Below is a portion of a schematic from a classic Philips CD player.
The IC on the left is the SAA7210 (decoder) and the IC on the right is the infamous SAA7220 (oversampling DF).
Notice internal "X-Tal Oscillator" in both ICs. Their respective pins connect to common 2-pin crystal oscillators.
There has always been controversy as to how effective these IC's internal oscillator-drivers are. To improve performance, many external clocks have been developed by Tentlabs, or even DIY projects. I've used them myself.
But suppose all one wants to do is re-use the stock 2-pin oscillator, and keep the design simple?
To that effect, lets say we keep that orig. crystal and connect it to something like this:
SN74LVC1GX04 Crystal Oscillator Driver | TI.com
Has anyone used a driver like this? If so, what were your results? Speculate on any performance upgrades over the SAA72x0 internal drivers.
Thx!
The IC on the left is the SAA7210 (decoder) and the IC on the right is the infamous SAA7220 (oversampling DF).
Notice internal "X-Tal Oscillator" in both ICs. Their respective pins connect to common 2-pin crystal oscillators.
An externally hosted image should be here but it was not working when we last tested it.
There has always been controversy as to how effective these IC's internal oscillator-drivers are. To improve performance, many external clocks have been developed by Tentlabs, or even DIY projects. I've used them myself.
But suppose all one wants to do is re-use the stock 2-pin oscillator, and keep the design simple?
To that effect, lets say we keep that orig. crystal and connect it to something like this:
SN74LVC1GX04 Crystal Oscillator Driver | TI.com
Has anyone used a driver like this? If so, what were your results? Speculate on any performance upgrades over the SAA72x0 internal drivers.
Thx!
Last edited:
1. The frequency at pin 11 and at pin 9 (19) may be different due to multiply by / divide by.... can't tell without checking the pdf's, but this would be one of the functions of the "IC's internal oscillator-drivers". The other one would be - fixing the waveform shape/amplitude to suit different IC's needs. There may be different impedance requirements as well, compared to XTAL-only "drive" ("raw" XTAL signal cannot be used to drive anything directly). A quick check with the oscilloscope, with probe set to 10:1, will reveal a lot...
2. Search the web for "KWACK-CLOCK-7.gif" - see how Elso decided to buffer the oscillator signal. If you can't find any info, go to "the other side" and ask over there -> politely!
3. I personally used third-party oscillator boards with dual outputs, to overcome the issue you are referring to; i.e. the first output would go to pin 11 on one IC, and the second output would go to pin 19 on the other IC
2. Search the web for "KWACK-CLOCK-7.gif" - see how Elso decided to buffer the oscillator signal. If you can't find any info, go to "the other side" and ask over there -> politely!
3. I personally used third-party oscillator boards with dual outputs, to overcome the issue you are referring to; i.e. the first output would go to pin 11 on one IC, and the second output would go to pin 19 on the other IC
Last edited:
My clocks
Frankly, I no longer have the DIY drive to build clock projects, such as the KC or Flea. That's why inquired about anyone with experience with single-chip solutions (as in the case with the TI driver).
If one Googles "oscillator driver circuit", some interesting stuff pops up, including:
...etc...etc...
I think the type of electronic oscillator required for digital audio -- among myriad types -- is a Pierce oscillator??? Correct me if I'm wrong. Or would other types work as well?
The TDA1541 requires a divide-by-two (I've built several of those). For my current project, based on a TDA1387, I won't need a divider.
I built a KC7 years ago and have the schema saved. I also built a PFM Flea, and a few simpler clocks
I had no problem attaching three separate leads (for decoder, DF and DAC) to the same (and only) clock output of KC7, Flea, etc.
Frankly, I no longer have the DIY drive to build clock projects, such as the KC or Flea. That's why inquired about anyone with experience with single-chip solutions (as in the case with the TI driver).
If one Googles "oscillator driver circuit", some interesting stuff pops up, including:
...etc...etc...
I think the type of electronic oscillator required for digital audio -- among myriad types -- is a Pierce oscillator??? Correct me if I'm wrong. Or would other types work as well?
Last edited:
What you call a "2-pin crystal oscillator" is actually a crystal. The oscillator is inside the chip - it is the section marked 'oscillator'! It is likely that the external TI part you mention is very similar to the internal circuit in the chip, so little advantage in swapping one for the other. You just add the complications of interfacing.
