Discrete Low Jitter Clock GB
Current status (February 2014):
I've now finished a new version, take a look here. There won't be any PCBs or kits until mid to late March, check back then.
Andrea Mori was also kind enough to make a proper phase noise measurement of the clock, it can be seen here, the phase noise is -108dBc/Hz at 10Hz, and the jitter is 0.26ps from 10Hz onwards.
I've just finished testing and enjoying a design I've been working on for some time; a low jitter clock. It's the second clock I've made, this one builds on the first (as well as learning from its mistakes). This clock has the following features:
- High quality oscillator circuit. It uses a differential Colpitt's oscillator, basically two Colpitt's oscillators running across a common crystal that operate out of phase with each other. This gives better power supply rejection than a single ended oscillator and means the clock doesn't need to be adjusted to have a 50% duty cycle output wave.
- High quality power supply. The power supply is both series and shunt regulated. The overall structure is a Schottky diode rectifier, CRC filter, LD1085V based series regulator and two TL431 based shunt regulators (one for the oscillator, one for the comparator).
- Power supply options. The PCB caters for two input power options; 220 - 240VAC or 7 - 20VDC. Different parts are populated depending on which option you chose. If there's demand I will alter the PCB to allow it to run from 110 - 120VAC supply as well.
- Small size. This was a big issue for me. The main failing of my first clock was its size, it worked well but was too big to actually fit into any CD players. This one is small enough to fit into all but the most jam packed players. It could have been smaller had I used surface mount parts, but I wanted to keep it easy to build.
- Low cost. I wanted to be able to afford to fit it to all of my favourite players, and there's quite a few of them.
Anyway, I wanted to gauge if there was any interest in a group buy. I was thinking of offering the following options:
- Bare PCB, about USD3.00 each
- Kit for 7 - 20VDC powered version, about USD25.00 each
- Kit for 220 - 240VAC powered version, about USD35 each
The cost of shipping around the world will vary, but should be in the USD5.00 to USD10.00 range.
If you're interested, please post how many and what type of PCB or kit you'd be keen on, and whether a 110 - 120VAC version would interest you.
I've posted up the schematic, board artwork and bill of materials (order numbers are for Farnell / Element 14) and some photo, please comment if you see any error I've made or improvements I could make.
2 x Kit for 7 - 20VDC powered version, about USD25.00 each
1 x Kit for 220 - 240VAC powered version, about USD35 each
all to suit Marantz/Philips players.
Thanks for your support stoolpigeon.
A few more things about the clock:
- The 7 - 20VDC powered version of the clock consumes about 40mA, so can easily be powered from a player's existing supplies in most cases. This current draw is fairly independent of input voltage due to the shunt regulators. The 220 - 240VAC powered version draws about a third of a watt.
- The oscillator circuit as shown is very versatile, and will work with a large range of crystal frequencies. I've successfully tested it with crystals ranging from 4.2338MHZ to 33.8688MHz, so it suits most of what you'll find in CD players. The only value I would still like to test is the 45.xxxx MHz crystal that is used in some modern Sony players.
- In the kits I can supply 11.2896MHz, 16.9344MHz or 33.8688MHz crystals. The 11.2896MHz are nice Citizen units, but the other values are generics. I recommend sourcing better crystals than generics, but my usual source doesn't have any.
- The size of the clock is 100 x 32mm for the 220 - 240VAC powered version, 75 x 32mm for the 7 - 20VDC powered version. This is in the board artwork PDF, but I thought it best to put it here as well.
- I've taken some traces of the output of the clock at various frequencies. It's getting somewhat rounded at 33.8688MHz, but the transitions are still steep which is what matters most. Please find these attached, the 'scope used is a LeCroy WaveSurfer 454.
- As built in the bill of materials I posted, the output wave of the clock is about 4V peak to peak, and is designed for 5V TTL or 5V tolerant CMOS levels. If you want to use it with ICs that only accept 3.3V signals, you'll need to reduce the amplitude of the output. This is best done by reducing the supply to the comparator. Replacing R205, currently 3.3kR, with a 2.2kR resistor will set the comparator supply to 4.2V and the output to about 3.2V peak to peak.
have you considered buffered clock outputs - incl inverted output
what type are citizen crystals that you offer - spec ?
Some background reading perhaps.
@samoloko: They're fairly standard AT cut quartz crystals. You can see the spec sheet here. There are probably better options out there, but availability in the frequencies you want is an issue, the frequencies CD players use are often not stock items.
The comparator at the output provides enought drive to not need further buffering as far as I could see, are you meaning multiple separately buffered outputs? That would be useful, but I've found I don't often need this feature, so I left it out to keep the board as small as possible.
Why would an inverted output be useful, are you looking at driving differential input logic? The comparator I used does have an inverted output (the PCB leaves it unconnected), it could be used if you wanted.
@EVUL: Yep, that's where I got the basic topology from, it's a worthwhile thread to read. I read almost every thread on clocks when I researched making this one. I tweaked that circuit a bit, added a waveshaper and power supplies, made a bill of materials, a layout and here we are.
may i ask few questions about it
1. how does it work? does the spdif signal come through it first then go out the dac?
2. does it only work on cd player or i can use it on all spdif device?
pardon me for my stupid questions, because i only have little knowledge about audio
thanks so much,
@bobkojek: No problem, I'm happy to answer questions. This clock replaces the master clock on a digital device. It provides a more precise clock to the digital filter and DAC ICs in the device so that the device's output is more stable and precise in the time domain. This will result in a very significant increase in sound quality in most devices.
It's useful in any device that has a master clock, such as:
- CD Players
- CD Transports
- DACs using some form of reclocking
Could you point out the datasheets for 11.2896/16.9344/33.8688MHz crystals, thank you!
@aparatusonitus: The 11.2896MHz Citizen crystals I used are these ones. Unfortunately Farnell doesn't stock brand name units in 16.9344MHz and 33.8688MHz frequencies, and I've been using generics.
I've just been looking, and I see that Active Components stocks all three (11, 16 and 33) frequencies in the Citizen HC49, I should be able to source them from there.
|All times are GMT. The time now is 04:32 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio