Hello
I found an interesting article about upsampling for a Dac.
http://www.mlssa.com/pdf/Upsampling-theory-rev-2.pdf
Btw, it's not directly related but maby this interesting part of that Upsampling-theory text are partly explain why Non-oversampling Dac sound good ?
Here's the text extract from the Upsampling-theory ;
"The ultrasonic image signal essentially frustrates any systematic interaction between the DAC’s differential non-linearity errors and the slow-changing baseband audio signal, which would otherwise create audible non-linear distortion. Instead, the interaction of the ultrasonic dither with the DAC’s differential non-linearity simply increases the noise floor of the DAC somewhat."
Bye
Gaétan
I found an interesting article about upsampling for a Dac.
http://www.mlssa.com/pdf/Upsampling-theory-rev-2.pdf
Btw, it's not directly related but maby this interesting part of that Upsampling-theory text are partly explain why Non-oversampling Dac sound good ?
Here's the text extract from the Upsampling-theory ;
"The ultrasonic image signal essentially frustrates any systematic interaction between the DAC’s differential non-linearity errors and the slow-changing baseband audio signal, which would otherwise create audible non-linear distortion. Instead, the interaction of the ultrasonic dither with the DAC’s differential non-linearity simply increases the noise floor of the DAC somewhat."
Bye
Gaétan
Hi luvdunhill,
DI DACs need Philips format I2S 64 BCK / WS in order for the timing module to interpolate correctly, if your I2S source has a different format, conversion is needed. Easiest way is the use of a SPDIF audio receiver, followed by a shiftregister reclocker (SPDIFDI2S module).
DI DACs interpolate, so they create higher initial resolution when compared to NOS DACs. Interpolation is divided over two DAC groups, one inverted, one non-inverted group. The balanced design reduces common errors, improves CMMR, and most important, provides a DC coupled output (no added distortion by coupling cap or transformer). Two I/V converters / channel are needed, these converters need to have excellent tracking (same gain / linearity), accuracy needs to be at least 0.1% or better. The DI 8M uses 0.01% bulk metal foil precision resistors to achieve optimal tracking. If gain / linearity differs in both I/V stages, additional bit errors are introduced.
The interpolation results in a slow rolloff filter that attenuates HF images enough to prevent problems in the rest of the audio equipment. Attenuation is sufficient to leave-out the analog filter, enabling excellent phase accuracy.
I can't guarantee optimal DI DAC performance when different I/V circuits are used. The current LM4562 class A I/V converters with bulk metal foil feedback resistors are extremely accurate, high linearity, very low THD, low DC offset voltage, and fast response. This is very difficult to achieve with any kind of discrete circuit, especially when two virtually identical I/V stages are needed.
If a balanced output is required, the diff amp can be left out, and both I/V converter outputs can be connected to a XLR output. The following amplifier with balanced input will then replace the diff amp stage in the DI DAC. However, when that amplifier isn't accurate enough (amplitude / linearity difference between inverted and non-inverted input), bit errors are introduced, degrading sound quality.
I use higher amplitude output signals (2V RMS), and short semi-balanced Litz wire interlinks. My MPA80 power amps have a gain of only 20x. This combination also produces higher resolution at low volume settings. Noise / hum levels are extremely low / inaudible, even when listening a few inches away from the speaker chassis with no signal applied. XLR interlinks are useful when using longer interlinks, but these longer interlinks could also degrade sound quality (capacitance). Another advantage of the RCA outputs is that the diff amp is located very close to the I/V converters.
Noise / hum is often generated by low-cost rectifier bridges. I received some HEXFRED diodes (fast / slow recovery) from onnosr, these completely eliminated the last bits of noise / hum introduced by the power amplifiers. I already used schottky diodes in the DI DAC (11DQ10). Both Schottky and fast-slow recovery diodes prevent switching noise and peak currents introduced by the diode itself during polarity change.
The DI 16 can be built for a reasonable price, that's also one of the reasons I designed it. It has the typical DI DAC properties, lot's of detail, analog natural sound and high phase accuracy. The only bottle-neck is the higher distortion of the TDA1543 chips, still it managed to outperform my twin TDA1541A DAC with 8th order Butterworth filter.
The Modified DI 8M is the top performer, building it is more expensive, but performance is on a whole different level, direct comparisons with high-performance 24 bit /192 KHz upsampling CD transports, and SACD players showed that the DI 8M revealed much more detail and sounded much clearer. The DI 8M has a very neutral and transparent sound, and fast impulse response.
