DUAL DAC board set upgrade project for
Exposure CD player by John Farlowe
Direct clock taken to extremes
First of all for this and some parent projects I must express my gratitude for VERY much hobby participation and fundamental research, set up and a wide variety of tests. What you read here and can hear from the resulting equipment wouldn't be possible without my brother's support.
For decades he is an enthusiastic listener, (co-) home constructor and also took the realisation of many of my development thoughts in his hands. Particularly hours and hours of library editing in my CAD program EAGLE and the transformation of my provided schematic drafts to real, accurately routed boards cannot be appreciated high enough here, beyond the direct clock development this project meanwhile includes also countless schematic extensions he designed basically - especially regarding digital and supply aspects.
So the dual mono logic used here was not my idea, but was re-developed by him using the example of some Micromega-, Rotel- and Harman dual DAC logic schematics, mind the sheer inquiry effort, how something like this can be "easily" realised. The first application of this logic was in a Philips CD751-Player, that forced the utmost out of two TDA1549 DACs from a self- and a donor- transplantation by utilising a completely self designed dual DAC board with not less than eleven (11!) voltage regulators.
Regardless this player may be ideal for connoisseurs use, but not really ready for production in the absence of any muting device, it already reached such a musically high base standard, that I am staggered myself looking - excuse me - ...listening to these results, that seemed for me impossible not so long ago.
However, for the various Symphonic/Blues-Line- and Exposure- modifications in autumn 2014 all clock module boards were used up, in parallel the "twin carburettor"-Philips mentioned above developed, the internal discussion regarding the Exposure CD player went to the more curious direction then, how the acquired perception would influence that Farlowe masterpiece, applied matching the given design style. Being an outstanding peace of gear already in its unchanged condition, with the direct clock extension that player had gained naturalness and outline, but one thing remained obvious: against theory one could still distinguish a pressed original CD from its burned CD-R copy - evidence for the drive to send the DAC more interference at low tracking contrast, than its only in the original mould unavoidable jitter.
The most obvious ways of crossfeed were power supply and ground coupling, the corrective against them was already tested and proved with the Philips CD751 project.
Applying the already existing logic board, thus my brother made a two board project from it, that also had to meet the Farlowe style in this case - not to taint the sonic design and character. Some respect is necessary...
the project is dedicated to the first designer, writing the name Exposure on the board was avoided for legal reasons
The logic board already combined the step up PLL and its slightly higher radiation with the dual mono logic for mounting it close to the drive control. That also has the advantage, that between the converter board and the digital part only the lower frequency master clock is on the long cable connect, which finally produces the necessary phase locked double or quadruple drive clock very close to the drive control chip exactly in the position of the original clock generator and can be injected on the shortest possible way to the ICs XO input.
above: layout and
below: implementation, a sample board of the predecessor version is added in the top of the picture
details of the logic board: the bigger ICs are shift registers for the dual mono conversion, the small IC (left board, left lower corner) is the PLL, the clock detection is placed on the front/ground side (right board, bottom left)
The latest innovation on the logic board now is a clock detection, hence from experience the master clock during switch on starts relatively slow because of its distinct, low noise decoupling - if the drive control initialises earlier, it remains without clock which leads to some choked mode. The chip then spits asynchronous data making audible noise.
On the other hand this IC also has a reset pin, which can be held down until at the oscillator input really gets the (external) clock signal. If the control is started with the external clock, it doesn't recognise its own crystal is missing and well behaved sends valid data to the converter/converters (that there are two now, it also doesn't even know...).
For a low interference conversion apart from the drive electronics the approach to the core of the Exposure player followed.
Until now the tiny "constant calibration" stereo converter TDA1545 was situated on the drive board, fed by its sources and connected to its grounding. A long shielded cable then directed the ground connection and both current outputs of the converter for about 25cm to the proprietary Exposure current to voltage conversion. That was designed very carefully, on the one hand side to supply it properly in the Exposure style (two LM317/337 for each channel with the traditional 10µF electrolytics at in- and output), all together fed by a considerable toroidal transformer and the proprietary slit foil electrolytic cans. On the other hand the whole current conversion and output stage is held as short and compact as ever possible, basically an adapted line stage from the top class preamps, bended from voltage- to current control. And already blessed with all sonic benefits of the Farlowe electronics.
