Sorry to butt in but IMHO, by the time you reach the "flow" stage, the thought process is no longer required, it has become an instantaneous thing. Before you even think, your body has started reacting to what the other side did. That's what separates the top dogs and the perennial also-rans.
DSD comparison to PCM.
- Thread starter U47
- Start date
Sorry to butt in but IMHO, by the time you reach the "flow" stage, the thought process is no longer required, it has become an instantaneous thing. Before you even think, your body has started reacting to what the other side did. That's what separates the top dogs and the perennial also-rans.
I wish I had one. There are legal definitions but, off the top of my head, premeditation might be equivalent to conscious cognition.
Yes, indeed you are correctI hope to also compare a NOS offering at the weekend if I can - a Metrum - still only 16/44 however & I know it has some shortcomings but it will give another datapoint.
Look forward to your observations!
Sounds like a reasonable definition. So if some cognition is subconscious the fact that a cognitive process takes place does not mean the response is premeditated, which is (if I understood correctly) what you originally asserted.
Kal Rubinson said:
If the congnition is subconscious, there is no premeditation, QED.
Last edited:
Aren't the Metrum DACs 'segmented current source' and capable of handling 24 bits?
The Audial no, but it looks very interesting. Not the CD-77 but the DP-777 very briefly and not in my environment. In any event, I will never comment on the sound of a component unless a) I'm sharing my thoughts on various rooms at a show, b) I've owned and lived with the component for a period of time.
Incidentally, the DAC chips in the PM2 look suspiciously like Ultra Analog D20400 chips. Opus, I know you have some strong thoughts about the issues of glitching in DACs - are you familiar with any techniques UA used in the D20400 to minimize this? I'm sure I read something to this end once a long time ago, but just can't find anything any more.
Mani.
The answer to both those questions is a 'no' they're resistor string and only 16bits. That's based on figuring out (within a margin of error of two or three possible chips) what the relevant devices are from specs and layout, because Cees likes to have the IDs scratched off his DAC chips.
Ah, for a moment there I was rather puzzled - I'd thought the PM2 an ADC. But I see on downloading the operating manual it has both ADC and DAC. You can't mistake D20400s as they're modules rather than chips, large hybrid things.
No, I'm not. I was involved in testing some DACs (a professional application) which used these modules and certainly their measurements were better than any chip available at the time. But none of the normal measurements show glitch, THD+N only does very indirectly. I understand they have a fairly standard 12 bit CMOS multiplying DAC inside - these are well known for being poor at glitching, so its entirely possible the UA module has a sample/hold inside. Its my understanding that Dan Lavry did most of the design work of the UA (though apparently he didn't see any of the profits) and he does use a S/H in at least one of his Lavry Engineering DACs.
Yep - my money's on that just being a badge-engineered D20400.
Incidentally, what's the ADC inside the PM2? It can't also be the partnering UA as that's S-D, a derivative of the Bob Adams dBx design.
Incidentally, what's the ADC inside the PM2? It can't also be the partnering UA as that's S-D, a derivative of the Bob Adams dBx design.
Last edited:
Here are the ADC modules:
They're about 1.5 times the size of the DAC modules. The PM2 designers claim they are true multi-bit ADCs.
This is what Michael Ritter said in an interview for Mix:
“The actual A/D converter in the Model One runs at 24 bits and 176.4 kHz currently; the Model Two will also convert at 192 kHz. We improve the linearity of our conversion with a high-amplitude broadband dither signal that we mix in with the program in the analog domain. The dither appears to be random, but the system knows at any given instant precisely what the amplitude of that dither signal is. And because we use our own custom, discrete, full-ladder converter with excellent amplitude and phase accuracy, we are able to apply an 'anti-dither' signal, exactly out-of-phase and matched in time, in the digital domain after conversion. That nulls the dither noise out of the signal… If it's going to be a 176.4 or 192kHz DVD-Audio release, then we will not decimate that signal; we use a proprietary filter [non-oversampled] optimized to that sample rate."
This is what Keith O Johnson said in reply to a question I asked him:
“Up conversion is not used at 176.4 kHz and 192 kHz sample rates. Instead, a passive analog filter with group delay correction removes alias products so that all 24 bits of information are captured from each sample by the ADC process. Conversion is made with ladder-pipeline architectures and analog added - DSP subtracted dithering, a first of its kind that is still used for instrumentation. Unlike sigma-delta methods, jitter sensitivity is low and accuracy does not degrade at high program levels.”
In any event, here are some test results of my particular PM2 unit (courtesy of Dave Peck):
ADC Multisine:
ADC THD -10dB:
DAC THD -10dB:
ADC/DAC Loopback:
With all this, can we be certain that the PM2 is indeed a true multi-bit ADC capable of of >16bit resolution? Interested in your thoughts...
Mani.
They're about 1.5 times the size of the DAC modules. The PM2 designers claim they are true multi-bit ADCs.
