know, that one or a couple may jump out for level (especially the higher key ones).

 

Peter, if you ever get a chance, do give this piano recording a listen:

CIABeethoven: Cello Sonatas/Wispelwey SACD - Classics TodayClassics Today

 

And feel free to compare it to any digital recording from the 90s

[Jud]

I *have* to be the one who is wrong, which is why I keep on seeking.

 

Regards,

Peter

 

Well not necessarily, which is what makes it all so interesting. Debates about "X must be better always!" are not interesting to me except to wonder how people can really believe it. Much more exciting to try to think of why it might be that Barry and Peter don't like what they hear, while Miska and Cookie M. do. (And so do I, not that it should mean anything to anyone else.)

[SoNic67]

They should recognizing what I am saying here, but with the notice that I could find no-one dislikig the sound (and I asked many explicitly). Still it should be so that "we" all recognize that not ALL cymbals have nails in them. Like not all cymbals are Chine (see earlier post).

 

This exactly what I experienced too. I have my test CD: Police-Reggatta De Blanc (SHM). The only what cymbals sound 'right' is when I play the CD trough a multibit DAC (PCM61 or TDA1451). Trough PCM 1792 sounds a little weird (decays are jumbled up around quitars), and trough CS4398 sounds even worse. I have the SACD disc Every Breath You Take: The Classics that has several songs from Reggatta ('Message In A Bottle' for example) and, even if SACD 'should' sound better, it suffers from the same cymbal issues. And, besides cymbals, it has other crappy parts on surround, but those are mixing issues I think. Those are real discs that I am using, not bootleg copies.

[elcorso]

This exactly what I experienced too. I have my test CD: Police-Reggatta De Blanc (SHM). The only what cymbals sound 'right' is when I play the CD trough a multibit DAC (PCM61 or TDA1451). Trough PCM 1792 sounds a little weird (decays are jumbled up around quitars), and trough CS4398 sounds even worse. I have the SACD disc Every Breath You Take: The Classics that has several songs from Reggatta ('Message In A Bottle' for example) and, even if SACD 'should' sound better, it suffers from the same cymbal issues. And, besides cymbals, it has other crappy parts on surround, but those are mixing issues I think. Those are real discs that I am using, not bootleg copies.

 

You need to know the source. This never happens to me in DSD recordings, but sometimes from PCM to DSD transfers, and less if the source is analog, but from old tapes with some damage.

 

Roch

[Miska]

and trough CS4398 sounds even worse

 

CS4398 sounds pretty good when fed with external digital filter at static 192/24 input rate. Even better when switched on Direct DSD mode. At least in my own DAC. As usual it of course also heavily depends on what kind of analog stage there's after the DAC chip. And how the DAC chip is programmed and configured.

 

and, even if SACD 'should' sound better, it suffers from the same cymbal issues. And, besides cymbals, it has other crappy parts on surround, but those are mixing issues I think.

 

What is the source and what modulator was used to produce the DSD stream? Saying that DSD sounds like something is like saying that 1/4" tape sounds like something. Of course it depends on the tape brand, the tape recorder, etc...

[PeterSt]

This exactly what I experienced too.

 

Interesting (to say the least). May I ask, through what DAC is this ?

[sjoc2000]

 

While the latter is an almost possible phenomenon to share (for a "go listen and you will see"), now take an upright bass. Go to a live concert (or invite Ray Brown to your living room) and notice that each single pluck (when not too softly played) is overwhelmed by the metal sound of the plucking itself (the zzzzing in rough on/off frequency). This is very special because it really needs "something" to bring that forward through loudspeakers. So this is what you can try (take a random recording with a profound double upright bass) and watch for the metal sound. It won't be there, or maybe you may say "hey !" for a couple of times only. It should always be there because in real life it is always there.

Peter

 

Peter;

 

Great recommendation for a little experiment some can try.

 

I just happen to have a copy of, "2002 - Ray Brown, Monty Alexander, Russell Malone (SACD-R 24 88.2) FLAC". Took ( almost did a toke again, :0) ) awhile for me to get my attention on the very beginning of the plucking of a string, and everything was there to my ears, the kind of click of the finger snapping the string, and the residuals of the mechanical action, in concert with the the appropriate vibration of the string. This judgement would be based on what I have heard when listening to the the bass violin live.

 

I used Audiogate, and converted this file to DSD128. Nothing was lost, and these actual mechanical sounds were actually more pronounced to these ears. If a musical piece is like a body of shimmering frequencies, it seems to me that DSD expresses every single aspect of this moving piece of the infinitude more faithfully than anything I have ever heard, other than a live performance, and very very close to that.

 

A DSD recording would I'm sure even exceed what I heard with my little experiment in conversion.

 

It is possible that I have never heard precisely what you are referring to. I am a long ways from having the experience you've had with critical listening, but I can hear pretty well, and I can't hear anything missing.

 

Love your posts, but in this case disagree with your conclusion. :0)

 

Jim

[SoNic67]

Interesting (to say the least). May I ask, through what DAC is this ?

The best results I think come from SM5845AF+PCM61J+2bits_analog (total 20 bit for 8xOS), using as I/V LM4562, not the internal OpAmp. Adjusted the MSB pot for minimum distortion (3rd).

