Inveterate meddling #101: playing inside the Marantz CD63
Having spent far too long playing with the guts of an old CD63SE we finally achieved something useful - so here it is for public consumption.
Recipe: Click the image left for diagrams showing the underside of the PCB before and after modification. |
The clock in the '63 is a simple Pierce-style oscillator implemented directly onboard the DAC chip using a dedicated logic gate. The relevant components are the NPC DAC (soldered to the underside of the board), CD02, CD03 and the 16.9MHz crystal, XD01. Having the clock built into the DAC is potentially ideal; but the clock in the 63 out of the box is known to be pretty jittery, obviously a product of mass-production in its execution, and those who've tried reclocking devices report tremendous gains, as you might expect after spending all that money.....
After much hacking about with an old 63SE we think we've found a good reason why. Go read the datasheet now and come back here afterward.
The crystal is connected across one end of the DAC chip, between pins 1 and 28. The 10pf caps CD02 & CD03 are tied to the ground plane via a PCB trace, basically forming a series resonant circuit with the crystal excited by the logic gate between pins 1 and 28 through RD02 (which you should not modify). The logic gate in the DAC even has separate digital supply and ground pins to decouple power supply noise; pin 2 is digital ground and pin 27 the +5v supply, fed through RD01 and decoupled by CD04 (electrolytic) and CD05 (tiny axial ceramic). So far, potentially outstanding butas laid out on the board:
Pin 2 (logic ground) is connected to the ground plane from the _far_ side of CD02 & CD03, so every time the logic gate changes state, the voltage change across the trace inductance will directly modulate the series resonant circuit, because CD02 and CD03 are no longer at quite the same potential. ... no wonder there's so much chaos in there.....The cure is to add one new capacitor and cut one track.
Our new capacitor is fitted directly across pins 2 and 27, as tight to the chip as possible, since low inductance is critical; stacked foil poly types work for me, X7R ceramics are another good option. You can also mount it by mirroring CD05 (which BTW is too small in value for the job it's required to do) below the board; the critical thing is keeping the lead lengths under 2mm from body of the DAC for the cap to remain effective into the VHF range. Since the DAC is a surface mount device the best way to achieve this is to pre-bend the leads until the cap just fits across the DAC pins, and clip them until the new capacitor sits tight against the top of the chip - when the board is reinstalled, you'll only have about 1/8" clearance from the case for this cap to fit into! Pre-tin the cap leads with solder off board, then holding the cap in place, when you're sure you're in the right place, momentarily touch the cap leads with your soldering iron and the cap will be soldered in place.
To get rid of the ground bounce problem locate the ground trace from Pin 2 and follow back around to the ground plane under the DAC. Take out the scalpel and carefully cut the trace between pin 2 and CD02 solder pad, as close to CD02 as you dare, and again as close to Pin 2 as you can, isolating the trace so it doesn't become an aerial. Having cut the trace out, you're done since pin 2 is still connected to the ground plane by a minute trace below the dac.
While the boards out we also recommend swapping RD01 for a small axial choke and a couple of ferrite beads and increase CD04, as suggested in Thorsten's articles on the TNT-Audio site. That's all there is to it - once you've got the board out and can trace the circuit this really is a lot easier than it sounds - when inspiration struck it took just 20 minutes to do this tweak - but then it took two years for the inspiration to strike!.
Incidentally we think the 'ground bounce' issue is why 'superclock' upgrades work so well. If an external clock is used, the signal is fed into XTO (pin 28) and CD02, CD03 and XD01 are removed. Which means the internal logic gate is still used on the DAC, but it now has the ground trace to itself and so cannot interfere with the oscillator. Those with a clock upgrade should still try the additional cap.
more adventures:
Postscript: four more tweaks to try which enhance the above:
These three mods together are definitely worthwhile, esp. the new regulated supply and take the logic gate-based oscillator about as far as it can go. Here are images of a particularly neat installation based on an LM317AT - Andrew L. Weekes notes
It's actually quite neat, I did a piece of board, 5 tracks wide, fitted the resistors / tant's and configured it such that there is a ground leg at each edge, the o/p in the middle. The ground leads are then bent at right angles and soldered to the ground plane (scrape away some resist) then the middle leg passes through the hole vacated by the resistor). With a 317 I chose resistor values that eliminated the need for an o/p load resistor for min. current, could be worth adding this to the site? A 120R and 360R gives 5V exactly, and 10mA quiescent current. I took the 10V feed from a link that sits before the 5V logic reg.
the payoff
So what does all this do for the sound ? In a word liveliness; the modified player no longer sounds a bit listless or forced. To our ears the gains have been terrific, esp. the vastly improved P.R.a.T - this player can now really carry a tune; bass freaks might be pleasantly surprised also... there's less fat, but better extension. But you don't have to take our word for it, for Andy comments:
I cannot isolate the DAC gate PSU from the effect of adding the analogue PSU, but in conjunction they elevate a humble CD63 with a good low phase noise clock [from LCAudio - mc] to unbelievable levels. In [the] system there were no obvious shortfalls or criticisms, just a huge desire to listen to and enjoy music. The bass has tightened dramatically since the mods, and often from the first few notes played in a song it was obviously better in terms of character of an instrument or voice. The resolution of detail and sense of low-level ambience in a recording was considerably improved, treble was awesome, one could tell exactly how, where and what with a percussion instrument was being struck. My brothers system also produces stunning imaging, not because he's into that specifically, but it's a function of the Royd Doublets he uses, and one felt like it was possible to walk into the sound stage and almost touch the instrument - really amazing. Rythmically the dramatically improved bottom end gives a much greater sense of timing and stop / start to a note, portraying it's shape and character, particularly when things get busy. In fact things never really sound busy, as everything is clearly portrayed where previously it could get lost in the mix. The most often made comment was 'sounds like a different CD' or sounds like a totally different bass guitar / guitar / drum set etc. Recordings that were previously impressive for example because of very low bass, but that lacked character to that bass came alive, with a real sense of the construction / string type etc, etc. Just much more interesting music, everything sounded like better musicians had been drafted in to play. I'm sure the CD63 as it now stands would piss over a standard entry level Naim CD player in every conceivable way. My second CD3.5 is just ahead owing to a clock mod and a good external PSU - not the most cost effective method I feel!
But there is plenty more to come from the CD63. Not content to leave well enough alone, we went so far as to do this to the longsuffering thing...
Have fun - don't attempt this with a chisel after the pub...
© the twisted pair 2000