Marantz SA-KI Ruby player and dac latest killer Delta Sigma??

[georgehifi]

Is this the latest delta sigma based wow machine player and dac, with what seems to be Marantz's own 1 bit D/A converter "Marantz Musical Mastering (MMM)"

https://www.us.marantz.com/us/Products/Pages/ProductDetails.aspx?CatId=referenceseries&SubCatId=&ProductId=SAKIRuby

 

Cheers George

 

Edited October 6, 2018 by georgehifi

[davewantsmoore]

50 minutes ago, georgehifi said:

Marantz's own 1 bit D/A converter "Marantz Musical Mastering (MMM)"

Rather than being a 'special 1bit converter' .... what they really mean is the 'secret sauce' is in the DSP code used for oversampling / remodulating and dithering of the audio.    Similar to the likes of PS audio, Meitner, or what people are doing with computers, eg. HQplayer, or xxhighend.... or in related way, MQA. 

[sir sanders zingmore]

It’s delta-sigma so it must be rubbish anyway, right?

[zenelectro]

5 hours ago, georgehifi said:

Is this the latest delta sigma based wow machine player and dac, with what seems to be Marantz's own 1 bit D/A converter "Marantz Musical Mastering (MMM)"

https://www.us.marantz.com/us/Products/Pages/ProductDetails.aspx?CatId=referenceseries&SubCatId=&ProductId=SAKIRuby

 

Cheers George

 

George,

 

It appears they are converting everything to 1 bit DSD after which you can directly convert to analog. Looks like a similar concept to the Mola Mola DAC

where it is basically a bitstream fed to a multi-element, equally weighted, DAC architecture. I think Chord Dave (pulse array dac) might be same.

So if you had say 16 elements (resistors), the DSD signal gets shifted along those 16 resistors, one at a time, with every clock cycle.   

 

The end result, at resistor summing point (OP) has some good properties:

 

- The HF noise is attenuated significantly as this acts like a low pass filter.

- Even though it is a low pass filter, it is transient perfect WRT the fact that it has zero ringing.

- The resistors don't need to be matched. Any mismatch does not affect linearity at all.

 

These DAC's can be made VERY linear so long as you take into account finite rise times, which can be addressed with RTZ type coding.

 

Having said all the above, expect more development of R2R DAC's. They haven't yet fully exploited this technology. The Holo Springs appears

to me to be one of the best implementations so far. I do also like the Kassandra DAC using AD1865's in a very smart configuration.

 

T

 

[georgehifi]

1 hour ago, zenelectro said:

George,

 

It appears they are converting everything to 1 bit DSD after which you can directly convert to analog.

It's odd they should say this, in their specs. under "specifications" then "specification +" instead of just naming a ESS or Wolfson ect ect seem to me they want you to believe they make their own dac chip.

 

 D/A Conversion = Marantz Musical Mastering (MMM)

 

Cheers George

 

Edited October 7, 2018 by georgehifi

[georgehifi]

These look to be the 1 bit D/A chips in yellow, and what could be discrete low bit R2R chips?? in red.

 

Cheers George 

[zenelectro]

They don't have to, nor would they likely anyway, make their own DAC chip.

It's a simple resistor array driven by high speed logic.

Or maybe they are using Sabre etc with everything disabled / bypassed.

These day there are so many ways to do conversion well - we are lucky.

 

George, on a separate subject, do you have any info on Burr Brown colinear (sign magnitude)

DAC arcitecture? I've seen the usual stuff on 1704 but am after more detail on how the

CCS's are arranged. I know there is an AES paper, might have to buy it. 

 

cheers

 

Terry

 

 

[zenelectro]

I checked out some other pics, it is as described earlier,  a type of pulse array DAC.

Edited October 7, 2018 by zenelectro

[georgehifi]

2 hours ago, zenelectro said:

I've seen the usual stuff on 1704

No but I do have this on the 1704 you probably haven't seen.

 

 

 

 

Japan PCM1704 pdf.pdf

 

Cheers George

Edited October 7, 2018 by georgehifi

[zenelectro]

Thanks George,

 

Yep, got that one too. It's far too simplistic.

I'll get the AES paper on Colinear Architecture by Burr Brown at some stage.

 

T

 

 

[davewantsmoore]

On 07/10/2018 at 1:54 PM, georgehifi said:

want you to believe they make their own dac chip

Like I said it's not, "their own DAC chip" .... as much as their own oversampler/modulator code to create the PDM signal (that just needs a low pass filter).

 

... but they need to state that in a way that people will not misunderstand .... so "we made a special dac" it is.

[georgehifi]

1 hour ago, davewantsmoore said:

so "we made a special dac" it is.

 

"D/A Conversion = Marantz Musical Mastering (MMM)"

 

Is what Marantz say in the specs+ section in their website on line.

And they or no one else mentioned any thing else to the contrary, except you Dave, if your so sure what is it? Manufacturer and model no.? 

