Those two AN8005 are bare minimum, I was rather talking about replacing them with discreet regulators, which is not always needed.
CS8412 is plentifully available on ebay. Look for the CS8412CP if you are after the DIP version.
Ebay search for CS8412.
Ebay search for CS8412.
One more thing has just occurred to me, that I forgot to mention, regarding the simplified schematic above. It will be obvious to most, but I will say it, just in case - the output capacitors (C13 and C14) need to be non-polarised, so either film caps or non-polar/bi-polar electrolytics. The value is not very critical - you could use 10x bigger or smaller without running into any serious trouble.
A point of clarification that seems likely to have already been addressed. I have one of those passive output NOS 4x1543 DAC's. As the TDA1543 literature calls for an inverting opamp connected to the output is it reasonable to conclude that these passive output DAC's require an inverter on the data line to produce the correct phase?
TDA1543 does reverse the absolute phase. Some people can hear the difference, but most do not. If you really want your DAC to produce the correct phase, you can either add an inverting opamp at the output, or use something like SN75179 to invert the digital input - but honestly, reversing speakers cables is much more sonically transparent...
Thanks. I am easily disturbed by an inverted phase signal from side to side, however this doesn't mean that changing absolute phase of both channels would be nearly that obvious. Personally, it is a matter of maintaining correct polarity.
I also want to do comparisons with other DAC's and don't like the idea of switching speaker leads to address the inverted phase of variant DAC's. The other issue is that all the other inputs to a system could end up being inverted by changing the speaker leads.
The SN75179 requires a 5 volt supply that isn't available in the unit. The DIR9001 receiver runs from 3.3volts. I was thinking of using the 5 pin surface mount SN74LVC1G86 "exclusive or" chip. In this way I can feed the data stream into one input and a high or low level into the other input to select the phase of the data stream coming out.
I also want to do comparisons with other DAC's and don't like the idea of switching speaker leads to address the inverted phase of variant DAC's. The other issue is that all the other inputs to a system could end up being inverted by changing the speaker leads.
The SN75179 requires a 5 volt supply that isn't available in the unit. The DIR9001 receiver runs from 3.3volts. I was thinking of using the 5 pin surface mount SN74LVC1G86 "exclusive or" chip. In this way I can feed the data stream into one input and a high or low level into the other input to select the phase of the data stream coming out.
Well, do as you see fit, but bear in mind that I was serious about the sonic transparency issue. There is a reason why the SN75179 was removed from the latest DAC revision - it flattens dynamics, and veils the sound. Any other solution that you might add to address the phase issue will also inevitably impact the sound. I'm afraid if you will have to put up with reversed phase or swapping cables - otherwise your comparison will be rendered meaningless (or at least unfair) by the additional components.
I should mention that the DAC is the MUSE 4xTDA1543 DAC. It has some serious issues, not the least of which is that some of these units are considerably more veiled than others. Nevertheless, my general concerns are also that digital artifacts can degrade sonic performance. I appreciate the trepidation in adding more digital components in the data stream. The risk of deterioration seems greater IMO because the TDA1543’s are fed from the output of a DIR9001 receiver, a device that has good inherent jitter performance and high speed outputs (less than 10nsec transition times).
I understand that noise, random or otherwise, exists at some level and can be superimposed on the digital signals. This is then fed into an input having some "fixed" trigger threshold (perhaps the DAC’s). The digital lines are always sloped to whatever degree, with the probability of jitter as dependent upon the slope of the digital lines and the magnitude of the noise component that triggers a subsequent component. Jitter is further dependent upon any variant propagation times of added active components that can be dependent upon its power supplies.
As you suggest, adding components appears hazardous if not impossible to maintain adequate jitter performance. In my own circuits I have even added a coil across the resistors feeding the DAC's on the clocking lines from the receiver to speed up the transition times.
I've done a lot of mods to the Muse 4*TDA1543 which I talked about a bit on HeadFi. The inversion can be done with one of the devices already on the board - that is if you don't want the optical input. I included attenuators between the DIR9001 (3V3 operation) and the TDA1543s which will accept 2V for logic 1. Other improvements came from isolating the ground fill from pin4s of the DACs and explicitly wiring star grounding wires - seems shared ground impedances add 'greyness' and sibilance to the sound. Also I deleted the cap shunt filters across the I/V resistors.
