DIYHiFi.org

For the sake of audio
It is currently Sun Dec 17, 2017 2:21 am

All times are UTC




Post new topic Reply to topic  [ 291 posts ]  Go to page Previous  1 ... 15, 16, 17, 18, 19, 20  Next
Author Message
PostPosted: Tue Jun 24, 2014 11:13 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4333
Location: Somewhere other than here.
A guru might not mind detractors, but if it harms a bidnis, by casting them in a false light, then we have a problem.

If that detractor lived in the USA, the person who owns that bidnis would probably want to take target practice under your tutelage.

Would be a waste of time, on me. Could not qualify on the rifle range, during my brief military career. That is how they found out I am almost blind in one eye.

Oooops.......


Top
 Profile  
Reply with quote  
PostPosted: Wed Jun 25, 2014 3:42 am 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Kuei Yang Wang wrote:
he MASH converter can only be considered a "3.5-bit" converter."

ball3901 wrote:
I was under the impression that you can't/don't send 1 bit DSD directly into a delta sigma modulator.


You cannot. A 1Bit bitstream is already Delta-Sigma, there is nothing for the modulator to modulate. Now if you lowpass filter the 1-Bit bitstream to analogue, of course you follow it with a DS modulator (and at a different frequency and number of bits).



I said I was done. I can't help it. Is there a group such as 'Recovering Audiophiles/fools Anonymous"?


Well, the only way I can figure when such a thing happens, that is, a 1-bit DSD signal is sent directly to the modulator, is when it is filtered first to remove the high frequency content as to not overload the modulator.

Of course, in such filtering, the signal isn't 1-bit anymore. For examples, see Cirrus Logic and AKM onboard DSD 'processing'.


One bit or not, my original claim was that you don't send delta sigma into a delta sigma modulator. Then Mr. Signalyst pointed out how MASH does exactly that. I don't know. Actually don't even care anymore. :dizzy:

Oh, and still trying to figure out what this magical multi-bit format is that 'isn't PCM.' Three evil letters, PCM.

Andrew


Top
 Profile  
Reply with quote  
PostPosted: Fri Jul 04, 2014 2:17 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Another question...


noised shaped PCM vs multi-bit delta sigma..

seem like the same thing to me? I am trying to figure out the difference. I am sure it is in the nuance, but for instance, in the Sonoma DSD workstation, certain DSPs involve 32 bit, 2.8 mhz PCM, which is noise-shaped to 8 bit 2.8mhz PCM.

Now, it uses a 5th order noise shaper to do so, but no where does it say this noise shaper is delta sigma...

But I am having a hard time conceptualizing how 8 bit, noiseshaped and oversampled binary PCM is really any different from multi-level Delta Sigma?

If I feedback the quantization error in an oversampled PCM signal, so that the average value of the PCM pulses over time equals the higher resolution source, I fail to see how this is any different than mult-bit delta sigma?

Practically speaking, they seem to be the same thing to me. But it seems as if people still make distinctions.


Top
 Profile  
Reply with quote  
PostPosted: Fri Jul 04, 2014 11:09 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
ball3901 wrote:
Another question...


noised shaped PCM vs multi-bit delta sigma..

seem like the same thing to me? I am trying to figure out the difference. I am sure it is in the nuance, but for instance, in the Sonoma DSD workstation, certain DSPs involve 32 bit, 2.8 mhz PCM, which is noise-shaped to 8 bit 2.8mhz PCM.

Now, it uses a 5th order noise shaper to do so, but no where does it say this noise shaper is delta sigma...

But I am having a hard time conceptualizing how 8 bit, noiseshaped and oversampled binary PCM is really any different from multi-level Delta Sigma?

If I feedback the quantization error in an oversampled PCM signal, so that the average value of the PCM pulses over time equals the higher resolution source, I fail to see how this is any different than mult-bit delta sigma?

Practically speaking, they seem to be the same thing to me. But it seems as if people still make distinctions.



Okay, so I get the technical distinction. Oversampled and noise-shaped PCM uses absolute binary word encoding. Multi bit delta sigma uses parallel bitstreams that create a relative amplitude value.