 

Cheers George

Edited October 8, 2018 by georgehifi

[davewantsmoore]

1 minute ago, georgehifi said:

if your so sure what is it? Manufacturer and model no.? 

There is no "DA converter" so to speak (ie. no DA chip) ...  Just a DSP and an amplifier (excuse the oversimplification)

 

You convert everything to PDM (DSD like), and then low pass filter it.     That's easy to say, but it's all about the choice of details when you do the conversion - as to what performance you get out the other side.

 

In the streophile article, they quote Marantz as saying  :"it's the first player/USB converter without a DAC—it doesn't need one!"..... but of course, they mean "no DA chip".

 

 

The tell that it's a PDM amplifier is right here.

Quote

By up-converting all file formats to DSD, using advanced filtering and processing and direct 1-bit conversion to feed the analog outputs

• Innovative Marantz Musical Mastering “MMM-Stream” up-sampling to DSD and filtering
• Custom-designed 1-bit direct “MMM–Conversion” from DSD to analog

up-converted to DSD at 11.2MHz using the proprietary MMM-Stream converter within the player, and the high-frequency signal produced is processed by the unique MMM-Conversion stage, used in place of a conventional DAC, to produce the analog output. The MMM-Stream section of the process replaces the oversampling filters normally used in digital to analog conversion and allows the implementation of the Marantz Musical Mastering filtering.

 

The pic you posted earlier is of the conversion parts (summing and low pass filtering)  .... and the output amplifiers are above it.   You can see an Altera (Intel) Max-V computer (the larger chip underneath your yellow circle) which is driving the filters, and it's interesting that they are using so many lanes.

 

By design, the analogue side of it is reasonably boring (although Marantz look to have done something relatively complicated) ..... where all the performance is wrung out of the thing is where the audio is converted in to PDM (ie. ~DSD).    This is in the digital side which is under the metal lid.   There's plenty of pics of that around - and you will see another Max-V chip, and two large Analog Devices SHARC DSPs.... and this is where the code is run that is much of the 'secret sauce'.   Pics of the SA-10 are also pretty close.

 

1 hour ago, georgehifi said:

"D/A Conversion = Marantz Musical Mastering (MMM)"

Is what Marantz say in the specs+ section in their website on line.

Indeed.   What does it mean?   It's marketing whimsy.

 

1 hour ago, georgehifi said:

And they or one one else mentioned any thing else to the contrary, except you Dave, if your so sure what is it?

Sorry, I'm not quite sure what you mean.    Anything to the contrary, of what?!

[georgehifi]

Here you can go to get some real close ups of the conversion process, that chip is not a Sharc but something like Max-V or something, they sanded off the maker but you can see it in other pics, also what the ones I thought were the d/a chip in red are sanded off as well. And the output stage has many surface mounted discrete transistors in the second pic. 

 

Cheers George 

 

 https://www.fidelity-magazin.de/2018/09/14/marantz-praesentiert-limitierte-ki-ruby-edition/#foobox-1/14/Marantz-SA-KI-Ruby-23.jpg

 

 

Edited October 8, 2018 by georgehifi

[davewantsmoore]

1 hour ago, georgehifi said:

Here you can go to get some real close ups of the conversion process

Yeah, that's the pic I had ... I wanted one where I could see the markings on the 4x ICs.

 

1 hour ago, georgehifi said:

 that chip is not a Sharc but something like Max-V

Correct (that's what I said before) ... There are multiple SHARC DSPs in the digital side of the player.

 

The Max V is directly connected to the 4 ICs ... it's not obvious here, but it is clearer in other models (eg. SA-10).

 

1 hour ago, georgehifi said:

in red are sanded off as well

The ones in your red circle (in other higher res pics) look like they only have coded bands around them.

 

[kukynas]

those transistors are Vishay MMA, those 4 ICs  looks like SNXX bit shift registers, Altera FPGA front of it will be most probably just driver/logic and error correction in similar fashion like other transistor ladder DACs, but PCM to DSD conversion will be most probably done by SHARC DSP as Dave pointed out earlier, Altera won't have enough juice for that

It looks as very promising concept, something like combined Chord/Denafrips approach  

[georgehifi]

5 hours ago, kukynas said:

those transistors are Vishay MMA

Yeah Daniel, but why so many in the output stage, must be over 100 of them, unless the top half are doing I/V duties aka Krell CD300 did after the PMD100/PCM1702's with all their transistors, supposed to be nice sounding machine. 

And then there's all those lined up resistors, like discrete  low bit R2R as well on the Marantz. 

 

Cheers George

 

 

Krell CD300

Edited October 8, 2018 by georgehifi

[kukynas]

32 minutes ago, georgehifi said:

Yeah Daniel, but why so many in the output stage, must be over 100 of them, unless the top half are doing I/V duties aka Krell CD300 did after the PMD100/PCM1702's with all their transistors, supposed to be nice sounding machine. 