This is what Michael Ritter said in an interview for Mix:
“The actual A/D converter in the Model One runs at 24 bits and 176.4 kHz currently; the Model Two will also convert at 192 kHz. We improve the linearity of our conversion with a high-amplitude broadband dither signal that we mix in with the program in the analog domain. The dither appears to be random, but the system knows at any given instant precisely what the amplitude of that dither signal is. And because we use our own custom, discrete, full-ladder converter with excellent amplitude and phase accuracy, we are able to apply an 'anti-dither' signal, exactly out-of-phase and matched in time, in the digital domain after conversion. That nulls the dither noise out of the signal… If it's going to be a 176.4 or 192kHz DVD-Audio release, then we will not decimate that signal; we use a proprietary filter [non-oversampled] optimized to that sample rate."
This is what Keith O Johnson said in reply to a question I asked him:
“Up conversion is not used at 176.4 kHz and 192 kHz sample rates. Instead, a passive analog filter with group delay correction removes alias products so that all 24 bits of information are captured from each sample by the ADC process. Conversion is made with ladder-pipeline architectures and analog added - DSP subtracted dithering, a first of its kind that is still used for instrumentation. Unlike sigma-delta methods, jitter sensitivity is low and accuracy does not degrade at high program levels.”
In any event, here are some test results of my particular PM2 unit (courtesy of Dave Peck):
ADC Multisine:
ADC THD -10dB:
DAC THD -10dB:
ADC/DAC Loopback:
With all this, can we be certain that the PM2 is indeed a true multi-bit ADC capable of of >16bit resolution? Interested in your thoughts...
Mani.
Last edited:
Incidentally, for those of you wondering why the hell I'm talking about a 13 year-old ADC, it is totally related to this thread.
It's almost impossible to talk about PCM vs. DSD in a meaningful way. Why? Because all modern ADCs start with a sigma-delta modulator. Yep, all of 'em start out as DSD. It's funny listening to recording engineers who talk of not liking DSD. One in particular springs to mind who uses a Metric Halo ULN-8 for all his work. But this is a DSD ADC, with its output then converted to PCM!
The Pacific Microsonics Model Two may be the only true multi-bit ADC in existence, capable of 24-bit. And only when material has been recorded on such a machine and then played back on a true multi-bit NOS DAC can the merits of PCM really be appreciated.
To my ears, true PCM sounds closer to the original analog source than DSD. And unless you've heard 'true PCM' as described above, you're not really in a position to judge, right?
Mani.
It's almost impossible to talk about PCM vs. DSD in a meaningful way. Why? Because all modern ADCs start with a sigma-delta modulator. Yep, all of 'em start out as DSD. It's funny listening to recording engineers who talk of not liking DSD. One in particular springs to mind who uses a Metric Halo ULN-8 for all his work. But this is a DSD ADC, with its output then converted to PCM!
The Pacific Microsonics Model Two may be the only true multi-bit ADC in existence, capable of 24-bit. And only when material has been recorded on such a machine and then played back on a true multi-bit NOS DAC can the merits of PCM really be appreciated.
To my ears, true PCM sounds closer to the original analog source than DSD. And unless you've heard 'true PCM' as described above, you're not really in a position to judge, right?
Mani.
Why do you think that is? Curious about why technically speaking PCM might be superior to DSD? Of course, i am a non-techie, so hopefully you can describe 'slowly' to a non-techie.
This isn't my field at all, so like you I can be considered a non-techie. But even so, the Lipshitz/Vanderkooy paper cited by Opus is incredibly compelling. It really resonates with my own PCM vs. DSD findings.
Here's something that I've doctored from a paper by the DSD proponent, Andreas Koch:
The top is typical of 99.999% of digital recording/playback chains. When most poeple play back a modern recording on their modern DAC, they're following this.
Koch and others argue that the second chain is better - take out the initial DSD-PCM conversion in the decimation stage and the PCM-DSD in the interpolation stage and stick with DSD all the way through. Makes a lot of sense... until you read Lipshitz & Vanderkooy.
So that leaves the third chain (inserted by me, and not in Koch's original diagram). Start with 'true' PCM and stick with it all the way. The anti-alias filter will need to be a non-oversampling type, in which case the rate really needs to be 4fs (176.4 or 192). And this is exactly what the PM2 does.
I'm sure the 'real experts' here can chime in on the theoretical/technical pros and cons of the second and third chains. But I've spent an inordinate amount of time comparing them and have decided that the third chain is easily the closest to the original analog.
HTH.
Mani.
Here's something that I've doctored from a paper by the DSD proponent, Andreas Koch:
The top is typical of 99.999% of digital recording/playback chains. When most poeple play back a modern recording on their modern DAC, they're following this.
Koch and others argue that the second chain is better - take out the initial DSD-PCM conversion in the decimation stage and the PCM-DSD in the interpolation stage and stick with DSD all the way through. Makes a lot of sense... until you read Lipshitz & Vanderkooy.
So that leaves the third chain (inserted by me, and not in Koch's original diagram). Start with 'true' PCM and stick with it all the way. The anti-alias filter will need to be a non-oversampling type, in which case the rate really needs to be 4fs (176.4 or 192). And this is exactly what the PM2 does.
I'm sure the 'real experts' here can chime in on the theoretical/technical pros and cons of the second and third chains. But I've spent an inordinate amount of time comparing them and have decided that the third chain is easily the closest to the original analog.