But, both are noise-shaped, both are oversampled. Both have multi-bit words, whether that word is coded binary PCM or the sum of multiple bitstreams.

So, I honestly fail to see any real practical difference? Or am I missing something? I see them as two slightly different ways of accomplishing the exact same thing?


Top
 Profile  
Reply with quote  
PostPosted: Sat Jul 05, 2014 9:45 am 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Hi,

ball3901 wrote:
Okay, so I get the technical distinction. Oversampled and noise-shaped PCM uses absolute binary word encoding. Multi bit delta sigma uses parallel bitstreams that create a relative amplitude value.


I think that gets very blurry.

If we use (say) 8 Bit to encode inter-sample differences or absolute values, there is little practical difference. Encoding inter-sample differences has (theortically) higher resolution at lower levels of differences, but requires us to give away substantial headroom.

ball3901 wrote:
But, both are noise-shaped, both are oversampled. Both have multi-bit words, whether that word is coded binary PCM or the sum of multiple bitstreams. So, I honestly fail to see any real practical difference?


Makes two of us.

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Sat Jul 05, 2014 4:56 pm 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Hi,

ball3901 wrote:
Or am I missing something? I see them as two slightly different ways of accomplishing the exact same thing?


I just read the debate over at the "Caf" (if we call that first "other place" Pub, then I think "Caf" - in the English/London sense of a cheap working men's breakfast/lunch venue also known as "greasy spoon" is spot on) that sparked this.

It is good to see the workers of the department of Obfuscation and Prevarication on DSD do overtime.

Let us be clear.

PCM encodes the difference of a given sample from an absolute reference at a given sample rate. External low-pass filters must remove any energy above 1/2FS to avoid creating aliases within the usable range. There are no limits on either the number of bits or the sample rate, nor is oversampling (sampling at more than Fcutoff * 2 forbidden or necessary). There is no absolute requirement for using linear coding (which we find in standard Linear PCM -e.g. CD), especially not in this day and age, there is no reason to not (for example) use ADPCM, except tradition and installed base. Noiseshaping may be applied to quantisation noise/distortion, but there is no absolute requirement to do so, however without noise-shaping we must employ adequate word-length for a given required SNR, regardless of bandwidth.

Delta Sigma encodes the difference of a given sample from the previous sample. External low-pass filters must remove any energy above 1/2FS to avoid creating aliases within the usable range. There are no limits on either the number of bits or the sample rate and there is no absolute requirement for using linear coding. As Delta Sigma underlying system invariably have very high practical levels of quantisation noise/distortion high order feedback loops (ananlogue or digital) are applied around the AD converter, to increase the usable SNR and reduce distortion within the bandwidth of interest.

The key issues are:

1) PCM has an absolute zero representation and cannot accumulate offset all the way to DC output at full scale, Delta Sigma can and does and hence the 0dB level must be set at low modulation indexes, loosing effective SNR, unlike PCM.

2) PCM in it's pure form (as opposed to purely a coding system derived from any sort of source and applied to any sort of converter with any sort of intermediate re-coding) operates as non-looped, inherently highly linear pure feed - forward system. Delta Sigma in all practical applications operates as looped, inherently highly non-linear pure feed-back system with a very high order of loop frequency filtering.

3) PCM is not subject to any slew rate limit as such but requires filters to limit the bandwidth which reduces the slew rate it can represent. Delta Sigma has an absolute slew rate limit and lacks normally any filters that reduce the input signals slew rate. As such Delta Sigma is subject to the issues and limitations as any other looped feed-back system when it comes to slew rate and is fully subject to factors as TIM/SIM and related issues.

In addition while often only modest filtering is applied for recording (one should think a filter to be 120dB down at 1/2FS plus slew rate limiting would be the minimum), severe high order filtering close to the audio-band is essential at lower sample rates to avoid the rise of noise from noise shaping.

4) PCM is trivial to edit, mix and otherwise process essentially lossless in the digtal domain.

Editing, mixing and processing Delta Sigma in ways that are lossless is highly non-trivial and currently not possible past "cut & paste" splices (and may never be possible unless DSD is recorded to another, intermediate format, be that PCM or Analogue).