And then there's all those lined up resistors, like discrete  low bit R2R as well on the Marantz. 

 

Cheers George

it's not output stage as in traditional DAC architecture (I/V or whatever), that entire part after FPGA is basically discrete D/A conversion stage you would normally find inside of DAC chip but in this case with 1 bit architecture, may be Dave oversimplified it when he mentioned low pass filter but it's bit more complex, I'm not resistor ladder architecture guru but there's plenty of info on google and on DIYaudio forum (no DAC project and similar)

you can find more of such designs in typical discrete R2R ladder DACs such as Denafrips, Holo, Soekris and as many knows MSB, some pictures of similar concepts below

Marantz seems to take slightly more complex approach with combining of resistor ladder and some discrete stage but I'm just guessing here because there's not enough resistor to complete entire bit shift cycle, may be something on the other side? or they used different techniques how to do it...

 

 

 

 

 

[georgehifi]

21 minutes ago, kukynas said:

you can find more of such designs in typical discrete R2R ladder DACs such as Denafrips, Holo, Soekris and as many knows MSB, some pictures of similar concepts below

Yes they all use discrete R2R resistors, hundreds of them for 16 18 or 24bit resolution .

But here in this Marantz there's only what looks to be 28 of them (after what you say are 4 x shift registers) followed by the 100 x smd transistors.

Then after this comes the output rca's and xlr's

 

Cheers George  

[kukynas]

4 minutes ago, georgehifi said:

Then after this comes the output rca's and xlr's

 

Cheers George  

yes, sure, sorry I should have been more clear...

because this is 1 bit (DSD) conversion you don't need as much of them as in case of PCM conversion but not sure if 28 is enough 

[kukynas]

Just checked the specs and it seems they are using their HDAM discrete output stage instead of traditional opamps stage so most probably that’s why so complex looking...

[davewantsmoore]

7 hours ago, kukynas said:

something like combined Chord/Denafrips approach  

... nothing too fundamentally different from 'bitstream' DACs of old .... most of the goodness is in the more sophisticated modulator.   Like people are finding, you can really tailor it to the specific output, and get quite a lot of optimisation.

1 hour ago, kukynas said:

it's not output stage as in traditional DAC architecture (I/V or whatever), that entire part after FPGA is basically discrete D/A conversion stage you would normally find inside of DAC chip but in this case with 1 bit architecture, may be Dave oversimplified it when he mentioned low pass filter but it's bit more complex, I'm not resistor ladder architecture guru but there's plenty of info on google and on DIYaudio forum (no DAC project and similar)

The Marantz is not using a "resistor ladder architecture".     In simplistic terms, what is after the (Max V) chip is an amplifier and low pass filter - but they do appear to be doing some sort of signal summing (noise control, or 'dither' of some sort?!)

 

1 hour ago, kukynas said:

you can find more of such designs in typical discrete R2R ladder DACs such as Denafrips, Holo, Soekris and as many knows MSB, some pictures of similar concepts below

The Marantz DAC is not at all like these.

 

1 hour ago, kukynas said:

Marantz seems to take slightly more complex approach with combining of resistor ladder and some discrete stage but I'm just guessing here because there's not enough resistor to complete entire bit shift cycle, may be something on the other side?

 

See above...

[davewantsmoore]

51 minutes ago, kukynas said:

Just checked the specs and it seems they are using their HDAM discrete output stage instead of traditional opamps stage so most probably that’s why so complex looking...

That's right.... it's an "amplifier".

[davewantsmoore]

2 hours ago, georgehifi said:

Yeah Daniel, but why so many in the output stage, must be over 100 of them, unless the top half are doing I/V duties aka Krell CD300 did after the PMD100/PCM1702's with all their transistors, supposed to be nice sounding machine. 

Looking at other pictures of recent HDAM circuits, it seems pretty standard to use that many

 

....   A quick discrete buffer circuit like this on the back of an envelope (like a WJ buffer) could use ~15 per channel .... and that's just my caveman sketch of it  (Marantz could surely use more parts to improve the performance).

 

1 hour ago, kukynas said:

yes, sure, sorry I should have been more clear...

because this is 1 bit (DSD) conversion you don't need as much of them as in case of PCM conversion but not sure if 28 is enough 

 

As I said before, it doesn't work at all the same way as a resistor ladder architecture.   The long line of 28 resistors are a red-herring

[georgehifi]

3 hours ago, kukynas said:

it seems they are using their HDAM discrete output stage instead of traditional opamps stage so most probably that’s why so complex looking...

Yes they are, but HDAM outputs are no where near as complicated as that, as you can get discrete HDAM knockoffs on ebay.

The Marantz's one's that are in the output stage before the RCA and XLR outputs I counted at over 112 transistors!!! Ok a few could be smd regulators, but still???

 

Cheers George

Edited October 9, 2018 by georgehifi