optimising kwak clock

[ced]

Installed a kwak clock on my naim cd3.5 over the weekend and, miracle, I didn't short out half the ics! (Though there was a moment of panic when I connected the output up to the wrong pin on the board and the cd drive reached about 3000rpm!)
But I have to say it works very nicely indeed (so thanks alot for the schematic Elso!!!).
I'm etching a new board shortly so I can fit it nicely to the chassis with a much tighter layout but was wondering.... would the addition of a ground plane as a shield on the component side and another board sandwiched to the other side of the tracks, again connected at the star ground, so effectively shielding both sides of the tracks be of any benefit or just act as a large capacitor? I really have little idea about ground planes or shields so any thoughts would be welcome.
Ced

 

[Elso Kwak]

KWAK-CLOCK

Hi Ced,
Glad to hear it is working nicely.
I considered a groundplane on the topside (componentsside of the PCB) but settled for a groundplane at the solderside. All space not occupied by components or tracks is groundcopper.
I made 3-dimensional clocks just by connecting the components in free space. It looks like a modern work of art or a exotic insect but it works. I did not hear any difference between these prototypes and a neat PCB.
Schematic version 6 is just completed. More bass-slamm and punch in the sound. Less glare. Still wondering how a 16.9344MHz clock can influence the sound as perceived

 

[ced]

Hi Elso,
thanks for that reply. The reason I asked was the datasheet for the AD8561 indicates that a ground plane is recommended for high speed performance (avoids ground bounce and eliminates stray capacitance) but numerous posts on this forum indicate you need to know current loops etc, and basically know what you're doing to design a good ground plane. I've star grounded instead but thought that some shielding might be effective- apparently this is the case with the didden jung regulators for example. I guess I'll just try your ground plane idea instead.
Elso, do you think a low jitter crystal would improve performance also or does it's quality really not matter?
And I know what you mean. Subjectively it seems to me that music is just cleaner and more sharply defined which improves everything slightly. I suppose logical if waveforms are being produced slightly more accurately.
cheers
Ced

 

[Elso Kwak]

"Low Jitter Crystal"

Elso, do you think a low jitter crystal would improve performance also or does it's quality really not matter?
And I know what you mean. Subjectively it seems to me that music is just cleaner and more sharply defined which improves everything slightly. I suppose logical if waveforms are being produced slightly more accurately.
cheers
Ced

Hi Ced,
The question for a low jiitter crystal I get over and over again.
Just what is a low jitter crystal AT-cut, BT-cut or.......?
I usyually advice to use the crystal from the CDP as you have it to remove anyway and it has the right frequency.
The qualifications I was giving with respect to the sound were for the new version 6. You builded version 5. The new clock is the "breakthrough" design and it is pitty it is not number 7. It is not difficult to modify version 5 to version 6.
Thanks go to Wildmonkeysects and On for stimulating suggestions for this ongoing project

Good luck.

 

[maczrool]

Do you have any jitter specs on your clock?

Stu

 

[Elso Kwak]

Jitter Specs

Do you have any jitter specs on your clock?

Stu

Hi maczrool,
No I don't have any jitter spec. A friend who builded my clock5 is working on that but the results will not be obtained before Christmas. If I may believe Guido Tent the jitter spectrum is also of importance but I don't own that kind of 30k$$$ equipment. And if I had 30k$ I would............. ......... and ...

 

[capslock]

Hi Elso,

could you mail me the current version of your project? Thanks!

ced,

regarding low jitter crystals, it might well be possible that some cuts have certain advantages. I would especially like to see results with those new telecom mesa-etched crystals, but they are not available in multiples of 44.1 kHz.

Anyway, all the crystals we usually see are AT-cut. An argument can be made that the higher order resonances are sharper, so an overtone crystal might be advantageous. On the other hand, it complicates the oscillator circuit. Pick your poison.

One thing is clear though: try to find the lowest series resistance you can. Those small cans which are only 3 mm high have small crystals. Typical R_s values are 2-3x what you get from the medium sized can (HC-49 or something). So even if you don't have access to specs, picking the biggest can you can find is a good idea.

Eric

 

[johnny1]

Hi Elso..
Is it possible to mail me a copy of version 6?
Thanks in advance

 

[ced]

crystals

Hi capslock,
Well there is one parameter already that sounds promising, and indeed looking at the crystals from the likes of LCaudio etc, they are all the large can ones not the small ones you mentioned and indeed were in my CD.
There seem to be several other parameters to crystals like 'frequency/temperature stability, calibration tolerance and operating load capacitance'; knowing what ones affect sound and what are irrelevant is the rub.

It does seem logical that the crystal is going to have an effect on ultimate performance like any other component in the circuit.
Having said all that, Elso's circuit really is good! Was listening this morning and after 4 days burn in it is sounding better and better.

ced

 

[Tazzy]

Would a kwak clock make any sense in a DacMagic II?

