Real or fake PCM63?

[PA0SU]

reclocking near the DAC-IC....

Ofcourse GuidoB, you are absolutely right! The layout of the PCboard is at least as important as the jitter-figures. I forgot this in my last posting. This is one of the very nice things of the 'TentDAC' digital board. The perfect layout on a four-layer-board.....

Bernhard, reread my posting about jitter again and do not forget the last alinea's.....

By the way: I do NOT use a VCXO with PLL or so. I route the clock signal in my cd-player directly to the DAC.....

 

[guido]

Maybe I should not speak up for PA0SU, but I think he was thinking of the problem of reclocking the BCK signal in case of the stopped clock operation, which would mean to apply that trick from the Tent Dac. So it's absolutely not only squeezing in a '175.

Ciao, George

Ps: When I was doing that digital PLL mod to my dac:

http://www.diyhifi.org/forums/viewtopic.php?p=26280#p26280

Then it was quite an amount of change in the sound, so maybe that "tremendous" is not so far from reality..

Ciao, George

I've seen the ammount of logic around the reclocking of the BCK signal in the schematic (compared to the other signals), i did not investigate further why it's there. Could you elaborate this, if you know why?

Don't see the connection between reclocking or pll?

 

[guido]

Re: reclocking near the DAC-IC....

I route the clock signal in my cd-player directly to the DAC.....

Try the other way around... You're not using the original clock of the player, are you


Hope this is still a bit on topic. Like "how to get the best envirounment to test if a PCM63 whatever is fake or not"

 

[Joseph K]

Guido,

"Could you elaborate this, if you know why?"

But YOU got it already, don't remember?

http://www.diyaudio.com/forums/showthread.php?postid=954265#post954265

Connection between reclocking and PLL: wanted to say, any manipulation with the jitter values seems to have a strong effect.
I remember when I reclocked my onkyo back in '98: the signal coming out of the digital filter had 1nsec peak to peak jitter [It was clearly visible, after some fiddling with a TEK TDS640 in persistence mode], and after the D flop I could not measure / see only the residual of the scope [45 pSec]

Now I would test it differently, especially if Herb would inform us about his phase noise setup [not the Wavecrest]..

Ciao, George

 

[Bernhard]

But. The title of the topic is: real or fake PCM63? Low level nonlinearity has nothing to do with factory grading - no wonder that we did not find correlation.
For real K or superior level grades we should test at FULL output level, and look for THD ~-96dB, for a K. Problem is, that You will not be able to do that with just an average sound card, and neither Your analyzer, Bernhard. Maybe mine, a HP3585, but I will have to control. And also the output stage in the test-bed dac should be up to the task. Better semi-pro sound cards would be good, like probably the EMU series.

Then the question is: does it really have a strong correlation with good sound, the better low level linearity? For me, I could not confirm that in my setup, but will have to spend more time on it. But still, if one feel it's important, then why don't just use a modern sigma-delta dac? They have excellent low level linearity, by definition.

Or, use a digital filter scheme before the multibit converter with correct dithering, like the PMD100 does, as I had shown here.
Or use an adjustable old converter, like the PCM58, and adjust the hell out of it.. By the way, on the contrary to some opinion here, I still enjoy very much the sound of my pcm58, I find it more relaxed and natural, then the pcm63... Though less attraction and fireworks...

George,


did you read the data sheet

Low level nonlinearity is the most important issue for factory grading.
That the grades are not reliable is another story.

That's why it would be the only possible proof for the Y chips that they are no fakes.

QUOTE from the datasheet:

Dynamic range, THD+N at -60dB Referred to full scale.

typ. 100, 104, 108dB for plain, J, K

That is low level performance: distortion + noise, nothing else.

108 dB absolute means exactly the same as 48dB relative to -60dB analyzer input signal as stated below:

QUOTE from the datasheet:

K grade typical -48dB at 991Hz / -60dB

Data is given also for 0dB and -20dB, but most data and most important is about low level.

For comparison: Theoretical dynamic range for full scale 0dB would be 20 x 6 dB = 120 dB


So for real K or superior level grades we should not test at FULL output level, and look for THD ~-96dB, for a K.

So my analyzer is more than conveniant

Also I can confirm the superb sound of a DAC with chips that have good low level nonlinearity.

I tested PMD100 with PCM1702, not very good.

I also tested a bunch of PCM58 and found them 'unadjustable'
Adjusted or unadjusted, PCM58 is garbage.

PCM56 is the chip to go with.