The added tube diff amp reduces the effects of thermal memory, and the possible problems with the high feedback factor in the unity-gain LM4562 diff amp. Only the diff amp is a hybrid amp (OP-amp + tube OP-amp in parallel). The LM4562 diff amp delivers approx. 2/3 of the output signal now, both noise and distortion (THD) is much lower than a tube-only diff amp / output stage. Since the DI DACs have very high internal signal amplitude, a resistive output attenuator is needed, DI 8M output impedance is 600 Ohm typical.
DI DACs need Philips format I2S 64 BCK / WS in order for the timing module to interpolate correctly, if your I2S source has a different format, conversion is needed. Easiest way is the use of a SPDIF audio receiver, followed by a shiftregister reclocker (SPDIFDI2S module).
DI DACs interpolate, so they create higher initial resolution when compared to NOS DACs. Interpolation is divided over two DAC groups, one inverted, one non-inverted group. The balanced design reduces common errors, improves CMMR, and most important, provides a DC coupled output (no added distortion by coupling cap or transformer). Two I/V converters / channel are needed, these converters need to have excellent tracking (same gain / linearity), accuracy needs to be at least 0.1% or better. The DI 8M uses 0.01% bulk metal foil precision resistors to achieve optimal tracking. If gain / linearity differs in both I/V stages, additional bit errors are introduced.
The interpolation results in a slow rolloff filter that attenuates HF images enough to prevent problems in the rest of the audio equipment. Attenuation is sufficient to leave-out the analog filter, enabling excellent phase accuracy.
I can't guarantee optimal DI DAC performance when different I/V circuits are used. The current LM4562 class A I/V converters with bulk metal foil feedback resistors are extremely accurate, high linearity, very low THD, low DC offset voltage, and fast response. This is very difficult to achieve with any kind of discrete circuit, especially when two virtually identical I/V stages are needed.
If a balanced output is required, the diff amp can be left out, and both I/V converter outputs can be connected to a XLR output. The following amplifier with balanced input will then replace the diff amp stage in the DI DAC. However, when that amplifier isn't accurate enough (amplitude / linearity difference between inverted and non-inverted input), bit errors are introduced, degrading sound quality.
I use higher amplitude output signals (2V RMS), and short semi-balanced Litz wire interlinks. My MPA80 power amps have a gain of only 20x. This combination also produces higher resolution at low volume settings. Noise / hum levels are extremely low / inaudible, even when listening a few inches away from the speaker chassis with no signal applied. XLR interlinks are useful when using longer interlinks, but these longer interlinks could also degrade sound quality (capacitance). Another advantage of the RCA outputs is that the diff amp is located very close to the I/V converters.
Noise / hum is often generated by low-cost rectifier bridges. I received some HEXFRED diodes (fast / slow recovery) from onnosr, these completely eliminated the last bits of noise / hum introduced by the power amplifiers. I already used schottky diodes in the DI DAC (11DQ10). Both Schottky and fast-slow recovery diodes prevent switching noise and peak currents introduced by the diode itself during polarity change.
The DI 16 can be built for a reasonable price, that's also one of the reasons I designed it. It has the typical DI DAC properties, lot's of detail, analog natural sound and high phase accuracy. The only bottle-neck is the higher distortion of the TDA1543 chips, still it managed to outperform my twin TDA1541A DAC with 8th order Butterworth filter.
The Modified DI 8M is the top performer, building it is more expensive, but performance is on a whole different level, direct comparisons with high-performance 24 bit /192 KHz upsampling CD transports, and SACD players showed that the DI 8M revealed much more detail and sounded much clearer. The DI 8M has a very neutral and transparent sound, and fast impulse response.
The added tube diff amp reduces the effects of thermal memory, and the possible problems with the high feedback factor in the unity-gain LM4562 diff amp. Only the diff amp is a hybrid amp (OP-amp + tube OP-amp in parallel). The LM4562 diff amp delivers approx. 2/3 of the output signal now, both noise and distortion (THD) is much lower than a tube-only diff amp / output stage. Since the DI DACs have very high internal signal amplitude, a resistive output attenuator is needed, DI 8M output impedance is 600 Ohm typical.
- Status
- Not open for further replies.