Thanks for your extremely polite way for pointing out my errors
Unlike my wife who cannot stand any constructive criticism I love it as that is the way I can learn.O.K. I simulated the I/V opamp stage using an ideal opamp and found that if the input impedance is 0 then there is a DC gain of 100dB and it is a low pass filter with a corner frequency close to 100Hz, regardless of the capacitor and the resistor values. If the input impedance is 820R then the gain is 0dB and the response is a flat line. If the input impedance is infinite then the DC gain is -60dB and the response is also a flat line.
A couple of hours ago I found out that both the I/V opamps (OP275) and filter opamps (NJM4565) don't have the 0.01uF local bypass as indicated in the schematic! They are there in the schematic (some are marked as NC - not connected while some are not) but on the board I can't find any of them. So they are all NC. I can see the soldered pads though. So if I want, I can solder them on the existing solder pads. The sizes appear to be 0850 (2.00mm x 1.25mm).
If I am going to use LM6172 (100MHz) and AD8066 (150MHz), I think I have to follow the datasheets. But the fact that Marantz did not install them shows there must be some bad effects, perhaps as Abraxalito pointed out that they make the ground dirty, or they interact with the power supply regulators, or they introduce LCR resonance with track inductance.
Abraxalito, what bypass cap value do you use across the supply pins? Do you solder them on top of the +/- supply legs of the 8-SOIC? I can't imagine I can do that sort of sodering...
If I am going to use LM6172 (100MHz) and AD8066 (150MHz), I think I have to follow the datasheets. But the fact that Marantz did not install them shows there must be some bad effects, perhaps as Abraxalito pointed out that they make the ground dirty, or they interact with the power supply regulators, or they introduce LCR resonance with track inductance.
Abraxalito, what bypass cap value do you use across the supply pins? Do you solder them on top of the +/- supply legs of the 8-SOIC? I can't imagine I can do that sort of sodering...
Last edited:
There was a paper linked on this Forum where the author gave a lot of attention to locating the chip decoupling.
His preferred (if I remember correctly) was a copper plane over the whole area of the chip top surface. One side of the plane connected to the Power PIN and the other side had the decoupling cap between the plane and the other Power PIN.
Shame I cannot remember the author. I guess it came out in the 90's
Having read H.Ott, I can see some of Ott in that original paper.
His preferred (if I remember correctly) was a copper plane over the whole area of the chip top surface. One side of the plane connected to the Power PIN and the other side had the decoupling cap between the plane and the other Power PIN.
Shame I cannot remember the author. I guess it came out in the 90's
Having read H.Ott, I can see some of Ott in that original paper.
Here is another progress update.
I guess I may have now finished round 3 mod - unless I am not happy with it tonight.
I made some more careful measurements. For the high frequency SMPS small transformer, noise in one winding is reflected in another through capacitive coupling. For an example, the noise ripples in the +5V supply has almost exactly the same pattern as in the +12V, and the +12V shares about 80% of the noise pattern with -12V, etc. So even though only the +/-12V supplies affect the sound, in order to reduce noise, resonances in any of the windings need to be suppressed as much as possible.
Another round of LTSpice simulations were taken place. The snubber circuits were refined and reduced.
The newly bought 22uF 25V X7R caps did not work well. In simulations, in which all parasitics (cap inductance, resistance, wire inductance, parasitic capacitance, etc) were included, the caps should work well. In reality, after introducing these caps on the rails I could see more resonances on the rails. My guess is that they may resonate with the electrolytic caps or the circuits in whole. Subjective listening confirmed this as the sound lost quite a bit of resolution.
So instead of using two of them, I removed one of them, and add a 0.22R 0.6W leaded resistor in series with the remaining 22uF 25V X7R. This is repeated for each of the secondaries +12V, -12V, +5V and Controller. This worked very well in both simulation as well as in the real circuit.
I have also altered some values of the surrounding capacitors.
I previously suggested that after the initial round 3 mod all ringings disappeared while the pulse amplitudes increased. Part of that was not true as I later found out that I accidentally put the probe from X10 to X1 while the scope setting was on X10. After last night's (final?) mod, all major ringings disappeared, and the pulse amplitudes decreased. However, I found that minor ringings in much higher frequencies (above 25MHz, notably one near 100MHz) still persist. This may be due to the length of the leaded parts of the snubbers. Because these snubbers were "Add-On"s and used non-SMD parts, I thought they could not be expected to deal with frequencies above 25MHz. Perhaps that 22uF + 0.22R has already reduced the amplitudes of these ringings.
I put on a broadband clamp-on ferrite (Digikey 240-2124-ND) on the +/-12V and ground supply wires to the analogue board. There had been a small one put on by Marantz. This new one obviously reduced noise above 25MHz by a lot. Although I could not measure it from the analogue board, subjective listening found the high frequency noise much reduced. So I experimented adding another one, and once again, it was even better. Would the wire length accommodate a 3nd one? Would a 3rd one make it sound even better? I don't know. These broadband clamp-on ferrite beads were possibly the most effective, simplest and lowest costs devices used in this series of mods.