The dynamic range in the BB datasheet is not measured at 0 dB / full scale.
The dynamic range in the BB datasheet is not measured at 0 dB / full scale.
The dynamic range in the BB datasheet is not measured at 0 dB / full scale.

 

[Joseph K]

Bernhard,

You are partially right - I forgot about the -60db part in the data sheet, BUT also You should not forget that the 0dB specification is given for a K grade, too!
That is,
at 0dB = -100db typ
-20dB = -88db typ
-60dB = -48dB typ

Now, ALL my chips are exceeding that -48dB [typical, -44dB worst] specification!
Even the non K chips, and they are even better..

Then don't forget that MSB errors will show up only above -6db, so I would expect more difference in that range. You know it best, that the PCM63 is different from it's predecessors, [colinear] while for a PCM58 - 56 it's fully enough to test for -60dB, for the '63 you should test also for full range.

Then I'm not questioning your experience sound wise, personally in a fast try I could not confirm, but that's my problem.

Why do You say it's un-adjustable that pcm58 ? I had shown that graph, definitely you can get better than a 63K, and I did not finish yet..

Ciao, George

 

[Bernhard]

Only 25% ( 1x 0dB, 1x -20dB, 2x -60dB ) of data is for full scale, I wrote that:

Data is given also for 0dB and -20dB, but most data and most important is about low level.

The dynamic range in the second line of data table is for -60dB low level, not full scale.

Your graph for PCM58 shows -123dB ( or -60dB relative to input signal ) for K8 ( acid high order harmonic ) .

Adjusting bit 15 and 16 will not help here.

Your PCM63 shows a little more sweet K2 but I would rate it better.

 

[Joseph K]

Bernhard,

I don't want to discuss what is important soundwise, I'm only searching to test ALL relevant parameters before saying that testing these chips is senseless..
And MSB variation is an important factor, and while testing a 1541 you see the dist. due to it already with -60dB, in the case of pcm63 you do not see it, only with 0dB signal. This is why BB guarantees only -100dB THD for full output, instead of -108dB for -20dB test signal..

Ciao, George

 

[Joseph K]

Bernhard,

"Your graph for PCM58 shows -123dB ( or -60dB relative to input signal ) for K8 ( acid high order harmonic ) ."

If you take a look of the dithered output from a PMD100, You can note that it eliminates exactly those high harmonic components..

So, if it's a problem, just stick a good digital filter in front of that PCM58 and all will be sweet..

 

[Bernhard]

.
And MSB variation is an important factor, and while testing a 1541 you see the dist. due to it already with -60dB, in the case of pcm63 you do not see it, only with 0dB signal. This is why BB guarantees only -100dB THD for full output, instead of -108dB for -20dB did you mean -60dB ? test signal..

I think you mix up a few things.

100dB for full scale
88dB for -20dB
48dB for -60dB

Distortion goes up while level goes down.

Same effect as in class B amplifier: Crossover distortion.
Good chip sounds like class a amplifier.

 

[Joseph K]

Bernhard,

No, this time not:

I think you mix up a few things.

100dB for full scale
88dB for -20dB
48dB for -60dB

Exactly, as shown: suppose that You have a fixed value low level distortion, at -108 dB from full range. Then, in an increasing test signal order, You should get:

-48 dB at a -60dB test tone [48 +60 =108]
-88 dB at a -20dB test tone [88 +20 =108]
-108dB at a 0db test tone [108+ 0 =108] >>

BUT in this last case the factory specification suddenly falls back to only -100dB !!
What had eaten up those 8 dB if not the increasing distortion at full output?!

Ciao, George

 

[Bernhard]

What can I say ?

The distortion is not fixed.

If you listen to a 0 dB 1k sine it sounds much purer than if you listen to a -60dB 1k sine because the distortion of the 0dB sine is -100dB and the distortion of the -60dB sine is -48dB.

The problem of the colinear is the mismatch of the 19bit DACs and the MSB as it is always the MSB that makes trouble.

Higher distortion at higher level is more natural than the other way round.

 

[Joseph K]

Bernhard,

as I have already told, I'm searching clues for testing, and do not want to discuss the audibility aspects.. What bothers me that I did not see a difference for my standard chips, when applying the -60dB test signal. Would like to confirm if it's still holding up with full scale signals as well?
If I'm not wrong, then Beauty_Divine wanted to do the same, not?

Ciao, George

 

[guido]

Guido,

"Could you elaborate this, if you know why?"

