33.8688 MHz Oscillator: Discrete Choice

MOTIVATION

If your are looking for a 33.8688 MHz oscillator, you are pretty much out of luck. I looked at the usual shops and they are not in stock. Such frequencies were used for CD players and they are  not widely manufactured anymore. Modern disc players derive all the necessary frequencies (including the 33.8633 MHz frequency) out of a 27 MHz oscillator or crystal and an integrated multi-frequency clock generator.

I have a couple of Denon SACD/DVD players and wanted to see if the clock can be improved. On the DVD2910, The audio section of the player derives the 33.8688 MHz frequency from a 27 MHz crystal, generated by a SM8701 clock generator. The specification of the clock generator indicates that the output jitter for the 33.8688 MHz frequency clock is 150 psec. On the DVD1920, the audio section of the players uses a 33.8688 MHz clock derived from the SM8707H clock generator. To my surprise, this lower end model jitter specification is better than the higher end model DVD2910 (and also the DVD3910 and top of the line at that time DVD5910). The DVD1920 is a newer generation device and thus uses newer parts. The SM8707H part has a jitter specification of 70 psec (2X better).

The published jitter figure is “period jitter” and in order to compare it with a standalone oscillator one needs to compare the phase-noise plots. For example, the popular Crystek C33xx oscillators are specified at around 1 psec Jitter RMS if measured at the 12kHz~80MHz phase noise band (this is the accepted norm for telecom applications). However, if one uses the “audio band” which starts at 10 Hz, the jitter measurement results in around 30 psec. Under similar measurements the top-of-the-line CCHD-950 measures 0.5 psec. I can’t find any reference as to how the jitter numbers were measured for the clock generators, but if one assumes that they have been measured within the “audio band”, then these numbers are not too bad.

Update: I figured a good rule of thumb when comparing period jitter (what is specified from chips) and phase jitter (what is specified from oscillators): period jitter ~ 10 X phase jitter [link]. Thus 150 psec of period jitter is the equivalent of 15 psec of phase jitter.

THE “TRADITIONAL” MOD

The traditional mod recommends replacing the crystal that feeds the clock generator. This is typically 27 MHz crystal connected to the clock generator. As can be seen from the above analysis, replacing this crystal with a low jitter oscillator will do nothing to lower the jitter of the generated clocks because of their inherent jitter. Thus if reducing jitter is the goal, the clocks to replace are the generated ones out of the clock generator chip.

I wanted to explore the option of discarding the derived clock and using a separately generated 33.8688 MHz frequency. Note: the derived 33.8688 MHz clock is synchronous with the 27 Mhz master clock which is used elsewhere. At this point, it is not proven (or dis-proven) that a asynchronous 33.8688 MHz clock would work in the device. I cannot find any report on the success or failure of this proposed mod.

CHOICES

There are only a few possible (reasonably priced) candidates for a 33.8688 MHz oscillator:

1- The eBay solution (around $30). There is a review in diyaudio on these oscillators:

 

2- The famous Tent Labs oscillator (available in many frequencies including 33.8688 Mhz), >$30:

The Tent is the only one that is backed by a phase noise plot (just happen it is the phase noise plot of the 33.8688 frequency):

If we calculate the RMS jitter from the plot at 10 Hz forward (which is what Crystek uses for their measurements), we get: ~0.32 psec RMS (see sidebar for link to jitter calculator). This is an impressive number! Equivalent to the Jitter figure for the Crystek  CCHD-957 oscillators. Note the figure from the chart gives 0.9589 psec jitter RMS because it is measured from 1 Hz.

3- The next option is the Silicon Labs programmable oscillator Si570. These are also non-existent in the usual parts distributors, but you can find them from Ham Radio web sites, in particular KM5H.com (~$20) and SDR-Kits (~$23).

Si570 photo from here: [link]

The Si570 is a favorite of Ham Radio aficionados because it allows you to “dial-in” any frequency. There is a ($40) diy kit that lets you do just that:

And it is amateur radio people like it because of its low phase noise. Clifton Labs has done an extensive study on phase noise for the Si570 and other oscillators. These guys are even more serious about jitter than audio people 🙂

The only potential “disadvantage” of this oscillator is that it requires a local uP to program the oscillator every time it is powered on.

4- DIY discrete oscillator from the diyaudio group buy (~$25). This is a differential Colpitts design with dual shunt regulators.

In terms of performance (noise), the diy discrete oscillator is probably  not the best, but also not the worse. Although the designer of the oscillator has not measured its phase noise performance, there is some measurement on phase noise for a the discrete Colpitts oscillator from Clifton Labs:

My home brew Colpitts oscillator, designed without paying particular attention to low phase noise, has slightly lower phase noise than the Raco oscillator module, most noticeable when viewed with 1 and 10 KHz spans.

The designer further says:

Unfortunately I don’t have the equipment necessary to produce a quotable figure for jitter. I have taken both objective and subjective measurements (comparisons of the jitter at a CD player’s output and listening tests) that indicate that it has significantly lower jitter compared to a stock CD player master clock.

Further, there is also the fun factor in building your own oscillator.

DECISIONS, DECISIONS

Seems the best choice for this mod, based on jitter specification is the Tent clock. However, I have already bought the diyaudio discrete clock kit and whether it is better or not than the published 70 or 150 psec jitter of the 33.8688 MHz clock remains to be seen. I would guess it is probably better than a derived clock, so worthwhile to implement.

INSIDE A CAN OSCILLATOR, CAN CRYSTAL

Inside a can crystal

FURTHER READING

• Introduction to Quartz Frequency Standards [link], pdf: [link]
• HP oven controlled oscillator: http://www.realhamradio.com/GPS-oven-journey.htm.
• HP 10811A/B Quartz Crystal Oscillator Manual: http://www.hparchive.com/Manuals/HP-10811AB-Manual.pdf
• Discussion on clock selection: http://www.diyaudio.com/forums/digital-source/194965-clock-selection-experts.html
• This is how Ham Radio people build their oscillators [link]

Such type of design have been found to provide good noise figures as described by this site: [link]