5) PCM and DSD cannot be converted to each other in a manner that is lossless.

As such we can see that each format has distinct limitations.

PCM's limitation requiring high order filters at relatively low frequencies can be overcome by "oversampling" (here used as in sampling at a rate higher than Fcutoff * 2), which relaxes filtering requirements. Increasing the DSD sample-rate reduces the needed filtering as it pushes the rise of noise higher and/or can allow the use of less severe modulator orders.

Multibit Delta Sigma does not actually exist as "format" so debating it is a bit like debating the colour of cheese the moon is made from. The precise formulation as a format internal to a given chip and/or software may be anything and it generally exists as only as an intermediate format.

Bottom line, DSD256 via DOP is the same as 352.8KHz PCM in terms of data used.

Both can be very good if done end-to-end, read PCM ADC, no or minimal, tasteful editing and PCM DAC. Heck, you should hear 2L's DXD stuff played back on a non-oversampling true PCM DAC, never mind HRx from Reference Recordings done on a PM Model 2. Read the same for DSD at high rates. The few decently managed DSD256 recordings we have are surprisingly good, even if the ADC actually invariably is MBDS at the core. By comparison DSD sounds way to "Hendrix" to my taste.

But no we are into subjective qualifications...

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Sun Jul 20, 2014 1:47 am 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Switching gears from DSD in my jack-of-all subjects, master of none thread..


Class C, B, A, A+, A++, A+++++, ad infinitum. I saw where a recently reviewed DAC garnered an A+ rating in a certain magazine. That wouldn't be a big deal, except this DAC only cost $299!!!! Seems odd. Seems very, very odd. Seems rather fishy, actually, that a 300 dollar DAC makes it into the best of the best category normally occupied by only the most esoteric. Well, not really most esoteric, as there are several options in class A+ for as little as 2 grand. But still, several order of magnitudes greater expense than this $299 V-DAC thingy.

So, according to these dudes, this $299 class A+ DAC is actually BETTER than class A stuff from AMR, Arcam, Classe, Electrocompaniet, Bryston (I could go on) and is even better than the SAME company's $3000 offering!! I guess not only will this new cheap DAC put all the big boys out of business. It will also cause this company to have a fire sale of all its more expensive DAC offerings.


REALLY? Oh please. Of course, it measures almost perfectly by the magazine's measurement standards. I am sure that has a LOT to do with the ratings, and the subjective preference as well. I mean, what if something sounded exactly the same but measured more poorly? Well, it sounds like class A+, but that x amount of second order harmonic distortion just doesn't measure up to the best, so let's make it Class B. Or at least, that is how the conversation must sound here in my head.


But seriously, other than being up to the task on this particular suite of 'pet' measurements (surely this piece of kit wasn't engineered solely to pass this magazine's particular tests), something isn't right here. I am sure that it is a fine sounding product. But I refuse to believe it makes everything else on the market a waste of money. For isn't that exactly what they are saying? If I could buy the very best A+ performing product in the world for a mere $300 bucks, everything else is a total waste of time and resources, and hi-fi as we know it would be dead. Very dead.

Hi Fi isn't dying anytime soon, though. Hi Fi magazines may be dying pretty soon, though. All I know is their credibility just took a huge hit with this person.


Top
 Profile  
Reply with quote  
PostPosted: Mon Jul 21, 2014 8:24 pm 
Offline
Benjamin
User avatar

Joined: Mon Dec 13, 2004 3:29 pm
Posts: 4333
Location: Somewhere other than here.
ball3901 wrote:
Class C, B, A, A+, A++, A+++++, ad infinitum.


Time for someone to find, and post, the Three Stooges "Singing the alphabet" ditty. Has to be out there, somewhere.

Then you will have your answer.


Top
 Profile  
Reply with quote  
PostPosted: Tue Jul 22, 2014 12:43 am 
Offline
Pig

Joined: Wed Jun 29, 2005 6:34 pm
Posts: 1068
Location: East Coast USA
It's just one guy's opinion! The review could be about pizza, music, or wine. :shock:

These reviews are not the magic answer to Life, The Universe, and Everything.


