Advice on Crystal Oscillators

[Herbert]

Hi,
I am not looking for overpriced oscillator boards to upgrade your CD-Player that promise 0.1ppm.
I am just looking for the xtal-cans themselves. Still the aim is upgrading a CD-Player, in my case it is 16.9344mHz.
Looking @ Mouser and Digikey it looks like the best value is 10ppm. I understand the data of the pF needed,
but not the data of ESR - should resistance be higher in order not to stress the circuit?
And what about voltage. Usual voltages in older CD-Player logic circuits were 5V, later it became 3.3V.
Are Xtals dependend on that? Any brand to look for?
All the best and thanks,
Herbert

 

[DonH56]

Crystal oscillators are a huge subject. The frequency spec (ppm) and drift affects accuracy but not really the circuit itself. Crystal oscillators are designed for a given amplifier, load capacitance, shunt capacitance, and ESR among other things. Changing one thing affects everything else. The frequency spec is good for a given capacitance and drive level, which in turn depends upon the ESR and amplifier. The drive level is a key parameter because, if it is too high, the crystal can be damaged. Voltage is set by the amplifier, and the swing is part of what determines the drive level. For a crystal oscillator, amplifier gain is typically expressed as -R (negative resistance = gain), and you need to know that to do an accurate design. Output noise (jitter) is usually determined by the amplifier, not the crystal, though again they are interrelated. And the amplifier's characteristics, such as gain stability and sensitivity to noise and temperature, also has an impact on the output signal that often swamps whatever the crystal itself is doing.

Searching a bit, here is an Epson description of drive level and load pull for a crystal circuit. It does not have the equations but provides an overviw of the issues. https://www5.epsondevice.com/en/information/precaution/design.html

I imagine there is a lot of leeway in the design, but without knowing the parameters of the components around the crystal it's a crap shoot. Just swapping a crystal could provide better or worse performance, and worst case could end up damaging the crystal.

HTH - Don

 

[Herbert]

Hi Don,
thanks a lot for your kind reply. The xtal to be replaced was used in a very common setup
in CD-Players of the early eighties using the Sony CX23034 digital filter. The CX23034 had also a divider
to supply the circuitry with the system clock derived from 16.9344. Unfortunately, no data is given
About the drive levels in the CX23034 datasheet as well as in the circuit drawing...
All the best,
Herbert

 

[DonH56]

The load capacitors are those two 10 - 30 pF caps on either side of the crystal. They are usually adjusted for the particular crystal and amp to provide the correct frequency and drive level. The crystal has shunt capacitance from the package and motional RLC values representing the resonator (crystal) itself. Unfortunately, while I know a bit about crystal oscillator design and such, I know virtually nothing about the specific circuits used in CD players. I defer to @restorer-john , @SIY , @Wombat, @Scott wurce, and others who have more relevant experience.

Just plugging in another crystal could change the frequency, noise, and potentially damage the crystal. A lot of the newer smaller crystals I see in my day job have much lower drive level than the larger crystals used in the past (e.g. 100 uW instead of 1 mW or more). Chances are you'd never know, nothing bad or good would happen, and life would go on with the CD player working as always.

Are you using the analog outputs of your CD player? If not, this does not matter. If so, it may not matter anyway; depends upon how the DAC is implemented.

I strongly suspect there are other changes you could make that would have much greater sonic impact than changing the crystal.

 

[Herbert]

Thanks for your thoughts. The reason I want a new Crystal: I am stripping out the complete DAC section of a Sony CDP-103
and implementing a modern DAC like the Khadas Board
In this Player, the system clock is produced close to the D/A converter chip using an 67,7376 xtal.
Dividers feed the filter and Decoder. The D/A converter and analog stage are to go.
This is why I need a 16.9344 xtal at the digital filter. This still takes lesser space than
using an external board with a Clock.

 

[solderdude]

chances are the circuit won't work properly with a freq. 4x what it was designed for. It may work but possibly may not work optimally.
You can give it a try.
In my younger years, before the internet, I built pitch controls in CDP's for DJ's and dance schools.
It consisted of a VCO which was connected to the X'tal input pin.
I already soon found out that some CDP's could easily do +/- 20% but other players not even a few percent.
Of course this was also mechanically related with the servo circuits and stuff.
There were none that one could push much more than 20% though.

X'tal oscillators (the ones with 4 pins) can be be more stable than a simple X'tal or ceramic oscillator (which are really crappy).

 

[Herbert]

chances are the circuit won't work properly with a freq. 4x what it was designed for. It may work but possibly may not work optimally.
You can give it a try.

