Hello.
I've just arrived blinking from the world of valves into the 80s and got a CD player because everyone's throwing them out. It's a Trio (Kenwood) 770. Nothing special, so far. It's got a heavy steel chassis, chunky pressed-steel mech with jiggling linear sled for the cast-alloy laser assembly with its delicious plastic focus centering spring as illustrated. It's logically laid out with bits that, mostly, can be seen with the naked eye and soldered with bare hands.
My issues centre on the current PWM-output DAC.
I've attached pics of the relevent sections of the player, which closely follows the layout and circuit in the datasheet app. notes. Also a picture of the following integrator output for 1kHz full scale, with a section of the higher trace magnified on the lower (the multiple images arise because 44.1 isn't an integer so the picture's jiggling).
Datasheet is here:
CX20152 datasheet pdf datenblatt - Sony Corporation - Dual 16 Bit, 88kHz, Multiplexed D/A . ::: ALLDATASHEET :::
Data for one channel are shifted into two 8-bit registers. When done, these registers are used to set two timers, and then become available for the data for the other channel. Meanwhile, two current sources are switched to the first channel current output. One current source, timed by the data from the less significant byte, is 1/256th of the magnitude of the other. Consequently for full scale the counters must count to just 256 with some time to spare before the current sources are required for the other channel. The frequency is supposed to be around 65MHz according to the datasheet. Sheets for other similar DACs show different frequencies...so precision seems more important than accuracy here.
In addition to the CCS outputs, there are three others intended for downstream audio functions as shown in the datasheet: one "discharge" pulse for each channel intended to trigger the discharge of the respective integrating capacitors, and a clock intended for the timing of the sample-and-hold stage that is supposed to fill in the missing triangular sections of output, so it should end up looking like a staircase with no chasms between the steps.
Have I got this right? In particular, the datasheet doesn't actually say that there are two timers set by the two 8-bit registers, and there is no evidence of the change in gradient of the integral voltage that would ensue whenever the LSB and MSB are not identical, which is nearly always.
Secondly, the operation of the sample-and-hold circuit is clouded in my mind by the bias and feedback circuitry. I can't see how it maintains a stable voltage when the opamp input is disconnected and left floating. I'm quite new to transistors and opamps as well as digital...
Thirdly, the Kenwood implementation differs in that AFAICS there is no crystal for the timers' clock. There is a transistor and a couple of caps and that seems to be all. Does anyone know, or guess, how the clock is generated? There is what looks like a ceramic resonator some distance away, and in between I guess there's a SM interface chip on the bottom of the board so the resonator will be for that, I guess.
Fourthly, there appears to be no ground plane on any of the (Elna) pcbs, and precious little screening anywhere. Is this likely to prove fatal to my audiophile aspirations?
Nearly lastly, what's the quarter-frequency "SCLCK" output for?
My feeling is that there is lots of room for improvement but I'm having difficulty deciding if it's worth the effort. At the moment I use the player and don't want to take it to bits until I feel more confident there's something I can do. It has grunt but not a lot of finesse. The tinkling bells in my Dvorak's New World Symphony are a bit toneless and lost in space. Not nearly as bad as they were on the first, very modern, CD player I tried, but not as good as the Marantz CD50 I also listened to.
This and very similar DACs are used in many early Sonys, including the iconic CDP-101 and CDP-7F, so improvements may have wider value. I've looked around for work already done but it seems restricted to replacing opamps with expensive audio-specific types, which I think may be missing the point.
The output from the opamp after the switches looks like mayhem, but so far I'm not sure if my scope probes are partly responsible because I think its a high impedance node with high frequencies present in abundance. There may be much narrower but much steeper chasms between the steps, followed by large sharp overshoots. I'll post a pic if can when I work out how.
So, finally, and crucially, is there a better way of doing the sample-and-hold thing, or better components for the job as it is done, such as faster or more pure switching devices. Is it even necessary to convert the DAC output to voltage PAM (pulse amplitude modulated), or is there an alternative route, perhaps via conversion to voltage PWM?