Top
 Profile  
Reply with quote  
PostPosted: Thu Jul 24, 2014 12:54 am 
Offline
Goat

Joined: Sat Mar 04, 2006 5:26 pm
Posts: 576
Location: Nova Scotia
ball3901 wrote:
So, according to these dudes, this $299 class A+ DAC is actually BETTER than class A stuff from AMR, Arcam, Classe, Electrocompaniet, Bryston (I could go on) and is even better than the SAME company's $3000 offering!! I guess not only will this new cheap DAC put all the big boys out of business. It will also cause this company to have a fire sale of all its more expensive DAC offerings.

REALLY? Oh please.

If you're building reasonable quantities, you can build a really kickass DAC for a sale price of $300.

Say your manufacturing price is $150, to make room for your own profit and store profit. $150 will buy you a multilayer PCB, low noise power supplies, a low phase noise clock, all the digital bits you need, a high end DAC chip, decent I/V, and a humble looking case. Pretty much everything you need to make a DAC that measures excellent on Audio Precision gear, at least close to the limits of the DAC chip that you're using.

Now what can you put in a $3000 DAC that you can't put in a $300 DAC? I guess custom SC cut quartz, individually tested/binned components, etc... could all be done, but who's gonna do that? Chances are, the $3000 just features a whole lot more bullshit and voodoo.


Top
 Profile  
Reply with quote  
PostPosted: Thu Jul 24, 2014 5:22 pm 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Hi,

gmarsh wrote:
If you're building reasonable quantities, you can build a really kickass DAC for a sale price of $300.


What is "reasonable QTY"? Millions (like iPhones)?

Anything with less than 100ku annually is "Boutique", sold in "Boutiques"...

gmarsh wrote:
Say your manufacturing price is $150, to make room for your own profit and store profit.


Many stores will not touch anything below 50 Points unless it is a "must carry" product like the phones made by Apple), some expect even more.

Often it necessary to have a local distributor for market access etc., they also like to make some money.

It is nice of you to grant the manufacturer some profit, considering that we have to design, assemble and test these units (or farm that work out) AND to handle most of the promotion etc.

gmarsh wrote:
$150 will buy you a multilayer PCB, low noise power supplies, a low phase noise clock, all the digital bits you need, a high end DAC chip, decent I/V, and a humble looking case.


And the typical retail sales price for something like that using (say) ESS 9018 will be above 1,500 USD from most manufacturers.

gmarsh wrote:
Now what can you put in a $3000 DAC that you can't put in a $300 DAC?


300 USD worth of parts and enclosure, the way boutique audio high end usually works?

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Thu Jul 24, 2014 7:11 pm 
Offline
Pig

Joined: Wed Jun 29, 2005 6:34 pm
Posts: 1068
Location: East Coast USA
Retail sales outlets, and whatever distribution is needed to support them, costs a lot. It's a really high percentage of the selling price, especially for specialized products. This is true whether it's audio gear, higher quality clothing (ever wonder why clothing stores have no problem with 50-70% off at "season's end"?), or dozens of other products. But, the people involved are all part of the economy, too. They all contribute to our own incomes by directly or indirectly purchasing the products and services we offer in our day jobs.

Complex systems are, well, complex.

I wonder if the focus on cheapness affects the brain's capacity to enjoy music, especially electronically reproduced music. Kind of a parallel idea to why DBT's might not be really helpful - the testing forces the test gear, I mean listeners, to multitask and divert brain processing away from listening.


Top
 Profile  
Reply with quote  
PostPosted: Mon Aug 25, 2014 4:01 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Kuei Yang Wang wrote:

Mr. Michael who is the designer of Mytek

"Ok,the way DSD is implemented in our DAC: DSD 1 bit> digital DSD 3 pole filter @ selectable 50,60 or 70kHz>6 bit output > 6 bit DAC. The filter, standard in all sacd players (typically analog) helps to filter out of band noise inherent n DSD. Such filtering requires multiplication which produces result in more bits than the 1bit input......"



Yeah, I am dredging up an old topic.. but I have nothing better to do this morning, and the topic has been on my mind.