By now, 90% of my original complaint about high frequency noise of the player has been removed. I played 4 blu ray opera discs last night to verify the sound with soprano voices. Amazingly, for the very first time I could not hear the "siblings zzz..." in their singing any more, as it is truly gone, and their words became more intelligible. The song "Sweet Poison" in "The Fairy Queen" disc is a classic example. There is this massive improvement comparing to the original stock player.
Hearing some volins, violas and cellos though I believe the tonal accuracy of the player will still need to be improved for the player to be classified as a high end player.
Is it still possible to improve the SMPS further in order to make the volin tone more correct? or is that the fault of the low bandwidth opamps in the DAC board? I really don't know the answer.
What should I do next? mod the analogue board with the newly purchased AD8066 and LM6172?
I guess I may have now finished round 3 mod - unless I am not happy with it tonight.
I made some more careful measurements. For the high frequency SMPS small transformer, noise in one winding is reflected in another through capacitive coupling. For an example, the noise ripples in the +5V supply has almost exactly the same pattern as in the +12V, and the +12V shares about 80% of the noise pattern with -12V, etc. So even though only the +/-12V supplies affect the sound, in order to reduce noise, resonances in any of the windings need to be suppressed as much as possible.
Another round of LTSpice simulations were taken place. The snubber circuits were refined and reduced.
The newly bought 22uF 25V X7R caps did not work well. In simulations, in which all parasitics (cap inductance, resistance, wire inductance, parasitic capacitance, etc) were included, the caps should work well. In reality, after introducing these caps on the rails I could see more resonances on the rails. My guess is that they may resonate with the electrolytic caps or the circuits in whole. Subjective listening confirmed this as the sound lost quite a bit of resolution.
So instead of using two of them, I removed one of them, and add a 0.22R 0.6W leaded resistor in series with the remaining 22uF 25V X7R. This is repeated for each of the secondaries +12V, -12V, +5V and Controller. This worked very well in both simulation as well as in the real circuit.
I have also altered some values of the surrounding capacitors.
I previously suggested that after the initial round 3 mod all ringings disappeared while the pulse amplitudes increased. Part of that was not true as I later found out that I accidentally put the probe from X10 to X1 while the scope setting was on X10. After last night's (final?) mod, all major ringings disappeared, and the pulse amplitudes decreased. However, I found that minor ringings in much higher frequencies (above 25MHz, notably one near 100MHz) still persist. This may be due to the length of the leaded parts of the snubbers. Because these snubbers were "Add-On"s and used non-SMD parts, I thought they could not be expected to deal with frequencies above 25MHz. Perhaps that 22uF + 0.22R has already reduced the amplitudes of these ringings.
I put on a broadband clamp-on ferrite (Digikey 240-2124-ND) on the +/-12V and ground supply wires to the analogue board. There had been a small one put on by Marantz. This new one obviously reduced noise above 25MHz by a lot. Although I could not measure it from the analogue board, subjective listening found the high frequency noise much reduced. So I experimented adding another one, and once again, it was even better. Would the wire length accommodate a 3nd one? Would a 3rd one make it sound even better? I don't know. These broadband clamp-on ferrite beads were possibly the most effective, simplest and lowest costs devices used in this series of mods.
By now, 90% of my original complaint about high frequency noise of the player has been removed. I played 4 blu ray opera discs last night to verify the sound with soprano voices. Amazingly, for the very first time I could not hear the "siblings zzz..." in their singing any more, as it is truly gone, and their words became more intelligible. The song "Sweet Poison" in "The Fairy Queen" disc is a classic example. There is this massive improvement comparing to the original stock player.
Hearing some volins, violas and cellos though I believe the tonal accuracy of the player will still need to be improved for the player to be classified as a high end player.
Is it still possible to improve the SMPS further in order to make the volin tone more correct? or is that the fault of the low bandwidth opamps in the DAC board? I really don't know the answer.
What should I do next? mod the analogue board with the newly purchased AD8066 and LM6172?
Nearly 10 months passed, I finally completed the the opamp upgrade job yesterday.
Replaced OP275 with LM6172.
Replaced njm4565 with AD8620. This is an old JFET opamp for possibly better RF immunity. It seesm everyone raved about this opamp though. I previously bought LM4562 for it but changed my mind.
Put SMD 0.1uF X7R directly on the power pins shunt to ground.
I did not do the soldering myself but a friend with the right soldering tools helped with the job.
I also changed the DAC output 4th order low pass filter from 105kHz down to 55kHz.
I measured the rails and did not find resonances other than normal rail noise and some residuals of SMPS noise, possibly at level of 3-5mV.
Does it sound better? I think so.
Replaced OP275 with LM6172.