It's a great American orchestra and a great American recording. I really should try to get the bells right.
cheers, Ian
I've just arrived blinking from the world of valves into the 80s and got a CD player because everyone's throwing them out. It's a Trio (Kenwood) 770. Nothing special, so far. It's got a heavy steel chassis, chunky pressed-steel mech with jiggling linear sled for the cast-alloy laser assembly with its delicious plastic focus centering spring as illustrated. It's logically laid out with bits that, mostly, can be seen with the naked eye and soldered with bare hands.
My issues centre on the current PWM-output DAC.
I've attached pics of the relevent sections of the player, which closely follows the layout and circuit in the datasheet app. notes. Also a picture of the following integrator output for 1kHz full scale, with a section of the higher trace magnified on the lower (the multiple images arise because 44.1 isn't an integer so the picture's jiggling).
Datasheet is here:
CX20152 datasheet pdf datenblatt - Sony Corporation - Dual 16 Bit, 88kHz, Multiplexed D/A . ::: ALLDATASHEET :::
Data for one channel are shifted into two 8-bit registers. When done, these registers are used to set two timers, and then become available for the data for the other channel. Meanwhile, two current sources are switched to the first channel current output. One current source, timed by the data from the less significant byte, is 1/256th of the magnitude of the other. Consequently for full scale the counters must count to just 256 with some time to spare before the current sources are required for the other channel. The frequency is supposed to be around 65MHz according to the datasheet. Sheets for other similar DACs show different frequencies...so precision seems more important than accuracy here.
In addition to the CCS outputs, there are three others intended for downstream audio functions as shown in the datasheet: one "discharge" pulse for each channel intended to trigger the discharge of the respective integrating capacitors, and a clock intended for the timing of the sample-and-hold stage that is supposed to fill in the missing triangular sections of output, so it should end up looking like a staircase with no chasms between the steps.
Have I got this right? In particular, the datasheet doesn't actually say that there are two timers set by the two 8-bit registers, and there is no evidence of the change in gradient of the integral voltage that would ensue whenever the LSB and MSB are not identical, which is nearly always.
Secondly, the operation of the sample-and-hold circuit is clouded in my mind by the bias and feedback circuitry. I can't see how it maintains a stable voltage when the opamp input is disconnected and left floating. I'm quite new to transistors and opamps as well as digital...
Thirdly, the Kenwood implementation differs in that AFAICS there is no crystal for the timers' clock. There is a transistor and a couple of caps and that seems to be all. Does anyone know, or guess, how the clock is generated? There is what looks like a ceramic resonator some distance away, and in between I guess there's a SM interface chip on the bottom of the board so the resonator will be for that, I guess.
Fourthly, there appears to be no ground plane on any of the (Elna) pcbs, and precious little screening anywhere. Is this likely to prove fatal to my audiophile aspirations?
Nearly lastly, what's the quarter-frequency "SCLCK" output for?
My feeling is that there is lots of room for improvement but I'm having difficulty deciding if it's worth the effort. At the moment I use the player and don't want to take it to bits until I feel more confident there's something I can do. It has grunt but not a lot of finesse. The tinkling bells in my Dvorak's New World Symphony are a bit toneless and lost in space. Not nearly as bad as they were on the first, very modern, CD player I tried, but not as good as the Marantz CD50 I also listened to.
This and very similar DACs are used in many early Sonys, including the iconic CDP-101 and CDP-7F, so improvements may have wider value. I've looked around for work already done but it seems restricted to replacing opamps with expensive audio-specific types, which I think may be missing the point.
The output from the opamp after the switches looks like mayhem, but so far I'm not sure if my scope probes are partly responsible because I think its a high impedance node with high frequencies present in abundance. There may be much narrower but much steeper chasms between the steps, followed by large sharp overshoots. I'll post a pic if can when I work out how.
So, finally, and crucially, is there a better way of doing the sample-and-hold thing, or better components for the job as it is done, such as faster or more pure switching devices. Is it even necessary to convert the DAC output to voltage PAM (pulse amplitude modulated), or is there an alternative route, perhaps via conversion to voltage PWM?
It's a great American orchestra and a great American recording. I really should try to get the bells right.
cheers, Ian