HTH.
Mani.
Just a quick comment on the third chain:
You'll need a Pacific Microsonics Model One/Two as no other multi-bit non-oversampling 24/4fs ADCs exist to my knowledge. You'll also need a non-oversampling and non-filtering DAC capable of 24/4fs replay. There's only one that I'm aware of, and it's in my signature.
With these restrictions in mind, I have to say that the future of high-quality digital sound may very well lie with pure DSD. My experience of chain two was that it was considerably better than chain one. And I'm in two minds about this. One the one hand it's great knowing that there's a possibility that many people will be able to enjoy better digital sound than they've had before. You can do this today pretty cheaply. Just get a Mytek (or similar) DSD-capable DAC and listen to a few pure DSD downloads. You'll be rewarded with very, very good sound.
On the other hand, you'll not be listening to the best digital sound available. That lies with pure hi-res PCM... IMHO.
Mani.
You'll need a Pacific Microsonics Model One/Two as no other multi-bit non-oversampling 24/4fs ADCs exist to my knowledge. You'll also need a non-oversampling and non-filtering DAC capable of 24/4fs replay. There's only one that I'm aware of, and it's in my signature.
With these restrictions in mind, I have to say that the future of high-quality digital sound may very well lie with pure DSD. My experience of chain two was that it was considerably better than chain one. And I'm in two minds about this. One the one hand it's great knowing that there's a possibility that many people will be able to enjoy better digital sound than they've had before. You can do this today pretty cheaply. Just get a Mytek (or similar) DSD-capable DAC and listen to a few pure DSD downloads. You'll be rewarded with very, very good sound.
On the other hand, you'll not be listening to the best digital sound available. That lies with pure hi-res PCM... IMHO.
Mani.
I have read great things about Phasure. Congrat's on your choice of SOTA digital.
Yes - from a quick eyeballing of the FFT, I estimate the ADC's resolution is somewhere in the region 17-18 bits. Nowhere near the 24bits claimed, but then nobody has a true 24bit ADC in the audio bandwidth.
Worth noting that the 1st example in the diagram is focused on DSD going to PCM and seems to miss out the ones that do not go to 1-bit recording (sacd/dsd); the studios that are always 24bit/192khz or even 24bit/96khz and possibly 176.4khz.
Linn Records used to do SACD, possibly one of the 1st to do native SACD but eventually they decided to go PCM as all their editing was being done there anyway and now 24bit/192khz.
Quite a few engineers suggest it is very difficult to upsample/downsample from DSD to PCM and PCM to DSD perfectly.
I do tend to think that while mathematically it seems simple to go from various PCM sampling rates and DSD, the reality when analysing the music offered that has been upsampled or downsampled is that a fair amount of it has problems, and these problems can even be related to native PCM release change of sampling rates - freaking nightmare to find consistently excellent downloads.
This is seen when HiFi-News catches such releases with their measurement software reviewing hirez downloads.
Cheers
Orb
Linn Records used to do SACD, possibly one of the 1st to do native SACD but eventually they decided to go PCM as all their editing was being done there anyway and now 24bit/192khz.
Quite a few engineers suggest it is very difficult to upsample/downsample from DSD to PCM and PCM to DSD perfectly.
I do tend to think that while mathematically it seems simple to go from various PCM sampling rates and DSD, the reality when analysing the music offered that has been upsampled or downsampled is that a fair amount of it has problems, and these problems can even be related to native PCM release change of sampling rates - freaking nightmare to find consistently excellent downloads.
This is seen when HiFi-News catches such releases with their measurement software reviewing hirez downloads.
Cheers
Orb
I just went to the first three ADC chip manufacturers I could think of (Cirrus, ADI and TI) and pulled up the datasheets for their top of the range ADCs. Turns out all three companies are using multi-bit modulators, not single bit as is used in DSD. So what Andreas is saying might have been true back in the 1990s when DSD was first formulated, but its simply not the case today. A multi-bit (typically they're using 5 or 6 bits) modulator puts out noise-shaped PCM, not DSD.
<edit> Here's a 4th example, also multi-bit modulator : http://www.ardatech.com/1201.shtml. This part has a DSD output, it must be non-native DSD because the modulator isn't single bit, thus requiring conversion.
Similar threads
Members online
- wbass
- Lojay
- Alex G
- the sound of Tao
- MarkusBarkus
- Kingsrule
- andromedaaudio
- schlager
- Xampai
- McDee275
- PurePoor
- Artnet
- hopkins
- Christiaan Punter
- SlapEcho
- wbfhet
- MusicFellow
- kasin
- itay123
- simorag
- apachedna
- Charlie B.
- pjwd
- TLi
- Hiroshi
- bonzo75
- kabrio69
- ChasFreeland
- ashandger
- joeling
- RCanelas
- Sarge
- TWJeff
- Ampman
- rdg
- alex3
- Johan K
- wl215
- iaxel
- Rexp
- Juiced
 | Steve Williams Site Founder | Site Owner | Administrator |  | Ron Resnick Site Co-Owner | Administrator |  | Julian (The Fixer) Website Build | Marketing Managersing |