Replaced njm4565 with AD8620. This is an old JFET opamp for possibly better RF immunity. It seesm everyone raved about this opamp though. I previously bought LM4562 for it but changed my mind.
Put SMD 0.1uF X7R directly on the power pins shunt to ground.
I did not do the soldering myself but a friend with the right soldering tools helped with the job.
I also changed the DAC output 4th order low pass filter from 105kHz down to 55kHz.
I measured the rails and did not find resonances other than normal rail noise and some residuals of SMPS noise, possibly at level of 3-5mV.
Does it sound better? I think so.
But surprisingly, cap signature makes an even bigger impression on sound.
I initially used some high ESR Nichicon caps after the 317/337 outputs in hope that the high ESR will damp any possible ringing due to the track L interacting with the 0.1uF X7R. Near the opamp I already had some Rubycon ZL low ESL 470uF caps for local bypass. The sound was harsh and worse than before the opamp upgrade.
Replacing the Nichicon caps with Rubycon ZL 100uF 25V (0.13R) got rid of the harshness and made the overall sound better than before the opamp upgrade.
I initially used some high ESR Nichicon caps after the 317/337 outputs in hope that the high ESR will damp any possible ringing due to the track L interacting with the 0.1uF X7R. Near the opamp I already had some Rubycon ZL low ESL 470uF caps for local bypass. The sound was harsh and worse than before the opamp upgrade.
Replacing the Nichicon caps with Rubycon ZL 100uF 25V (0.13R) got rid of the harshness and made the overall sound better than before the opamp upgrade.
I was totally shock when I measured the DAC power supplies.
The 5V supply has some mild resonance in the MHz region with PtP noise of about 5mV at the regulator output, but 10mV at the DAC power input.
The 3.3V supply has resonance in the MHz region with PtP noise of about 20mV!!! It is apparent that the regulator, electrolytic capacitor, and the local bypass 0.01uF X7R at the power pin and the tracks interact to cause the resonance.
They may be measurement artifacts but I am not sure. I soldered anchor points in the middle of the analogue and digital ground planes, and put the probe at the DAC power input and found those resonances.
The player still sounds pretty fine. Perhaps the digital circuitary is not sensitive to power supply noise?
The 5V supply has some mild resonance in the MHz region with PtP noise of about 5mV at the regulator output, but 10mV at the DAC power input.
The 3.3V supply has resonance in the MHz region with PtP noise of about 20mV!!! It is apparent that the regulator, electrolytic capacitor, and the local bypass 0.01uF X7R at the power pin and the tracks interact to cause the resonance.
They may be measurement artifacts but I am not sure. I soldered anchor points in the middle of the analogue and digital ground planes, and put the probe at the DAC power input and found those resonances.
The player still sounds pretty fine. Perhaps the digital circuitary is not sensitive to power supply noise?
Digital circuitry can be very insensitive to supply noise as I understand, but not always. Ideally, I would think digital would be affected by supply noise only at extreme levels.
What would concern me is if the supply noise comes with ground noise. Also, any resonances radiate to other parts of the board. After all the energy in the resonator is half stored in the magnetic field of the inductor.
There is also the possibility that resonances are somehow increasing the performance of the circuit. I think in some cases this is possible, and that it is also possible for resonances to cause pleasant distortions.
What would concern me is if the supply noise comes with ground noise. Also, any resonances radiate to other parts of the board. After all the energy in the resonator is half stored in the magnetic field of the inductor.
There is also the possibility that resonances are somehow increasing the performance of the circuit. I think in some cases this is possible, and that it is also possible for resonances to cause pleasant distortions.
Now when I think of it, it makes sense to me that unless the noise level is approaching some threshold towards the 1 and 0 bit processing level of the DAC the noise should not affect the 1 and 0 streams so normal level of noise in the digital supply should not affect sound quality.
However, the noise in the ground can, or radiated EMI interference.
In this CD player, the digital ground and the analogue ground is connected via a 2200pF polyester capacitor. There may be other capacitors to connect the digital ground to the chassis which the analogue ground is directly connected to.
However, the noise in the ground can, or radiated EMI interference.
In this CD player, the digital ground and the analogue ground is connected via a 2200pF polyester capacitor. There may be other capacitors to connect the digital ground to the chassis which the analogue ground is directly connected to.
for recognising the 0 and 1 level noise does not matter that much.
But the DAC reads the rising slope of the change from 0 to 1 and reads the falling slope of the change from 1 to 0.
If these slopes at vertical 1MV/µs rise/fall time, then the DAC has no problem with timing.
But the slopes are not vertical. And worse, the slopes are contaminated with noise and interference.
Now the threshold for reading the rising slope can be altered by the noise on the slope. Similarly the falling slope threshold can be altered by the noise content.
The timing is critical to what comes out of the DAC.
Similarly if the "clock" signal has noise on it, that too affects the timing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.