Phillips CD304 MKII Restoration&Tubization

I know I’m late to the party. By 20 years or so. CD player modification and tube based upgrades was all the rage in the late 90’s and early 2000’s. But I just couldn’t resist the offer to work on Phillips CD304 so here it goes.

This CD player has quite the reputation for being an over-engineering marvel and shear weight of it tells you it’s really so. All aluminum chassis and CDM-1 CD transport screams about those times when things where build “because we can” and not “because it’s cheap”. It’s really more or less build like a tank and I’m sure it would survive a series from AK-47 just fine. At least on some angles.

There wasn’t a lot wrong with it and cosmetically it was in a decent shape. Just a few of annoying old age tantrums, like drawer needs help when closing, display plexiglass caught a bad case of glaucoma and read errors here and there. Nothing that couldn’t be cured by all-through re-cap, some much needed lubrication and a few new parts.

First things first. I decided to switch mains transformer from 220V winding to 230V as this is the nominal voltage now all across the EU. This lowers the heat output from all the voltage regulators quite significantly. Then CDM1 transport was disassembled and spindle together with laser pivot arm got a good amount of fresh lube. Motor board was probably oldest SMD job I have ever seen, so I instinctively re-flowed all components with a fresh solder. When assembling spindle motor I set the laser focus distance and adjusted laser voltage as per CD304 service manual. After close inspection, there was nothing wrong with all the mechanical settings of laser arm. Also factory locking-lacquer was still in place so I decided to not touch anything there. I’m not including any CDM1 adjustment write-up here, as it’s already has been done. Couldn’t do any better if I tried.

Unfortunately this improved things just partially. Now all CD’s and even CD-R’s could be read, but when putting new disc, player would randomly throw error and just refuse to play it. This very puzzling behavior would intensify after player was warmed enough, so it’s obviously something temperature related. After playing with this problem for a couple of days I was about to give up. But then I noticed something peculiar. If I would open and close the drawer, same disc would play just fine 80% of the time! So it’s not laser, it must be spindle motor related. It was time to download CDM1 service manual and dig into schematics.

After scrolling down to motor control section I had Aha! moment. Some technician wrote a hand note “motor not starting… (will start if you give it a push)”. Sounds very familiar, doesn’t it? And what do you know, after getting error message I pushed CD with finger and it started to spin! I also could manually spin CD and find a “dead spot” where CD wouldn’t budge when play button was pressed.

Above schematic is most primitive 2 phase brush-less motor controller used in CD304. Control signals for both phases are formed by using two hall sensors that are tracking permanent magnetic fields on the rotor. Half of LM324 op-amps are configured as inverting drivers with output transistors in the loop and other two are DC servos for appropriate phase lead-lag angles. So what did go wrong here? Rotor magnets lost some if their initial permeability? Hall sensors degraded? Probably a little bit of both.

Changing integration constant on those servos will increase starting and operational current, as well as motor torque. In other words, even if hall sensor is not yet active, phase current will start to push the motor. After installing suggested cap’s there where no starting hick-ups anymore. Motor current increase was quite moderate and windings just a little bit warmer to touch. Total victory was declared.

Next it was a time to re-cap two main boards. Quite a boring janitorial work. Especially with hand operated solder-suction pump. If I would do it more often, I would invest in automated vacuum de-soldering station. Something like Hakko FR-410 or JBC CV-2E would makes this job a walk in the park. Watch out for those “solder-blobs” on the ground plane (circled in red).  These are nutorious points of failure, even though there is some conformal coating on the board. Always reflow all them with fresh solder.

I bought a whole revision set from It’s 51 caps, new display glass and new belt. You can’t beat that for 39€, highly recommended. Their service manual is also great. Only thing that I would change in their set is main reservoir PSU caps right after bridge rectifiers. This is not a place for low-esr caps. Don’t get me wrong, these Panasonic FR caps are great and I use them in all my restoration projects. It’s just that original diodes will “ring” into that low impedance and make a lot of noise. Therefore they should also include fast “soft recovery” diodes or use general good quality +105C caps. Actually, service manual shows BYV26C as replacement diodes, but they are not included. Luckily I still had some in my stock.

Changing caps on a display board is very painful as two of them are hidden beneath the display itself. No other way to get in there except de-soldering whole 23pin package all together. Plexiglass change is pretty straight forward procedure. Just put on some nitrile disposable gloves and blow dust with compressed air from between plexiglass and dark back-cover. With new transparent glass, every single dust-spec is very visible.

Decoder board needed some additional modifications. I decided to replace ceramic DEM caps with Panasonic 1uF SMD film ones. This one mod has most significant sonic impact on every TDA154x player that I ever worked with. Standard through-hole film caps works too, but looks terrible and wouldn’t fit here. Also make sure to solder them sideways! Otherwise there is too little spacing between them and there will be bit-to-bit leakage. It will sound and measure just dreadful. Trust me. And don’t ask me how I know that 🙂

Another “must have” mod on these boards is re-routing bit currents directly to analog GND pin. I have made remarks on PCB layout above. Also you can notice cut tracks on photo with new DEM caps. And before someone starts arguing with me that “Phillips engineers knows best” I should like to remind you that this is a single layer board. No way they could’ve done that without bodge-wires. It is extremely important that these currents won’t mix with anything else on the board.