Oscillators are a big subject. Oscillator noise/jitter (two ways of looking at the same phenomenon) is a bigger subject. For audio what matters most is mid-frequency jitter and this will be determined mostly by things such as power rail noise, grounding and crystal quality. Other applications of crystal oscillators may be more interested in long-term frequency stability or wideband noise, which matter less for audio. The fact that someone says that a circuit is a good crystal oscillator does not necessarily mean that it optimises the things needed for audio; indeed improving one aspect of performance may degrade others.
Oscillators are a big subject. Oscillator noise/jitter (two ways of looking at the same phenomenon) is a bigger subject. For audio what matters most is mid-frequency jitter and this will be determined mostly by things such as power rail noise, grounding and crystal quality. Other applications of crystal oscillators may be more interested in long-term frequency stability or wideband noise, which matter less for audio. The fact that someone says that a circuit is a good crystal oscillator does not necessarily mean that it optimises the things needed for audio; indeed improving one aspect of performance may degrade others.
When I added a Kwack Clock or Flea Clock to a DAC or CDP, sound quality always improved despite "complications" such as lead length (connection wire from clock to various ICs). Devices such as SAA72x0 are pretty noisy. So I'm sure clean, isolated power was a big factor.
Both the KC and Flea are mostly thru-hole circuits, and I'm sure all the trace length doesn't help.
With the TI driver (at $0.25-0.50 ea.) , one turns a 2-pin crystal into a 4-pin oscillator. Tentlabs VCXO cost $66.39 US Dollar. So some economy to be realized, perhaps.
This would allow the device to be used with, say, a Flea clock, or a Tentlabs clock. Might be interesting.
Not sure about which manuf. of crystals or 4-pin oscillators uses high-quality crystals.
I have some Tentlabs 4-pins in my kit (but not the $$$ Tent VCXO).
Others mentioned Russian-made Morion being pretty good.
Are there other goodies?
I use a tuned plate, crystal grid oscillator and it works.
The advantage of crystal oscillators is that they are accurate and have very little close-in phase noise, so it certainly makes sense to use a crystal oscillator.
Compared to other crystal oscillators, Pierce oscillators have several advantages that make them very popular. (I use the term Pierce oscillator in the general sense: an inverting voltage-to-current amplifier with two capacitors and a crystal, no matter how it's biased.)
A fundamental-mode Pierce oscillator doesn't need any inductors.
Because of the filtering effect of the load capacitors, a fundamental-mode Pierce oscillator has practically no tendency to unintendedly oscillate at an overtone.
This filtering also reduces the impact that harmonics generated in the active part have on the frequency.
The resonant current doesn't need to flow through the amplifying part but only through the crystal and the load capacitors, so without disturbing the resonant current, you can very simply limit the amplitude by letting the amplifying part run out of current.
Of course none of this means that there are no other options!
As a general rule if you are concerned about phase noise or jitter and not concerned too much about correct frequency (i.e. digital audio) then it may be best to avoid VCXO. The reason for this is that the arrangements to give voltage control of frequency (e.g. so the oscillator can be used in a PLL) will degrade phase noise. The same applies to temperature compensated oscillators. Audio requirements can sometimes be the opposite of most other applications of crystal oscillators. The only reason to choose a VCXO or TCXO would be the assumption that a precision oscillator may (we hope!) use a good quality crystal and have good supply arrangements. People sometimes ask about (or brag about) GPS or atomic clock locking of audio oscillators; this improves something we don't need (long term stability) and degrades something we do need (short term stability).
Use of an external oscillator like you're describing allows for a nice low noise power supply. The SAA7220 is an electrically very noisy chip pulling about 1W of power, so its power supply pins are very hard to keep clean.
I just noticed the andrea_mori topical thread here:
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator
It would be interesting to compare the performance of a andrea_mori's design to that of one based on the aforementioned TI driver
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator
It would be interesting to compare the performance of a andrea_mori's design to that of one based on the aforementioned TI driver
There is an outstanding thread at "the other side" (as I call it...)
DIYHiFi.org • View topic - To clock or not to clock, that is the question...
