1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
  2. Where we are now: The site is back up and running! The character set encoding issue has been fixed so old posts should now look ok. I am currently working on themes/skins etc and I will be deleting the temporary green skin very soon as it has many issues. I will do a very quick variation on the stock blue skin first just to get a light-grey background and san-serif font, and I will set this as default. Later this week I will hopefully add a professional third-party skin in pfm colours (life is way too short to do this crap myself, and I've found one I really like than needs next to no tweaking). We are getting there slowly...
    Dismiss Notice
  3. May I please request people check and update their email address if it is out if date. I ask as I’m getting countless bounce errors as the server fails to send notifications through etc. I’ll eventually figure out how to send these to a black hole somewhere, but it makes sense to address it at source as quite a few folk are obviously not getting the thread and conversations notifications they have set in their preferences.
    Dismiss Notice
  4. Things you need to know about the new ‘Conversations’ PM system:

    a) DO NOT REPLY TO THE NOTIFICATION EMAIL! I get them, not the intended recipient. I get a lot of them and I do not want them! It is just a notification, log into the site and reply from there.

    b) To delete old conversations use the ‘Leave conversation’ option. This is just delete by another name.
    Dismiss Notice

CDX reverse engineering

Discussion in 'd.i.y.' started by teddy_pardo, May 16, 2006.

  1. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    In the past two weeks I’ve been spending time modifying my CDX. I have achieved great results so far and will post the modifications on a separate thread as soon as I’m done. In the meantime I thought I’d share some of the reverse engineering work that I’ve done on the CDX.

    The power supply
    When used with an XPS, the CDX is powered by 4 power rails

    +/- 22V for the six OP42 opamps
    +/- 10V for the two DACs (one per channel)
    Two +15V for the digital and mechanics

    When powered by the internal PSU, the 22V and the 10V are both connected to a common +/- 20.5V rail, and the two 15V are connected to a common 13V rail.

    The +/- 22V is internally regulated (317/337) to +/- 18V, then connected directly to the 6 OP42. Each OP42 gets a couple of 100nF SMT decoupling capacitors. These regulators are located at the back left of the PCB, near the DIN connector (these are the first candidates to be replaced by ALWSR, which is what I did)

    The +/- 10V is internally regulated to +/- 5V to power the two DACs. For each DAC there are two 337 (one for the digital part and one for the analogue part). There is only one 317 for both the digital and the analogue (I don’t understand why). In total 6 regulators located on the left side of the PCB. The negative analogue and the common positive will probably benefit from being replaced by ALWSRs.

    The two 15V rails are first regulated to 9.3V and then to 5V. There is one 9.3V regulator per 15V rail, and then many 5V regulators are connected to the output of the 9.3V regulators. Note that the 9.3V regulators carry a current of about 1A (one of them has a heat-sink). They are located at the front of the PCB near the transformer.

    The PDM100 (described below) gets two 317s, one for VDD1 and one for VDD2, as recommended by the datasheet. VDD2 is a candidate for ALWSR. These two regulators are located at the back of the PCB near the Burndys connector.

    The circuit

    Main components

    PCM1702 – DAC. Two are used, one per channel
    TDA7073 – Dual BTL power driver (for the servo). Two are used
    SAA7376 – servo processor and CD decoder
    74HC04D – hex inverter
    PMD100CQ – HDCD decoder and filter. Datasheet can be found here

    The crystal is connected to inverter #2. The output of inverter #2 goes to inverter #4, then to #5, then to #6. The output of inverter #4 (the first one) is connected via a 47R resistor to the PMD100, and the output of inverter #6 (the last one) is connected via a 330R resistor to the SAA7376 decoder.

    Crystal -> #2 ->#4 -> #5 -> #6 ->

    The PMD100 generates an internal clock which is used as an input clock for the two D/A converters (connected via two 10R resistors).

    Note that the unlike other Naim CD players, the D/A converters are not connected to the inverter but to the BCKO (bit clock output, pin 34) of the PMD100. The diagram is shown on page 19 of the PMD100 datasheet.

    One strange thing though is that the PMD100 datasheet says: “The PMD-100’s internally generated jitter on the bit clock output is lower when it is operated at 256 Fs when compared to 384 Fs (XTIM low versus high). When implementing the PMD-100 into a 384 Fs and bit clocked DAC design, we strongly recommend using re-clocking circuitry in order to attain the best possible sonic perfomance.”

    Although the CDX is running at 384 I have seen no re-clocking. So I guess that performance may be gained here (maybe it was done on the CDX2, anyone?).

    I have not received yet my XO module so I cannot comment about the improvements of the Flea to the CDX, but given that the clock is not used to feed the D/A converters, I suspect that the Flea would not be as beneficial to the CDX as it is to other players. Martin, what do you think?

    I’m very interested in what has been done in the CDX2, anyone willing to share pictures?
  2. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    A small correction. On one of the 15V rails there are two 9.3V regulators (those without the heatsink), that are three in total.
  3. bivalve

    bivalve pfm Member


    I can send you a couple of CDX2 internal shots, look out for my pm.

  4. zanash

    zanash pfm Member

    Any thoughts on the post dac and output stages......?

    I was loking at the cdx .......

    the obvious mods for me where the addition of a set of rca sockets...to allow connection to more modern gear ;-)

    The opamps are pretty good ....but are the tants in the signal path ?

    There are also a couple of resistor that are gold banded[1% or better ?] and may well be in need of swapping to bulk foils or an other.

    This cdx is not mine so I couldn't whip the pcb out to have a look at the pcb traces.