Like I said I have also SAA7220+TDA1541A (16 bit for 4xOS) and YM3417D+PCM61 (total 18 bit for 8xOS) and although they sound better than PCM1794 and CS4398, I think I like better the first combo. Maybe because of separated I/V OpAmp and 20bit following the OS...

I am tempted to modify a player with CXD2500BQ from Sony as CD driver and error recovery (and same SM5845+PCM61), just because it has a double buffer - 32kB of RAM instead of usual 16kB. That might help with the mechanical, low frequency jitter.

 

Also, I am tempted to get something based on the PCM100 filter, but they are very few remaining on eBay.

[sandyk]

SoNic67

 

If the LM4562 isn't a surface mount device,have you tried replacing it with the more expensive metal can LM4562HA/LME49720HA ?

Despite having the same chip internally, they get quite warm, which indicates improved thermal management, and many report they sound even better too after 24 hours of use. One of the original NS design team reported they sound better too, in DIYAudio

Regards

Alex

[SoNic67]

If the LM4562 isn't a surface mount device,have you tried replacing it with the more expensive metal can LM4562HA/LME49720HA ?

DIP8 capsules in those old players. Cheaper than metallic can and better dissipation in free air:

SO (SMD)=145°C/W

NA (DIP8)=102°C/W

HA (Metal Can)=150°C/W

HA can be better cooled with an external heat-sink (theoretically down to 35°C/W)... But if you don't use a heat sink, is worse.

I am waiting fro some browndog adapters to try single OpAmps.

[sandyk]

SoNic67

In another forum we are using them with T05 heat sinks.

Regards

Alex

[sandyk]

See also

 

National opamp inflation - diyAudio

[Miska]

Like I said I have also SAA7220+TDA1541A (16 bit for 4xOS)

 

I had Marantz CD-60 with this combo. It just sounds so much like 80's digital with all the nasty digital glare etched on top of everything. It definitely sounds digital. But on those old days (SAA7220 is from 1985) they couldn't put good enough maths into such low cost chip, especially so that it would work without huge heatsink. It's just cascaded 2x120-tap FIRs performed in 16-bits. Quite a bit of ripple and distortion.

 

I don't know about SM5845, but SM5846 is pretty ordinary 8x DF with 3x cascade FIR.

 

I still have two of my old designs, DF1700 + PCM63P combo with very good analog section and DF1700 + PCM1700 combo also with very good analog section.

 

For the PCM1704 I would recommend just pairing it with DF1704 or DF1706 as it was designed to be. Or even better, run these with digital filter in a computer to avoid the limitations of the hardware ones.

[One and a half]

Well, this is a professional audio manufacturer's opinion.

 

"In my opinion DSD is completely unnecessary." - Charles Hansen March 25, 2013

 

So based on that opinion, Ayre are now making a DSD upgrade for their DAC, so what changed and what's the care factor?

[Miska]

I still have two of my old designs, DF1700 + PCM63P combo with very good analog section and DF1700 + PCM1700 combo also with very good analog section.

 

Of course I still also have a DAC I designed around CS4328 chip which is practically a pure "DSD DAC" (although it has only PCM inputs), since it operates at 1-bit 64x fs.

[SoNic67]

I agree that SAA7220 sounds bad, but I kept the player for the NOS experiments with TDA1541. Still better that the CS4391 (4th order D-S) that I have in a Harman Kardon receiver. CL engineers knew/heard something when they abandoned the 1 bit D-S cips and evolved to multibit D-S (like anybody else in industry)... Or you think you know more than them?

As for SM5845 - it's the original Alpha processing in Denon. Outputs 20 bit from 16 after OS. IMO that's the only way to preserve original dynamic, not adding noise shaping instead.

I did hear the DF1700 and I was disappointed. I didn't like it as I was expecting.

 

PS: Maybe a mode can split those last posts (#237 down)?

[Jud]

Well, this is a professional audio manufacturer's opinion.

 

"In my opinion DSD is completely unnecessary." - Charles Hansen March 25, 2013

 

So based on that opinion, Ayre are now making a DSD upgrade for their DAC, so what changed and what's the care factor?

 

Charles feels his ADC has no need of DSD to sound its best. OTOH, there are SACDs and DSD files out there that people may want to play, so why not give his DAC customers the option (at a price that may be a self-fulfilling prophecy regarding how many customers will want to take him up on the offer)?

[mav52]

Charles feels his ADC has no need of DSD to sound its best. OTOH, there are SACDs and DSD files out there that people may want to play, so why not give his DAC customers the option (at a price that may be a self-fulfilling prophecy regarding how many customers will want to take him up on the offer)?

 

They way I view it, You build it, someone will buy it.. So for Charles it's, Get on the bandwagon or "miss out on sales". It's make good business sense to bring in revenue when you have the opportunity than let your competitors get it..

[Jud]

mav, ordinarily I might think the same, but (1) I'm not sure Charles is the bandwagon type, and (2) that is definitely not a bandwagon price.

[wisnon]

Yeah well, anyone who does not get on board will suffer in the next few years, as DSD has become a business fundamental. The level of DSD is what is now nice to have, but DSD64 is now a must have in order to hold market share.

 

Yes, there are many not interested, but even some of those will consider DSD as a purchase criterion because of future-proofing or resale value.

 

That is the reality of the marketplace.