I suggest reading every single post because Jocko's posts, in particular, should be turned into some sort of white paper on the topic of crystals and accompanying oscillators jitter/phase noise issues. Very helpful. He also referred to Andrea Mori's design... in a typical straight-to-the-point; no sugar-coating way, that Jocko is famous for.
DIYHiFi.org • View topic - To clock or not to clock, that is the question...
I suggest reading every single post because Jocko's posts, in particular, should be turned into some sort of white paper on the topic of crystals and accompanying oscillators jitter/phase noise issues. Very helpful. He also referred to Andrea Mori's design... in a typical straight-to-the-point; no sugar-coating way, that Jocko is famous for.
I respect JH's experience, but I really don't have time/patience to read thru all those pages. Similarly, I really don't have time/patience to read thru all those the pages in andreamori's threads in DIYA. These threads are simply too long.
If someone want to summarize for me, here in this thread, please do so.
I was a bit surprised to learn the $$ of the Laptech crystals suggested in Andrea's project!! E.g.: 11.2896 MHz fundamental -- Eur 27.00 each. Are they REALLY that good?!! I haven't come across a post that compares Laptech vs. generic $2.00USD -- please inform me otherwise!
... this may be one of the key posts:
DIYHiFi.org • View topic - To clock or not to clock, that is the question...
Good crystals are expensive. 27 euros is not expensive for a good crystal. There was a time when all crystals cost that sort of money, and all crystals were reasonably good. Then someone somewhere (Far East?) discovered a way to quickly and cheaply make bad crystals, which were nonetheless good enough for clocking CPUs etc. Now people expect all crystals to be cheap.hollowman said:
And because the Far East or mainly most of the time Far West cost killers, does a 1988 Philips embeded Crystal better than a brand new one in the goal just to give a second life to the CD player ?
Do all these fragile Xtals support the DIY soldering operations (temperature) ?
And at the end does it make sense to have an Xtal far from the dac chip with discontinued ground boards, massive ground loops, mixed current cross due to the tiny DIY board, unmatched impedance signal traces cause the lack of space for proper drawong or stacked antennas called hats ?
Some have listened to a difference when grounding the old big Crystal can to the close ground and putt some blue Tack for proofing behavior enhancement...
Do all these fragile Xtals support the DIY soldering operations (temperature) ?
And at the end does it make sense to have an Xtal far from the dac chip with discontinued ground boards, massive ground loops, mixed current cross due to the tiny DIY board, unmatched impedance signal traces cause the lack of space for proper drawong or stacked antennas called hats ?
Some have listened to a difference when grounding the old big Crystal can to the close ground and putt some blue Tack for proofing behavior enhancement...
Some DIYers and audiophile-gear manufs swear by the quality of crystal. Hence, all the interest in Andrea's Laptech-based projects.
You can compromise a bit: use a socket. Makes 'crystal rolling' a real joy for bored DIYers 😉
Outboard clocks are not uncommon in high-performance pro gear and super-high-end audiophile stuff (e.g., dCS).
In all my experiments with 'outboard clocks', including crude breadboard projects, I've noticed sonic improvements.
All that said, trace length is important--hence my interest in the compact TI driver.
I use the tack/putty myself. I have seen proper damping jackets for xtals in some audiophile gear.
I never tried this, but I've seen some design based on NJU6368. I guess you may be interested.
There may be good technical reasons why a rack full of digital audio stuff needs to have one single clock. "Super-high-end audiophile" is more likely to be for marketing or fashion reasons.hollowman said:
Expectation bias is powerful, even when you know it is happening. Changing sound is easy; improving it is difficult; the more effort we have put into a change the more likely we are to perceive it as an improvement - especially where no actual measurements are involved.
http://www.njr.com/semicon/PDF/NJU6368series_E.pdf
The NJU6368 is an interesting part. It's "Oscillation Capacitors" Cg and Cg and load resistor are on-die. With the TI driver, these passives have to be added by user.
The TI driver's datasheet has an original copyright date of 2004. So it has been around a while.
Unfortunately, both the NJR and TI parts are SOT23 size. Too small to work with w/o an adapter (or custom PCB, if you want to go that far!).
External passives are likely to be better quality than on-chip devices. The design aim of that device is small size and convenience. This is implied by the fact that the datasheet makes no mention at all of oscillator quality (e.g. stability, noise, jitter).
- Status
- Not open for further replies.