I have been doing a series of modifications to the analogue stage of my
marantz CD60SE recently. This is a standard 16*4 Phillips chip set but is
the "Special Edition" version of the CD60.
I also listened to the standard version and they were very different with the standard version being thin and harsh sounding with less detail. I bought the workshop manual for the CD60 and as far as I can tell the differences between the two is the extensive use of Elna cerafine electrolytics for power supply decoupling next to most chips and a pair used back to back for output coupling.
The analogue filter stage has a curious mixture of discrete and surface mount resistors but the caps are Phillips polypropylenes. Deemphasis is switched with a fet.
My first change was very conventional. I removed the NE5534 dual opamps
and replaced them with AD712s (A$2.30 each).
This change was assessed double blind by changing one channel (but I didn't know which) then getting a friend to decided which channel was better and why. We had exchanged impressions and were in fair agreement. Then the circuit was traced and we found the changed channel was the prefered one.
Essentially the change could be described as more detail but slightly darker sound and less tizzy.
The second change was to replace the AD712s with AD827s (~A$9.50 ea). This
gave a smaller, further improvement in detail and restored the original
tonal balance. At this point I was convinced that any noticable improvement
was good value at under ~A$20 (~US$15).
For convenience the original opamp was replaced by a turned pin gold plated socket which made trying different opamps easy. For further opamp improvement I expect you would have to use single opamps with extra decoupling caps for each but this could still be mounted on a component header to make for easy and reversible changes. I haven't tried this however.
All the above is fairly conventional and not at all original.
The third change was more radical and was an original idea (though it turns out that it is essentially the same as PS Audio do in some of their DACs and the idea has been adopted by at least one other person who has developed a fancy stand alone DAC ).
Basically my thoughts were as follows: Current output DACs like the TDA1541A don't like to see a voltage on their output and ideally would like to see a very low impedance to ground. The ideal current to voltage converter would otherwise be just a resistor which would produce a voltage across it proportional to the current. (I read that in an application note for an AD opamp recommended for current to voltage conversion).
I then wondered what impedance opamp I/V converters present to DACs in usual applications and was told that an opamp set up for I/V conversion presents a virtual earth to the DAC as its output feedback works to keep its inverting input (DAC connection) at earth potential.
So I thought what if I use a small resistor to generate a small voltage? This should be the ideal I/V converter but for one problem: as the resistor gets smaller the voltage across it for a given current gets smaller but the DAC is (presumably) happier.
The small voltage will need to be amplified and if it is too small noise becomes a problem [but not all bad as the noise voltage of such a small resistor will also be small]. My friend did a calculation for the noise figures of the AD797: one could use a 10 ohm resistor for a 2mA peak current (-> 20mV) amplified to 2V out and still get 100dB SNR.
If the resistor is bypassed with a capacitor the voltage across will be determined at low frequency by the resistor but at high frequencies will be less and less with increasing frequency. This appealed to me because the ultrasonic output of the DAC need not even enter the first acitve analogue device and so its performance at >audio frequencies would be of less consequence. Also I could shunt the sampling frequency to earth with a series inductor-capacitor.
The result of this is a 27 ohm resistor across the output of the DAC
conveniently mounted on a component header. The value of the resistor was
obtained by trial and error to match the output volume (with the gain stage
below) to that of my phono stage.
Across this is about 133nF of capacitance made of a mix of polyesters and polystyrenes I happened to have. My calculations say this gives the same cutoff frequency as was in the player originaly and a 6db/octave slope. I also have 1mH-820pF series circuit across the resistor for the notch at 4x44.1kHz (the inductor has 3ohm DC resistance so it is not a deep notch).
Also across the resistor is a flying lead out of the player to a separate gain stage. This gain stage is followed by an RC filter to give a steeper slope to the filter at higher frequencies and the notch again, this time much deeper. This then sees the 100Kohm vol control of my preamp.
I was able to leave the original circuit of the player intact with this arrangement which will be useful for later sale of the player or return to the original while doing further mods to the outboard gain stage.
The gain stage is [was] a single 12AT7 valve/per channel used as a shunt regulated push pull (SRPP) circuit which is class A, linear and uses no conventional feedback.
Listening tests indicate that the noise of this circuit is low enough to not cause any problem. It is also satisfying that this stage is essentially the same as the second gain stage of my phono amp which makes any CD/LP comparisons a bit fairer. Subjectively the sound has moved closer to my turntable but still not as good. It seems more open and detailed, less tizzy and generally more emotionally involving with this circuit compared with any of the opamps in the original circuit. The tonal balance is much the same as original -only the AD712 seemed to change this.
