what about the ecdesigns version without any vref resistor ?
http://www.diyaudio.com/forums/digi...ding-ultimate-nos-dac-using-tda1541a-416.html
post 4152
maybe with 6 or 7V Vcc and some different bias?
http://www.diyaudio.com/forums/digi...ding-ultimate-nos-dac-using-tda1541a-416.html
post 4152
maybe with 6 or 7V Vcc and some different bias?
This is an interesting question. I checked the reference supply as being 2.08 volts with 2ma and 2.07 with 6ma. Hence between states there is 10mV. This is likely logarithmic in terms of current. In the off state it was about 2.09 volts. In other words it takes significant noise voltage to cause any noise current from the zero current state within the DAC's With a separate current source as being much higher than the off state current of the reference in wouldn't "normally" add any significant noise current to the output by being shorted out with an external capacitor. Of more interest, given the above measurements are the issues involved in using a single resistor for setting currents in all TDA1543's. 10mV change in a 2volt supply to the extent changing the current by 4mA suggests a low output impedance of about 2.5ohms. What this means is that reference voltage variations between DAC's as being connected together, having as little as 1 percent variation between them can cause considerable differences in the current setting of each DAC. One could be operating well above or below its normal operating point. This is not to suggest it wouldn't sound better with greater mismatch, however it may be better to control each DAC separately. In other words, you can't necessarily use a resistor 1/4 in value to produce equal currents in 4 DAC's.
In the circuit you provided as Building the ultimate NOS DAC using TDA1541A it shows a TDA1543. If the diagram is correct, P1 and P2 are adjustable to the positive and negative rails, hence you can set the potential anywhere you like, this is independent of the rail voltages.It isn't clear what the diagram means by setting the voltage to 0 Volts on P2 given the unknown values of R and L. This suggests setting the voltage to 0 volts across the unknown resistors R and L. The pots interact hence seems to require fiddling to set the voltage at 3.2Volts.
The one thing about this circuit is that unknown resistors R and L can be low and hence limit the swing of the output and hence improve distortions. The problem in doing so is that the reduction is activated by C3. C2 is also in play. Both exist as active in the audio signal path even though the voltage remains stationary across them.
On another note, I have been examining Peter Daniels differential TDA1543 design. This seems more intriguing and wonder if that design could be improved upon, this by using 4 chips, 2 for differential left and 2 for differential right with the opposite channel being disconnected in each chip.
There appears several advantages in using differential mode and using more chips.
1) A differential signal is more immune to power supply fluctuations and any common mode noise. The common mode noise signals are rejected by the differential.
2) The amplitude is doubled creating a better signal to noise ratio.
3) The center of a differential mode signal has no audio AC component, meaning that capacitors attached at the center of 2 identical resistors for some form of decoupling or filtering are not active to AC audio signals. In an ideal situation where no noise exists the capacitors are not in the audio signal path.
4) Having independent chips for every phase of both channels allows separate Vbias adjustments that eliminates the need for chip selection.
5) Perhaps the most important of all is that left and right digital data and clock signals can be shut off while the other channel is taking in data.
In NOS designs the numbers of digital edges the TDA1543 has to deal with is as low as it gets with no oversampling. IMO this can be one of the reasons why this chip works so well in such applications. It can be that the TDA1543 is accidentally good at rejecting such edges already. If true, then cutting the edges in half by switching off the unused left or right clock and data lines can further clear the background. Furthermore, by switching off the inputs to the off channel the ladder switches for the unused channels would never change. By eliminating as many digital signals as possible within the chips this could yield a substantial improvement in background.
Although all of the above may be true this may still require added circuits at the output to deal with the audio becoming too analytical.
The one thing about this circuit is that unknown resistors R and L can be low and hence limit the swing of the output and hence improve distortions. The problem in doing so is that the reduction is activated by C3. C2 is also in play. Both exist as active in the audio signal path even though the voltage remains stationary across them.
On another note, I have been examining Peter Daniels differential TDA1543 design. This seems more intriguing and wonder if that design could be improved upon, this by using 4 chips, 2 for differential left and 2 for differential right with the opposite channel being disconnected in each chip.
There appears several advantages in using differential mode and using more chips.
1) A differential signal is more immune to power supply fluctuations and any common mode noise. The common mode noise signals are rejected by the differential.
2) The amplitude is doubled creating a better signal to noise ratio.
3) The center of a differential mode signal has no audio AC component, meaning that capacitors attached at the center of 2 identical resistors for some form of decoupling or filtering are not active to AC audio signals. In an ideal situation where no noise exists the capacitors are not in the audio signal path.
4) Having independent chips for every phase of both channels allows separate Vbias adjustments that eliminates the need for chip selection.
5) Perhaps the most important of all is that left and right digital data and clock signals can be shut off while the other channel is taking in data.
In NOS designs the numbers of digital edges the TDA1543 has to deal with is as low as it gets with no oversampling. IMO this can be one of the reasons why this chip works so well in such applications. It can be that the TDA1543 is accidentally good at rejecting such edges already. If true, then cutting the edges in half by switching off the unused left or right clock and data lines can further clear the background. Furthermore, by switching off the inputs to the off channel the ladder switches for the unused channels would never change. By eliminating as many digital signals as possible within the chips this could yield a substantial improvement in background.