My TV only has an optical output so I don’t want to eliminate this feature. Like you, I also deleted the capacitor shunt filters across the I/V resistors and abandoned several other normally sound engineering practices as well. I don't want to confess to some of these mods lest reaping the full wrath of the "gods". In some trepidation of confession I also removed all the ceramic caps across the DAC's and added a loosely wound coil in series with the power supply line to the DAC's. This seems of similar advantage of star grounding by removing the tight control of the power supply regulator and other components connected to alternative grounds. In other designs, like that of Peter Daniel, he uses power supplies of less tight regulation that can improve the effectiveness of existing grounding, those that may not be of single point design. Both techniques could be of advantage in making a new PC board.
The TDA1543 contains an intermal digital serial to parallel converter. This digital network is in some proximity to the analog current sources in the chip and will couple some capacitive effects into the output analog signal. This is to whatever extent is audible or otherwise. Unlike the digital data line going into the first in a series of serial shift registers, the input clock lines acts on all registers doing the shifting inside the TDA1543, hence it can be concluded that higher capacitive elements, as a summation of the capacitance in all the registers, can have greater impact on the analog outputs than that of the singular data line input. Of all input lines the digital clock seems would have greatest impact.
It is understood that issues exist when the input capacitance of the DAC’s are attempted to be instantly charged from some receiver or other logic gates. However, as soon as a slope is introduced this invites jitter from a number of sources of noise, including random noise. IMO it is important to maintain the jitter capability of the receiver as minimally affected by subsequent slope deterioration. Ultimately it seems the jitter characteristics from the two input clocks are most important, not so much the data line. It may be a disadvantage to speed up the data input line, given its irregularity and that it acts on so little of the input circuits of the DAC. Using an attenuator as you suggest could be an improvement on the data input line to prevent alternative capacitive issues potentially of greater significance.
Using an inductor across the resistor from the receiver chip and in series with the DAC’s characteristic input capacitance forms a two pole filter. This permits faster transitions times as compared to just a series resistor and can support a faster high frequency roll off. I use approximately 4 or 5 loosely wound turns of magnet wire about 3/16" diameter as spread to connect to the ends of the clock line resistors on the Muse DAC.
Well if you don't need the coaxial input then the gate used for that is freed up to use as inverter. Or do you have something else feeding that? If you do then be sure to rewire the grounding on it so as not to spread CM noise across your grounds.
I use ferrite beads for the same purpose, they're smaller and cheaper than hand-wound coils. Save hand winding for passive output filters
That rather sounds bulky but an interesting idea
I might try implementing it with many ferrite beads in series. Be sure to put some kind of damping resistor across your coils - I find even ferrite beads need these.Nearly all comparisons and further adjustments are being done at a friend’s place using various digital coaxial cables including MIT high resolution musical instrument cable. The digital cables gave better results than optical cables. Nevertheless, I am using a generic optical cable for initial modifications and testing at home. The reason is that jitter, noise, etc. has the potential of being handled entirely inside the DAC in a shielded environment akin to a Faraday pail. However the galvanic isolation from ground loops and noise still appears less important than issues using the optical cable.
I also use ferrite beads in conjunction with a “loose” coil on the output lines and in other areas of the circuit as well. For many of the coils I just tightly winding magnet wire on some cylindrical tool, bend the ends and use a scalpel to remove the varnish from the ends. In all cases the coils needed to be pulled apart for some unknown reason. It may be related to reducing the inter-winding capacitance from end to end of the coil. Ultimately inductance is difficult to work with in relations to sonic influences, too much inductance is as bad as too little.
I also replaced the AMS1117 3.3volt regulator with a high current LM334 shunt regulator set for around 3.3 volts. The existing regulator has a 150 ohm series resistor (R15 on the board of the Muse) from the input power supply and looks like a 1 watt resistor. I didn’t change this. In a shunt configuration the 150 ohm resistor now dissipates about 0.7 watts from a 12 volt supply. This hasn’t been an issue.
The LM317 regulator is questionable in this application and appears well worthy of being targeted or replaced. The issue is that these regulators have a feedback loop that is far too slow to even attempt to respond to the switching speeds of the TDA1543. The transients of the TDA1543 imposed on the power supplies can be considered open loop in speed as equal to the digital switching speed. The LM317 has no hope.