But YOU got it already, don't remember?

http://www.diyaudio.com/forums/showthread.php?postid=954265#post954265

Woops... I didn't remember, but after re-reading it, it makes sense again.

Had a look at this back then and continued with i2s. Until i got hold of those pmd100's and started playing with those. replaced a 7220 in a philips player

Should join this pcm63 measurement exersice, just to get some skill using the spectrum analyser. Got two K's from japan and two from korea, both 1997. Mmm

 

[PA0SU]

Re: Re: Jitter....

Unfortunately I have no instruments up today to measure jitter.
I use Accuphase DP70 as a transport and after putting in a Kwak Clock, things improved significantly.
But the changes in sound quality from the low jitter clock have nothing to do with low level performance of DAC chips.
Two totally different things.

I'm so sorry for mr. Kwak, but his clock is at least 20 dB worse than most Tent-XO's looking in the frequency domain. (I did not check the jitter with Henk on his WaveCrest....) If you have any notion of oscillators (it is a pitty that most of you do'nt read Dutch otherwise you could find the theory and practice of how to build a GOOD oscillator on my web site) then you could know that this oscillator produces a lot of noise: try to find out how this oscillator stabilizes in amplitude.... There is no agc....

Bernhard, I make the statement that the low level performance in a 'jittery' environment far overwelms the distortion in the DAC itself !!!!
The distortion you measure is a question of amplitude-distorsion, but the ear is (also) very sensitive to the precise zero-crossings of the audio-signal. Errors here not only 'enlarge' the sound sources in the audio image (in width and depth) but also makes the sound harsh. Pure amplitude-distortion is not that bad moreover when it is about 40 dB under the noise threshold of the recording.
It seems strange, but heavily amplitude-compressed audio-recordings (thus, with a small dynamic range) sound mutch better with a low-jitter-clock.........

 

[PA0SU]

jitter in XO and VCXO

For all certainty: VCXO's always produce more noise/jitter than XO's.

 

[Bernhard]

Re: Re: Re: Jitter....

I'm so sorry for mr. Kwak, but his clock is at least 20 dB worse than most Tent-XO's looking in the frequency domain. (I did not check the jitter with Henk on his WaveCrest....) If you have any notion of oscillators (it is a pitty that most of you do'nt read Dutch otherwise you could find the theory and practice of how to build a GOOD oscillator on my web site) then you could know that this oscillator produces a lot of noise: try to find out how this oscillator stabilizes in amplitude.... There is no agc....

So why not just post a competing clock ?

 

[Joseph K]

(see my web site www.by-rutgers.nl)

Ciao, George

Edit: look for: Reproducible Low Noise Oscillators

 

[Joseph K]

By the way, dear PA0SU, I have to admit You don't make life easy for us..

I have found in Your site essentially all that I wanted to ask - [the articles about pre- post DBM amplifiers - exactly what I missed for a phase noise setup...] but not only in dutch, but also in graphics PDF... if it were at least some kind of text, one still could use a translator..



(the smilies are for "desperation" )

Edit: uhm, having had a better look, the post DBM amp I need is a low noise DC amp, your's not apply..

Also, it's very interesting, never heard about it, do You have it better explained somewhere?

but the ear is (also) very sensitive to the precise zero-crossings of the audio-signal. Errors here not only 'enlarge' the sound sources in the audio image (in width and depth) but also makes the sound harsh. Pure amplitude-distortion is not that bad moreover when it is about 40 dB under the noise threshold of the recording.
It seems strange, but heavily amplitude-compressed audio-recordings (thus, with a small dynamic range) sound mutch better with a low-jitter-clock.........

Ciao, George

 

[PA0SU]

Good clock & how to measure noise

What happens now is what I was afraid for: people want to know my experiances which has been built up in my nearly 70 years life. This would take the rest of my life!
I published a lot in the field of radio receivers (and transmitters) oscillators and so on. This is one of the reasons Guido Tent asked me as a free lancer for pre-development. You will understand that I do not want to publish all my new findings here, because of Guido.....

Nevertheless:
Using the scheme of the oscillator in: 'Reproducible low noise oscillators', be careful with Xtals. The suggestion there to replace the LC with a Xtal, will dammage the Xtal because of overdissipation. In that case the BAT81 should be connected to earth. Also look at the here attached scheme of my DC-receiver with which I measure noise-figures of oscillators.
Sorry George, no further explanation of the cicuit. I could send you the project-documentation (I always write during development) with a lot of, for you, redundant information, but I'm afraid for another barrage of questions......