Finally we are disconnecting op-amp I/V circuit from TDA1541A by removing two jumper wires (red arrows) and modifying MUTE circuit. This is necessary if you are planing to use tube I/V stage. Hearing how an output cap is quickly charging to 150V at turn on is not funny at all. This might damage your pre-amp inputs or even pop your speaker woofer. Original mute circuit was modified as per schematics above and now it provides 30s delay, before muting relay is disengaged.

All mechanical parts took a bath and I started working with a drawer mechanism. The reason for slow or incomplete closing was stiff and overstretched rubber belt. No wonder as 35 years is almost eternity in rubber time. Now looking back I should’ve replaced that yellow belt gear. It looked suspicious but still strong and of-course it failed after one month or so. It’s still readily available on eBay, Alliexpress etc.

Word of caution about those Chinese R-Core transformers that you can buy everywhere. Above is a 20W version, but I bet they’re all build up to same standards. I wanted to change leads for longer and softer ones and I didn’t liked what I found beneath the first isolation layer. Those wires are almost shorted there! Some heat-shrink will fix that, but yeah… doesn’t inspire confidence. It’s cheap though! And it has shielding between primary and secondary, which is mandatory for all gear I build. Also if you keep load below 80% of rated, then magnetic field leakage is really low.

Now the fun part called system integration begins. It’s great for everyone to have a lot of ideas. Tubes, transformers, capacitors, rectifier boards, you name it. Yet finally someone has to come and put it all together. But how do you do that? And more exactly where in this case? Space is at premium inside 304 because of all the shielding and heavy sub-frame. I tried googling if somebody managed to squeeze it all in there, but I failed to find a single example. I was not going to go full “Lampizator” mode and put holes in covers with tubes outside. OK, challenge accepted then!

After a lot of measurements and a lot of question marks I finally managed squeeze it all in. Very tight fit, but still doable. I installed additional aluminum cross-bars for holding capacitors and made a new frame for tubes. When all was said and done it was time for final assembly.

Also changed original TDA1541 to newer TDA1541A R1 and installed a gold-plated socket for updated SAA7220P/B digital filter. This way board can be reconfigured for NOS operation by installing couple jumpers, although I’m not NOS fan in medium-modified CD players. Too much treble energy is lost.

All primary rectification, delayed 6.3V tube heater supply and HV filtering was assembled on proto-board. I don’t think it needs further explanation. Classical delayed regulator on LM317 and simple mosfet cap-multiplier. All tube circuitry was mounted on the sockets. 6N62P-EV is ECC83 equivalent and needs heavy selection, as do almost all soviet tubes. It has enough gain for a very low I/V resistor of 15Ω  so 2mA bias current for TDA1541A output can be omitted. First stage output resistance is very high so it needs buffering.  E88CC/6922 works as a cascoded cathode follower and lower mosfet set’s the current around 10mA. Depletion mode DN2540 would be better there, but at the time it was out of stock everywhere. This stage has around 110Ω of output impedance and is capable of driving any cable of reasonable length.

And above is the final result. Clean! If I may say so myself 🙂 Really happy how it turned out. You can say all you want about tubes, but that worm light is really soul-melting and it’s even before any sound comes out! And boy does it sing. Difference between op-amp and tube I/V stage always blows my mind. So much more musicality and details. I understand that this is just an illusion. Tube distortion is doing some masking of all the flaws intrinsic to every CD player. And there is a lot to mask here – terrible EMI from all the motor drivers and servos, jittery as hell master clock, awful digital filtering, large 5V TTL signals, hideous decoupling of digital circuitry and the list goes on and on… I might even run out of synonyms for word terrible. And then with a help of a long forgotten vacuum device you flip the switch and this mess suddenly becomes something that you can actually enjoy. It’s still a very long way from what TDA1541 can do in it’s ultimate implementation, but nevertheless a delightful presentation.

Above is a couple of measurements I’ve done to make sure everything runs as intended. Here first red graph is 0dBFs 1kHz signal at 1.9Vrms. Distortion is totally dominated by 2nd harmonic at -60dB which translates to about 0.08% THD. Low enough for me. This is a very typical spectrum for ECC83 and it’s relatives. Next blue graph is cross-talk between two triodes of the same tube (other half has 1.9Vrms signal at 0dbFs). This is a price to pay for sharing same high-gain double-triode between both channels. On the other hand best vinyl has 40dB of channel separation and sounds stellar. So 75dB here is again good enough in my book. And last magenta one is system noise-floor. This spectrum was captured with no CD in tray.

As I already mentioned 6N2P even in its special “EV” version is still all over the place. I bought 20pcs and only 3 where about same gain and didn’t had any defects like “popcorn” noise, increased distortion or grid-leakage. But those 3 selected ones are on-pair with best ECC83 out there like Siemens, Telefunken, VALVO etc. Can’t hear any difference. This is not the case when comparing to “modern” JJ production, which is an obvious step backwards from any good vintage tube. Haven’t tried TAD yet. Anyhow, here is some more images of the final photo-shoot.

Copyright © 2022 MV Audio Labs. Legal Disclaimer.