    If there are any caps in the signal output I'd be thinking of swapping them for film caps or polypropes.
  5. martin clark

    martin clark pinko bodger

    There are tants in teh signal path, but with the healthy voltage bias they have it's not trivial to sonically better them if you like what Naim gear does in the first place. FIlm caps do work, but the gains are not large. The OP42 opamps are good too - not worth messing with, probably larger gains to be had more easily elsewhere first. Both these things IMO of course.

    Teddy - have a look for the I/V stage following the dac (1702s?) swapping any I/V resistor in the opamp feedback loop for a Dale RN60 type will reap surprising rewards (smoothness and detail)
  6. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    I replaced them with OPA627 and the difference is very very big. The sound transformed completely and became much more open and detailed, also the bass is much better. I used to have a strong mushy mid bass, which was replaced by a tight firm bass. I must say that before changing the opamps I wasn't very happy with the CDX, now I changed my mind...

    Thanks for the tip, will do.

    Martin, I have a question. The output of the clock is going through three inverters as described in my post above. Can I use the Flea and do the acoustica mod, to connect directly to the 7376 and the PMD100?
    BTW, I received a notice from the post, I think that my XO has arrived, so I will install the Flea tomorrow :)

  7. martin clark

    martin clark pinko bodger

    Interesting what you say about the 627, it's the one opamp I'd make an exception for ;) When I wrote the above I was thinking (given the retail price of 627s) that it would be worth looking for 'easy wins' first. I still think the OP42 is a good opamp (much better than OP604 for instance) though; it likes a nice clean PSU, but don't they all...

    Any chance of some photos of the inverter layouts? I'll come up with a proposal (want to have a good read of datasheets first for any funky timing issues). That said, MartinC has fitted a flea bypassing the inverter*, and seems happy with it i.e. he's not been posting here much in the last couple of weeks!

    * see posts near current end of 'Got Fleas?' thread
  8. MartinC

    MartinC pfm Member

    Hi Martin.

    Currently enjoying a needed break in Hawaii, so no bodging at the moment - back at the soldering iron next week!

    The CDX / XPS modding threads have gotten pretty interesting recently, so should be interesting.

    I took a few internal pics of both sides of the CDX board that I can link to when I get back.


  9. martin clark

    martin clark pinko bodger

    Silly sod for looking in. Get back on that beach.
  10. teddy_pardo

    teddy_pardo Trade: Teddy Pardo


    Which threads are you refering to? I looked for information on the "got Flea" and didn't find much info on the CDX.

    I will post pictures tomorrow (my memory card reader is broken, I'll have to get another one)

  11. Adrian

    Adrian pfm Member

    Not the same, but I have just added a Tent VXCO with psu based on the flea to an external dac using the PMD100 and 1702s. There was a noticable performance gain.

    My dac uses 256Fs rather than 384Fs, so you may get the additional jitter noted on the PMD100 data sheet. The reason the 1702s aren't clocked directly from the clock is that the PMD operates with a stopped bit clock during work clock transitions. This is a waste of time for bit clocked dacs like the 1702, but means that to retain the sample integrity you need to use BCKO rather than the clock to clock the dacs.

    Assuming your crystal is connected to both sides of inverter #2, I would have thought this makes the oscillator #2 with #4 to square it up. So you could take the output of the Flea and inject it in place of the output of #4. This would retain the probagation delay of #5 & #6 if necessary.

    [thanks to this board for the flea design - though I better register to say thanks!]
  12. zanash

    zanash pfm Member

    what are the resistor values in the sig path .....so I can pass on the info and a few dales :)
  13. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    I may have said something stupid. The datasheet refers to bit converters, while I think that the PCM1702 in the CDX is a 20bit DAC. That may explain...

    The PCM1702 receives both the bit clock and the word clock, according to the PMD-100 datasheet the word clock is of good quality.

  14. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    After all, I was just following the Acoustica recommendadtions... BTW, I got the OPA627 on eBay for $3.3 each, not expensive :)

    Top side of the PCB (sorry for the bad quality, let me know if you need a better one)


    back side


    Cheers, Teedy
  15. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    One more thing, looking at the PCM1704 datasheet, BB recommend OPA627 for the I/V and OPA2134 for the post filter. What do you think?
  16. martin clark

    martin clark pinko bodger

    They would do - they make it!

    I've read very good things of using 627s for for I/V purposes (FET input, excellent bandwidth and slew rate etc - but so does the standard OP42) - and if you have some to hand I'd be very tempted while it's all in pieces.

    (I'll have a detailed lok at the pics tomorrow, many thanks)
  17. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    Some more reverse engineering of the CDX CD player. Here is how I think the analogue stage looks like:


    The 2n2 and 1n capacitors are Polystyrene, the others are SMT (Ceramic?), The output capacitor is the usual blue tantalum. All 1K Resistors are 1%, the others are 5%.

    I hope I did it the reverse engineering correctly, any feedback would be appreciated.

    Does the CD3.5 or other Naim players look the same?

    My plans are to replace all the resistors by Welwyn, and the SMT capacitors by Polystyren, any thought on that?

  18. trancera

    trancera pfm Member

    Well if it were me I'd try a single poly instead of the caps in parallel. Whenever I tried things in parallel in signal path I never liked it.
  19. teddy_pardo

    teddy_pardo Trade: Teddy Pardo

    Yup, it seems like cost was a factor here. I guess that 2n2 polystyrene + 16n SMT cost less than 18n polystyrene... That's probably also the reason for using a 6n8 SMT and 1n polystyrene...
  20. PigletsDad

    PigletsDad pfm Member

    The 6n8 and 1n are in different places in the circuit - not in parallel. This is something to do with response shaping - I guess the circuit is trying to be some magic roll-off shape - maybe Bessel for constant group delay?

    There is a total of 40dB/octave of eventual rolloff - 6dB from the initial inverting integrator, followed by two 18dB/octave stages.

Share This Page