Instead of changing the LM317 this chip was limited in responding to the transients. In this regard I adding both a capacitor from the output of the LM317 to its reference terminal and an inductor in series with the resistor connected from the reference terminal of the LM317 to ground. The capacitor reduces or shunts the changes in the difference voltages internal to the LM317 that would normally activate its feedback loop correction mechanism. The inductor also raises the impedance to ground that would normally increase the rate of charge or discharge of the added capacitor.
This deregulation of the LM317 can be thought to shift the sonic dependency from the LM317 to other capacitors connected from the power supply to ground. At the moment I am using a Muse cap on the power supply that is too large to fit in the box. The sonic weight and sense of three dimensional space is thought could also be a function of looser regulation that could be alternatively dealt with in Peter Daniels design. However, looser regulation might also increase of alternative issues. Adding a ceramic cap across the Muse dramatically reduced weight and space, a condition that suggests of greater influences with looser regulation.
I also use ferrite beads in conjunction with a “loose” coil on the output lines and in other areas of the circuit as well. For many of the coils I just tightly winding magnet wire on some cylindrical tool, bend the ends and use a scalpel to remove the varnish from the ends. In all cases the coils needed to be pulled apart for some unknown reason. It may be related to reducing the inter-winding capacitance from end to end of the coil. Ultimately inductance is difficult to work with in relations to sonic influences, too much inductance is as bad as too little.
I also replaced the AMS1117 3.3volt regulator with a high current LM334 shunt regulator set for around 3.3 volts. The existing regulator has a 150 ohm series resistor (R15 on the board of the Muse) from the input power supply and looks like a 1 watt resistor. I didn’t change this. In a shunt configuration the 150 ohm resistor now dissipates about 0.7 watts from a 12 volt supply. This hasn’t been an issue.
The LM317 regulator is questionable in this application and appears well worthy of being targeted or replaced. The issue is that these regulators have a feedback loop that is far too slow to even attempt to respond to the switching speeds of the TDA1543. The transients of the TDA1543 imposed on the power supplies can be considered open loop in speed as equal to the digital switching speed. The LM317 has no hope.
Instead of changing the LM317 this chip was limited in responding to the transients. In this regard I adding both a capacitor from the output of the LM317 to its reference terminal and an inductor in series with the resistor connected from the reference terminal of the LM317 to ground. The capacitor reduces or shunts the changes in the difference voltages internal to the LM317 that would normally activate its feedback loop correction mechanism. The inductor also raises the impedance to ground that would normally increase the rate of charge or discharge of the added capacitor.
This deregulation of the LM317 can be thought to shift the sonic dependency from the LM317 to other capacitors connected from the power supply to ground. At the moment I am using a Muse cap on the power supply that is too large to fit in the box. The sonic weight and sense of three dimensional space is thought could also be a function of looser regulation that could be alternatively dealt with in Peter Daniels design. However, looser regulation might also increase of alternative issues. Adding a ceramic cap across the Muse dramatically reduced weight and space, a condition that suggests of greater influences with looser regulation.
Removing SN75179 etc?
Rather than removing the 75179 chip and the 10nF caps can I just bypass/ jumper straight from the "in" and "ground" of the BNC connector to RXP/RXN of the CS8412 and leave the bypassed components in place?
Or do I need to actually remove 75179 and/or the AN8005 (U1) regulator?
Thanks
Rather than removing the 75179 chip and the 10nF caps can I just bypass/ jumper straight from the "in" and "ground" of the BNC connector to RXP/RXN of the CS8412 and leave the bypassed components in place?
Or do I need to actually remove 75179 and/or the AN8005 (U1) regulator?
Thanks
Last edited:
Sorry, should have reread the thread - back in April, 2011 Uncle_Leon wrote: "Get rid of the SN75179 and all the parts that feed it (U5, U1, C5a, C5b, L3). Remove the two film capacitors at Digital In (C6 and C7)".
Back to work!
Back to work!
Technically, you could just bypass the SN75179, but it would still be powered and so might have some minor influence on the sound via power supply or inductive/capacitative coupling. It's best to remove it altogether. It's actually a very easy job - since you are getting rid of it anyway, you can just cut its legs off or overheat the hell out of it.