 

[ced]

I believe Elso has a version for an outboard dac. Do a search on kwak and check the threads. If its what you want Elso was really helpful and sent me a schematic very swiftly.
Ced

 

[Elso Kwak]

Low Jitter Crystal

Hi All,
Capslock is right the cystal in the CDP is usually AT-cut. That seems to be a good compromise between frequency-tolerance vs. temperature. Theoretically a larger crystal is better than a small one but in real world I did not hear any difference between a HC-49U (large) and a HC49S (small) 16.9344 MHz crystal.
I have received a e-mail that a BT-cut crystal is better with respect to jitter but I did not found a source yet.
Overtone crystals might be usefull but these oscillate at the fundamental in the KWAK-CLOCK! F.a. a 33.8688 MHz crystal used for Panadonic MashDac oscillates at 16.9344 Mhz in the KWAK-CLOCK. Useless!
Very recently I found a supplier that can offer fundamental mode 33.8688 MHz and 45.1584 MHz BT-cut crystals. The latter frequency being used in a lot of Sony CDP' s and SACD-players. Hopefully I can get these crystals as I am not a OEM. Oops again ugly three letter word= Original Equipment Manufacturer.
Inverted mesa technology crystals are interesting as these oscillate at the fundamental too at much higher frequencies i.e. above 45MHz. I will try a inverted mesa 100MHz in the Asynchronous Reclocker.

The version of the KWAK-CLOCK for a outboard DAC is essentally a Asynchronous Reclocker with a low noise oscillator and 74VHC74 flip-flops. I look at it as a anti-jitter device. It is installed between the digital input receiver and the DAC-chip in a NON-OS DAC or between the digital filter and the DAC. See also Fedde's site:
http://home.student.utwente.nl/f.s.bouwman/audio/nonoz.html to get a idea of the implementation.
The Asynchronous Reclocker can also be used successfully in a one-box CD-player.

All schematics are available for DIY-use by hitting the e-mail button underlining this post.

 

[Elso Kwak]

Re: Correction

Hi All,
F.a. a 33.8688 MHz crystal used for Panadonic MashDac oscillates at 16.9344 Mhz in the KWAK-CLOCK.

<B>CORRECTION</B>
That must be:
F.a. a 33.8688 MHz crystal used for the Panasonic MashDac oscillates at 11.2896 MHz in the KWAK-CLOCK.

I tried to correct but 30 minutes were expired! Time out.

 

[Brian Guralnick]

It's a third overtone crystal. You need a tuned inductor & cap to ground across the crystal input.

 

[Elso Kwak]

Overtone Oscillator

It's a third overtone crystal. You need a tuned inductor & cap to ground across the crystal input.

Hi Brian, I wished it would be that simple........

 

[Oli]

Why does Kwak not use a comparator with his asynchronous reclocker?

A Kwak-clock related question...

Elso uses a comparator on his clock (version 7) but does not use one on his asynchronous reclocker. Would there be any advantage to using a comparator after the colpitts oscillator to square-up the sine wave before feeding the flip-flops within the reclocker circuit? Is there a reason why a comparitor is not used, e.g. excessive dv/dt on rising edges?

 

[Elso Kwak]

Re: Why does Kwak not use a comparator with his asynchronous reclocker?

A Kwak-clock related question...

Elso uses a comparator on his clock (version 7) but does not use one on his asynchronous reclocker. Would there be any advantage to using a comparator after the colpitts oscillator to square-up the sine wave before feeding the flip-flops within the reclocker circuit? Is there a reason why a comparitor is not used, e.g. excessive dv/dt on rising edges?

Hi Oli,
The reason is very simple:
When using the AD8611 in the ASR sound was worse than without a comparator. The reverse is true for the regular KWAK-CLOCK.

 

[Oli]

Perhaps less is more

The man himself speaks with authority!

Cheers Elso, I was just wondering about the comparator, but it seems that you have covered this ground already and come to a empirical conclusion. i.e. it sounds better. Perhaps the comparator does inject nasty rising edges into the ics. Perhaps the principle of "less is more" applies. In either case it makes the circuit easier to build!

 

[Elso Kwak]

Hi Oli,
Well the idea was more to use a very fast comparator as the frequency is about 66 MHz. But the AD8611 is a nasty part, also prone to oscillations and giving bad sound. And the CD expects a squarewave. What the flip-flop expects I dunno, but it is working.

 

[Oli]

That probably explains it.

That probably explains it. As for what the flip flops expect, I suppose the key thing is that the wave is pure, hence consecutive peeks appear the same shape in order to keep the jitter low.

The slower rise time offered by the sine wave (as opposed to a square wave) must reduce ground bounce and potential multiple logic transistions. I am aware that designers purposefully place resistors in digital lines to reduce currents entering input stages to reduce ground return current and r.f garbage. In addition such resistance acts together with input capacitance to produce an RC low pass filter. Excessive bandwidth can indeed be a bad thing.