I think this is misunderstood. In the old CD-Player with Sony Cipsets CX23035 (Decoder), CX23034(Digital Filter) and DAC (CX20152),
one Xtal provides the base frequency. It is divided for the needs of every chip.
In my case, the Xtal is feeding the DAC with 67.7376. The Dac also sends the clock signal to the digital filter
But the filter needs 1/4 of the frequency, this is why the DAC already has a divider on board to generate 16.9344 for the filter.
The filter then sends the clock signal to the decoder.
But the decoder now needs 8.4672, so also the digital filter has a 1/2 divider fon board.
Because the D/A section has to go with my mod and because the chip newly implemeneted (DIT4096) needs 16.9344
to run, the filter CX2304 will stay. It can alternatevily generate a clock signal itself,
like in the schematic attached.
So it is just about finding a decent Xtal...

 

[solderdude]

The question remains if all the chips are designed to operate for the 4x higher clock each chip gets.

Just could buy cheap oscillator (67.8 Mhz) to check if it works in the first place before looking for a high quality VCO or custom made X'tals.
Would have to feed it with 3.3V and capacitor couple it to the X'tal input pin (1nF or so) to try.

 

[Herbert]

The question remains if all the chips are designed to operate for the 4x higher clock each chip gets.

To my understanding it does not. Again, this is the original setup:
DAC CX20152 is equipped with Xtal 67.7376. DAC CX20152 has a 1/4 divider on board >16.9344 to CX23034.
CX23034 has a 1/2 divider on board > 8.4672 to CX23035.
CX23034 can be fed from the DAC´s clock OR be equipped with its own Xtal.
In the CDP-103 the Dac gives the clock.
In the CDP-552ES -from which I took the schematics-uses PCM53 DACs from Burr Bown
They have no Xtal input. So the Digital Filter CX23034 has the configuration with it´s own Xtal.
Maybe the misunderstanding comes from that I am talking of the CDP-103 but show the CDP-552 schematics.

But anyway, my assumption is wrong: The CX23034 can drive it´s own Xtal - but only outputs the 1/2 frequency.
But I need also a second undivided output for the SPDIF circuit - so I need an external board anyway, that provides
several clock frequencies. I already own one.

So my question would be only relevant for those, who simply want to upgrade their player with a better xtal.

 

[solderdude]

Ah you are not replacing a 16.9MHz X'tal with a 67MHz clock. Just looking to replace the 67MHz clock with a less noisy one.

 

[Herbert]

No - I wanted to remove a chip (The DAC CX20152) with his 67MHz Clock attached.
But he DAC provides the system clock for the whole player.
The DAC is equipped with a 1/4 divider for the clock output = 16.9MHz.
Next in the clock chain comes the digital filter:
It can be clocked by the DAC OR it can be fed by a 16.9MHz xtal. (See schematics)
So I was looking for a decent / less noisy 16.9MHz XTAL.

Sony used two designs for distributing the Clock in their early players:

Design 1) 67MHz Clock equipped DAC > Digital Filter > Decoder (i.e. CDP-102 / CDP-103 / CDP-302)

Design 2) Clock equipped digital filter > Decoder. (All higher end models like CDP-552, but also Nakamichi, Accuphase etc.)
In these designs the 3rd Party DAC -like Burr Brown´s PCM53 or PCM56- is clocked by
the parallel data from the digital filter. The filter can be third party as well like from NPC.

The CX23034 Digital Filter was designed to work in both scenarios.

But again, my Idea is obsolete, as the CX23034 only provides a divided Clock output of 8.4672mHz
But I need a second undivided one for the SPDIF-Converter.
So I need to fit in a board that distributes 16.9344 (SPDIF-circuit) and 8.4672 (for the Decoder) anyway.
And I already have one.

But still my question might be relevant for those who have a CX23034 based player that runs on design 2.

 

[DonH56]

I still don't quite follow, but if you use two independent oscillators instead of dividing down from one, you will have to incorporate some means of synchronizing them to prevent bit errors, cycle/phase slips, and such. Unless the chips do that?

Epson and Fox are suppliers we use in my day job for GHz serial data transmission.

 

[Herbert]

I am not using two Independent oscillators.
The basic question of this thread is just what 16.9344 oscillators can be used
with the CX23034 digital filter. I have a CDP-552, when there is time I will open it to find
out what brand was used. Then I might find a datasheet and post it here...

 

[Herbert]

Again, it is just about fiting a CX23035 with an oscillator. See post #3.