It also is nice that it compliments the rest of my system which is a SRPP
valve phono amp and line amp and a pair of Leak TL12 valve power amps (much
modified) driving much modifed Allison One speakers. (The turntable is a
LinnLP12/Ittok/ortofon MC3000 with T3000 transformer to stepup to MM level)
For the listening tests I also used a large home made mosfet amp (AEM6000)
and a pair of Magnepan Typanis which were on loan.
The circuit lacks either the deemphasis circuit or output muting. So far I havn't encountered a disc with an obvious need for deemphasis and I understand they are rare but I haven't looked far yet. [I have now installed an LED which indicates when DEEM is required. It is only on a very few CDs and there are few discernable patterns as to which ones they are. I plan to install a switchable, parallel network of one resistor and two caps to do the DEEM in parallel with the I/V resistor].
Changing tracks etc does not produce any clicks or thumps so I don't know what the output muting in the original circuit was for -perhaps it helps for damaged discs?
This experiment was done from a position of semi-ignorance and I would enthusiastically welcome any comments or suggestions for further improvements. Here is the circuit of the amp I used for the gain stage:
+427V | plate ______grid | cathode | |____________||___________ out | 2Kohm | 1K5 | |________| | | | | | plate | | in__||_________grid 10Mohm 2M2ohm |1uF cathode | | 200ohm | | | |_____________|_______|_________|___signal earthIf you want to make one of these there are other considerations like a floating heater supply so that the max heater to cathode voltage is not exceeded for either triode. The version of this I used for the CD uses cascaded RC filters for the supply whereas the my phono amp has which uses similar gain stages has an IC regulator (but valve rectification) followed by RCs. The latter amp also has a slow heater turn on over 30secs after which the high voltage comes on and after the amp has settled an output relay is closed (at 1 min).
The 1uF cap allows grid current bias of the lower triode.
The rest of the system has changed a bit since I wrote the above. I now use a Tannoy Monitor Gold dual concentric speaker in a 300L reflex box with a very simple two stage triode input, single-ended pentode output amp with a small dose of feedback. There is now no line amp so there are just three valve stages between the DAC chip and the speakers. For phono there are four stages and a step up transformer from the cartridge to speakers. I have also done some further thinking and leaned a little more about how to use loadlines etc. I am now planning to modify the gain to replace the upper cathode resistor with a FET current source which should make the stage more linear still. This will require some changes to the valve type and I will not necessarily use the same type of triode for upper and lower. I will try a more conventional bias arrangement for the lower triode which will remove the input cap. The DC offset is small compared to the bias voltage so this should not matter. I may try a voltage reference with cap bypass instead of the conventional resistor and cap bypass or I may have an unbypassed cathode resistor which will introduce a few db of local feedback.
More time passed........
I have now done the fet current source idea. I had an ASCII diagram of it somewhere but can't find it. It is a modifiction of the same amp I used above so the power supply, output coupling cap, most wiring etc remains the same. I will describe it for you and I'm sure you could understand it. The I/V resistor signal end is direct coupled to the grids of both halves of a 12AX7. Their plates are also joined but they have separate, unbypassed cathode resistors to ensure current sharing and to provide bias. This eliminates the first coupling cap in the above diagram. The lower plates (12AX7) connect to the grid of half of a 12AU7 which replaces the upper triode in the diagram. The 2K resistor between the two triodes is replaced with a 2mA "constant current diode" which has a dynamic impedance around a megohm (from memory). Thus the lower triode operates at near full gain in spite of the lack of cathode bypass. The upper triode hopefully acts as a more ideal cathode follower. The drop across the lower triode is about 250V, with 1mA per half 12AX7 (and 1K or 1K5 cathode resistor? can't remember) -whatever seemed a nice linear part of the curves. Compared with the previous circuit I think this one sounds "faster" and more "open" and detailed. On one CD it seemed to introduce a little harshness to a female voice but I suspect the recording as it is an improvement on everythin else. The sound is still not an equal for my Linn LP12/Ittok/Ortofon MC2000II.
I have plans to make separate DAC with a better DAC chip, separate regulation for everything etc with my EE friend when he is ready to etch boards and do the design. He has bought a duplicate set of all the major parts. I will retain the same I/V circuit.
© Copyright 1997 Peter Campbell