For a general improvement there was nothing to do beside removing wear and some impedance matching, if anyway it still should be an Exposure afterwards. The options left: shorting paths, implying less interference during conversion and signal current transport, establishing smoother supply conditions and mainly: supplying the converters structural component correctly from the analog supply and pin it to analog ground, while the drive with all its servo convulsions is completely separated from the signal -supply and -ground. This time not with a mighty phalanx of passive buffering as in the own Philips projects, but exactly supplied as the Exposure output stages also are. That train of thought my brother translated to a twin converter board, which
- communicates data and clock signals with the drive via a galvanic isolation device,
- places an highly precise clock generator with a Tentlab XO and extremely low noise supply directly beside the two converters,
- feeds both its stereo converter chips from an own LM317 voltage regulator trough decoupling, supporting filters,
- provides an optocoupler separation as mute relay control and
- has outputs and screw joints exactly above the corresponding connections of the Exposure board.
and the realisation
the bottom side is also fit with components, obvious: the separation of analog and digital ground
...but wrong applied for the first attempt
When finally all core ingredients were gathered for my own player, one Monday morning I started assembling two board sets. Countless times I re-ordered smallest materials the next days, very soon I also understood I had to put on a finer soldering tip: a whole lot of SMD parts, especially resistors, in this design had an overall length of no more than just a millimetre, for manual mounting rather something for juvenile, sharp eyes and children's hands - I had to get used to this a little bit. Two days later the boards were about finished and allowed the first alignment procedures, now for both clock and converters there were exactly 5.00V available. As it is always, at many points after the first measurements re-calculation and re-design is necessary, also for the documentation - it was almost weekend, when the first setting up operation impended.
As I could not sort out where the proper screw joints provided for my case were, after the drive board revision I tried mounting the logic below it, pictures from the working Philips CD751 showed a similar position. The function of the new designed clock detection was still unexplored. Initially I connected it with a small additional circuitry at the input of the drive logic voltage regulator, so the whole supply was only started with the clock. That worked immediately. Also immediately the synchronous multiplied converter clock was available exactly where I wanted it.
The central drive control. At the SAA7345 the clock crystal was removed, also the converter (red spot) is disassembled
But the drive didn't work.
Either it tried to focus after getting clock signal and then gave up with "error", or it spinned up the disc motor at full throttle, that wasn't it yet - but why? With the similar constructed Philips it worked!?
What I was not aware of: the screw holes of the old and new logic boards were completely different...
Standing so far in virgin soil it is always difficult to get first reference, how the problem is basically structured, it could have been a mistake during the refurbishing of the board as well as a defective laser unit or a complete conception error.
After several failure attempts I began with taking another drive out of a for its bright display "cannibalised" CD733 and then disassembling its board Exposure-typical, refurbishing and testing it. In spite of a identical layout in difference to the former board this one was two-sided with an extensive ground plane on the top layer. This was obviously quite an essential difference, for now the drive initialised, fast track changes also worked - still I had fierce signal drop outs and jumping during playback.
Time consuming mounting efforts - left: bad logic position, right: complete switch on by clock detection
failure attempt: obviously the grey flat cable starts exactly beyond the laser unit connector - that way it didn't work
Which then brought up the proper idea: here it is the logic board that disturbs the signal path from the laser photo diodes to the pre-amplifier, for its output was situated exactly over the flex cable connector of the drive. The shielding of the board has reduced the problem. After re-mounting of the logic board to the primarily intended position (which I had not taken into account) everything was rather more easy to connect at the edge of the board, in addition several centimetres clearance came between the new electronic and the sensitive lowest level signals.
During the board exchange I also connected the clock detection directly and short to the reset pin of the drive control chip, assembled that way everything started immediately as intended - and the read errors had vanished completely.
There were some minor subsequent works, the correction of a right-left inversion at the digital signal lines was still necessary, but now it became obvious, that with a slightly different assembly everything would have worked from the very start - which it now did. The new clock detection: perfect function, new opto-coupled relay control - immediate usability, adjustment of the output boards input to the double converter current - exactly the intended, accustomed level at the output. Now there was only one agenda item to be checked: the ground relations analog - digital - enclosure.
The starting point (digital ground connected to enclosure as before, analog ground now "lifted" trough 100nF/4,3MOhm) seemed sonically inappropriate to me. Meanwhile analog ground is connected to the enclosure (which has no protective earth connection) and the digital part is "lifted", this way round it sounds better.
correct application: instead of 4mm before now 4cm space between the critical leads - next to the motor to the left you see the flat cable connector to the laser unit and below the laser receiver preamp IC.