The above I believe is actually correct. I am going to assume 1 bit DSD goes to the digital filter, where just like Mytek says, it is multiplied into a multi-bit signal. (of course, this process is lossy, and results in redundant samples ahem.. decimation) Just like everyone else who processes DSD. Cirrus, AKM, etc.. even Sonoma/Philips DSD-wide is treated the exact same way. At this point, the volume control is applied. I am now going to assume it works like DSD-wide. Volume control is applied at 32 bits, then is noise-shaped back to 6 bits. Either that, or the output of the filter is actually 32 bit, not 6.

The filter rolloff in the data sheet would seem to indicate the above scenario is very possible. Of course, then there is the ASRC. Can you send noise-shaped Data into an ASRC? I don't think you can. Correct me if am wrong.


The above is 'best case' scenario for DSD on the ESS chipset. Of course, there are several other possibilities. But they ain't talkin.


Top
 Profile  
Reply with quote  
PostPosted: Mon Aug 25, 2014 6:35 pm 
Offline
Muriel
User avatar

Joined: Thu Jul 26, 2007 5:45 pm
Posts: 1167
Hi,

ball3901 wrote:
The above I believe is actually correct.


Depends. In some jurisdictions advertising "Spam" made from pink slime (aka mechanically recovered meat) as actually "Meat" is perfectly correct. I will still not eat it.

ball3901 wrote:
Of course, then there is the ASRC. Can you send noise-shaped Data into an ASRC?


Of course you can send multibit PCM data to an ASRC. Noiseshaped or not does not come into it. The ASRC sits (if I understand right) before the volume control and seems wrapped into the digital filter.

ball3901 wrote:
I am going to assume 1 bit DSD goes to the digital filter, where just like Mytek says, it is multiplied into a multi-bit signal. (of course, this process is lossy, and results in redundant samples ahem.. decimation) Just like everyone else who processes DSD. Cirrus, AKM, etc.. even Sonoma/Philips DSD-wide is treated the exact same way. At this point, the volume control is applied. I am now going to assume it works like DSD-wide. Volume control is applied at 32 bits, then is noise-shaped back to 6 bits.


Wow, what a way to deliver pure unadulterated DSD.

Sounds like whoever calls that pure pre will marry a 40+ Year old streetwalker who has been at it since age 12 and has any STD imaginable and a few unimaginable ones to boot and call her "the purest and lovgirl I have ever known" (who knows, it may be true! - he may have only ever known toothless slapper's in their late 60's dying of HIV related complications).

So everything is relative.

ball3901 wrote:
The above is 'best case' scenario for DSD on the ESS chipset.


If that described the 'best case' of how my dinner was prepared, I'd be very sick right now.

Lucky my dinner was not like that...

Ciao T

_________________
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." Richard Feynman


Top
 Profile  
Reply with quote  
PostPosted: Mon Aug 25, 2014 7:10 pm 
Offline
Cow

Joined: Tue Apr 01, 2014 1:47 am
Posts: 131
Location: Knoxville, Tn
Kuei Yang Wang wrote:
Hi,



ball3901 wrote:
Of course, then there is the ASRC. Can you send noise-shaped Data into an ASRC?


Of course you can send multibit PCM data to an ASRC. Noiseshaped or not does not come into it. The ASRC sits (if I understand right) before the volume control and seems wrapped into the digital filter.




Ah, well, then. I think that answers it for me. Now, I still may be wrong, but I am satisfied enough for the moment. DSD is digitally filtered into a multi-bit format via the 3-pole filter. How many bits? Volume control is advertised at 32. But, that could simply mean 32 bit processing, then noise-shaped back to 6 bits.

But, seems like every other DAC that digitally processes DSD, and also seems like DSD-wide on the pro side.


You are right, and have always been right. This isn't DSD anymore. To me, it is just high sample rate PCM, with noise-shaping. I don't use the 'PCM' acronym anymore on the forums. Drives people crazy. I just say 'multi-bit intermediary'. Make of it what you will.


Best case? Maybe. Seems to be the best chance we have with the digital filtering. Higher sample rate, gentler slopes = better impulse response and less ringing. In theory, anyway.


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 291 posts ]  Go to page Previous  1 ... 15, 16, 17, 18, 19, 20  Next

All times are UTC


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group