Although all of the above may be true this may still require added circuits at the output to deal with the audio becoming too analytical.
R and L are not resistors but simply the R and L output of the DAC.
The output ground is taken at a potentiometer P2 to null out the DC on the I/V resistors to avoid using coupling caps. this is meant by the set to 0V DC, it has nothing to do with the bias of the dac. You can ignore this part of the schema.
The trick in this circuit is to leave the Vref pin unused, and to center the DAC bias by putting the I/V resistors not to ground but to a potential (via P1) to feed some current to the DAC via the I/V resistors.
R4 and R3 are the 820ohm I/V resistors
what is important is the voltage across P1 (set to 3.2v for 820ohm I/V and 5V Vcc)
The output ground is taken at a potentiometer P2 to null out the DC on the I/V resistors to avoid using coupling caps. this is meant by the set to 0V DC, it has nothing to do with the bias of the dac. You can ignore this part of the schema.
The trick in this circuit is to leave the Vref pin unused, and to center the DAC bias by putting the I/V resistors not to ground but to a potential (via P1) to feed some current to the DAC via the I/V resistors.
R4 and R3 are the 820ohm I/V resistors
what is important is the voltage across P1 (set to 3.2v for 820ohm I/V and 5V Vcc)
Sure - The issue is that this circuit will not hold the differential output amplitude constant under the condition that a different binary input value is held, particularly if the load is a coupling transformer. The 470uF caps, the external differential output load, resistors and current sources can have an AC and DC component.
Ultimately if the circuit can't hold a square wave at ultra low frequencies (DC) it has an AC coupled circuit passing audio through these caps. This makes it dependent upon the quality of the caps. This is to state that although the zero amplitude of the signal is nulled it has different gains for different frequencies or amplitudes.
This is different in the circuit by Peter Daniels, his can create a square wave, though this doesn't necessarily make it any better or worse considering the variant components being used.
The other concern is if you took the caps out you would be mixing the right and left channels.
Ultimately if the circuit can't hold a square wave at ultra low frequencies (DC) it has an AC coupled circuit passing audio through these caps. This makes it dependent upon the quality of the caps. This is to state that although the zero amplitude of the signal is nulled it has different gains for different frequencies or amplitudes.
This is different in the circuit by Peter Daniels, his can create a square wave, though this doesn't necessarily make it any better or worse considering the variant components being used.
The other concern is if you took the caps out you would be mixing the right and left channels.
my idea was to replace the P1 and the 470mF C2 by 2 alkaline cells in series.
(1.5+1.5v)
Maybe 2 for L channel and 2 for R channel, to have the cleanest bias possible
And maybe to power the 1543 with a 6v alkaline too... in that case the I/V resistors could tap to the middle of the alkaline cell series at 3,0V
(1.5+1.5v)
Maybe 2 for L channel and 2 for R channel, to have the cleanest bias possible
And maybe to power the 1543 with a 6v alkaline too... in that case the I/V resistors could tap to the middle of the alkaline cell series at 3,0V
I don't have anything against batteries necessarily. My main issue is more of inconvenience to monitor and replace batteries. Despite this, the type of battery used can be an issue, even in meeting the criteria of a perfect theoretical device. These components are still dealing with AC signals and can cause the system to become analytical, or otherwise. This can become distracting from the overall sense of the music. It isn't that the battery is necessarily responsible, rather that the component you replaced could be responsible for having voiced the system in a good way, perhaps to compensate for other lesser devices.
With that said, the problem in doing what you suggest is that a quiescent sinking current exists for zero signal input, hence a quiescent voltage drop exists across R3 and R4. The potentiometer P2 is used to adjust for that quiescent voltage drop across these resistors to set both points to the same potential. In other words your differential will not be zero between the two points by using the midpoint of the batteries. This leaves potentiometers of unknown sonic virtue in the signal path.
With that said, the problem in doing what you suggest is that a quiescent sinking current exists for zero signal input, hence a quiescent voltage drop exists across R3 and R4. The potentiometer P2 is used to adjust for that quiescent voltage drop across these resistors to set both points to the same potential. In other words your differential will not be zero between the two points by using the midpoint of the batteries. This leaves potentiometers of unknown sonic virtue in the signal path.
in the schematic it has two TDA1543's powered by 8.3v http://www.diyaudio.com/forums/audio-sector/187748-pushing-limits-tda1543-nos-dac.html#post2550934 , afaik the datasheet states that the vmax is 8.0v, is this acccurate/wise/desired? if so for what reason.
Does this introduce desired distortion? is heatsinking required?
Does this introduce desired distortion? is heatsinking required?
Last edited:
Oh no...I was looking forward to getting an all AudioSector setup...
Should have bought one when I had the chance...
People under-estimate them. I've owned two non-USB, went away from digital, but now I miss the music I can't get on vinyl.
They were always better than the rest of my equipment. They are insanely musical with power conditioning (good, not transformers).
Peter was also kind enough to solder on the pesky SMD chip with his kits.
Last edited:
There is no need to use any surface-mount parts. CS8412 is available in a DIP package, which sounds the same as SMD; I believe Peter chose SMD simply for its compactness.
Onw in the updated, balanced version. Nope. I have NIchicon KG mounted currently. Bypass caps are Vishay MKP 1837. Output caps are Auricap. Everything else as spec'd