In the core everything at best
Now that the drive sends flawless master clock synchronous dual mono data, a glance on the new converter board. According to all the already existing connections of the output board below it provides for all critical connections exactly perpendicularly above positioned contacts, all together there are the following transfer points:
- 2x analog ground (one of them for the "lifted" ground coupling)
- converter signal current right/left
- converter reverence voltage right/left
- relay ground/open collector control line
- digital ground/supply decoupled
- mute input, decoupled
- word clock input, decoupled
- L+L/R+R digital signal input, decoupled
- master clock output, decoupled
Many of these connection are directly drawn downwards using solid silver coated copper wire.
twin carburettor in double-deck-technology
For the connection to the drive all switching-, ground- and data lines were combined in a soldered flat ribbon cable, even the supplementary right-left-exchange is visible at the connector. Below the low jitter Tentlab-XO between socket and metal case a Sorbothane pad is inserted for damping reasons. The decoupled master clock leaves the board in a pigtail WLAN antenna cable. The green LED is part of the low noise parallel stabilisation in the clock supply.
...below the revised I/U output board with its fat supply
Apart from a wear eliminating set of new Panasonic electrolytics (of course with the original values) and the adaptation of two resistors per channel the output board is totally unmodified. With the dual mono principle applied, no more half a stereo converter works for each channel, but a complete one, whose outputs are directly paralleled. Together now the double number of current sources is working and delivering at any time the double current. Accordingly also the parallel (bias) current source and the conversion resistor has to be adopted to the double current - 220 Ohm (current source) become 110 Ohm and 2,00 kOhm become 1,00 kOhm (I/U conversion). Thus the former quiescent point (and the resulting voltage signal level) in dual mono mode is kept exactly as it was with only one TDA1545.
If now the signal comes out with exactly the same level, what is it all good for?
Concerning the dual mode first of all one gets a far better decoupling in any direction - the converter chips amongst each other (in this case also between the two channel signals), from the drive and generally from disadvantageous common current feed- and delivery- leads. Then of course with two converter outputs per channel the levels sum up to the double (=+6dB), while all converter errors and noise components combine rather geometrical, which is only 3dB. With the -6dB-adoption (together with all noise signals) that leads to an increased signal to noise ratio of the current input, below the line you get a little more freedom from interference of about 3dB.
Further absolutely not trivial are the sophisticated direct timing (as mentioned in the direct clock article) and the short connections to the output board.
And does it pay off?
In the very week of progress it lasted until Saturday, before I was able the first time to have an ear for the player. But because of such a lot stressed concentration before, no significant test was possible with that absolutely not attuned new device. One week of continuous operation later this had changed completely, by that time I also had the lent double TDA1549 player of my brother (which had ordinary drive problems) really fit again for comparison.
It plays as "Exposure" as never before, exactly the tendency for "outing" itself further as a peace of Farlowe electronic, but with less digital artefacts and even more tighter and clutching, than already the first direct clock version managed it. It is definitely demonstrated, that the converter uses no bitstream/delta sigma technology at all, but a parallel current source concept just slightly different from the traditional R2R-technology thus being audible far more friendly to low level signals than actual designs are. And who ever was fascinated by a Exposure XIV preamplifier knows immediately, what I appreciate about this player. For primarily it reduces the playback on discrete analogue aspects and keeps out of the usual CD player discussion - it does its job far more like a preamplifier, well recorded materials will not disturbed the listener with unnecessary occupation by digital aspects.
One more listening session later it had played itself into the hearts as an absolute complementary match for a Musical Fidelity A200, at best resolution compared to the rivals CD751 and PAN it showed particularly more "trace" combined with the ability to draw musicians and instruments as seizable objects, not to dissolve or disrupt anything spacial and also no melody line, but to represent events well proportioned and coherent.
In respect of effort this player is (disregarding self inflicted loop ways) one size smaller than the CD751 reconstruction, but if you ask me, which one of these two excellent players I would choose for myself in the context of the testing environments, the points win would be the Exposure's with its lean, logic line management and the for my ears a little bit cleaner converters.
The answer to the question "does it pay off" in my opinion for the customer (for someone who orders the revision or buys one of my revised players) is: more than worth it!
For the calculation above already subtracts what I paid dearly - you have the opportunity to get into a development line, that hardly has to do anything with actual arbitrary hifi, the result is some hobby enthusiastic, musical equipment on an excellent base - taken to extremes and absolutely fearless compared to REALLY expensive gear. Well, listening by oneself is always the best - but after all what I heard until now this is "very big value" anyway - just not yet for me, I must still think about some details, how the following examples can be mounted quicker by far. An obvious possibility from version 2.0 on: to let assemble rather than assembling myself.