Hi maxlorenz,
I attached both a timing and block diagram of a scrambler circuit. It has now been tested on both TDA1541A and TDA1543-based DACs, there is a significant increase in low level information, and much better L/R channel matching, even the DC-offset voltages are closely matched now.
This scrambler is already integrated on the new DI4T main board prototype, together with the highly effective tracker for source jitter blocking (replaces VCXO / PLL).
I used a separate 330R damping resistor for each TDA1543 clock input, this produced best (audible) results. The TDA1541A-S1 chips in my DI4T perform best with a 22R damping resistors on BCK
I attached both a timing and block diagram of a scrambler circuit. It has now been tested on both TDA1541A and TDA1543-based DACs, there is a significant increase in low level information, and much better L/R channel matching, even the DC-offset voltages are closely matched now.
This scrambler is already integrated on the new DI4T main board prototype, together with the highly effective tracker for source jitter blocking (replaces VCXO / PLL).
I used a separate 330R damping resistor for each TDA1543 clock input, this produced best (audible) results. The TDA1541A-S1 chips in my DI4T perform best with a 22R damping resistors on BCK
Attachments
New DI4T mainboard being tested.
Hi Zoran,
The constant current source and both diodes create a 1.4V reference voltage that (hopefully) tracks the DAC chip internal input diodes (3 in series, 2.1V). The 6K2 / 1K attenuator should produce approx. 800mVpp (+/- 400mV), this should be sufficient to drive the DAC WS and DATA inputs.
I don't advise using any attenuator on the BCK input as this will usually lead to increased jitter (verified with measurements), just use a damping resistor on this input.
I didn't use a DSP, it's just a tiny microchip PIC16F648A, some 0.1% precision SMD resistors for the 10-bit passive R2R DAC, and a discrete 2.8224 MHz VCXO. The whole bunch draws approx. 10mA @ 5V. I use the CS8416 FIFO buffer to compensate for slight phase / frequency differences that are allowed in the tracker.
The tracker uses precision frequency measurement and corrections, the micro controller basically needs one correction to match VCXO frequency to source frequency. Corrections are made with 2-second intervals.
I attached a picture of the new DI4 main board prototype. The chip closest to the insulated voltage regulator on the left is the micro controller for the tracker. The 10-bit passive R2R DAC is placed on the solder side.
I used resistive attenuators for DATA and WS signals, all digital outputs have damping resistors installed to minimize interference.
The new CS8416 receiver is placed close to the Toslink optical receiver and runs in slave-clock mode.
The small module in the center is the discrete VCXO with a supercap soldered directly to the power supply connections.
The 5 chips on the right are used for 4x interpolator, scrambler, and DEM / WS generation.
The UHS clock buffers are placed on the solder side.
Both GND and power supply connections are hard-wired on the solder side, this had to be done for achieving minimum crosstalk.
Hi Zoran,
The constant current source and both diodes create a 1.4V reference voltage that (hopefully) tracks the DAC chip internal input diodes (3 in series, 2.1V). The 6K2 / 1K attenuator should produce approx. 800mVpp (+/- 400mV), this should be sufficient to drive the DAC WS and DATA inputs.
I don't advise using any attenuator on the BCK input as this will usually lead to increased jitter (verified with measurements), just use a damping resistor on this input.
I didn't use a DSP, it's just a tiny microchip PIC16F648A, some 0.1% precision SMD resistors for the 10-bit passive R2R DAC, and a discrete 2.8224 MHz VCXO. The whole bunch draws approx. 10mA @ 5V. I use the CS8416 FIFO buffer to compensate for slight phase / frequency differences that are allowed in the tracker.
The tracker uses precision frequency measurement and corrections, the micro controller basically needs one correction to match VCXO frequency to source frequency. Corrections are made with 2-second intervals.
I attached a picture of the new DI4 main board prototype. The chip closest to the insulated voltage regulator on the left is the micro controller for the tracker. The 10-bit passive R2R DAC is placed on the solder side.
I used resistive attenuators for DATA and WS signals, all digital outputs have damping resistors installed to minimize interference.
The new CS8416 receiver is placed close to the Toslink optical receiver and runs in slave-clock mode.
The small module in the center is the discrete VCXO with a supercap soldered directly to the power supply connections.
The 5 chips on the right are used for 4x interpolator, scrambler, and DEM / WS generation.
The UHS clock buffers are placed on the solder side.
Both GND and power supply connections are hard-wired on the solder side, this had to be done for achieving minimum crosstalk.
Attachments
Hi Bernhard,
This is probably caused by TDA15xx on-chip tolerances between both L and R channels.
When simply paralleling these chips, it's possible that these typical errors are increased (correlated errors).
The scrambler circuit will prevent this form happening by swapping L/R channel data within the DAC chip. This is achieved by manipulating the I2S data for each DAC chip (group).
When paralleling 4 chips, I2S data would need to be as follows:
DAC #1, L/R (this is the standard sequence)
> L output = L output, R output = R output
DAC #2, R/L (channels reversed on the DAC chip)
> L output = R output, R output = L output
DAC #3, L/R
> L output = L output, R output = R output
DAC #4, R/L
> L output = R output, R output = L output
This is probably caused by TDA15xx on-chip tolerances between both L and R channels.
When simply paralleling these chips, it's possible that these typical errors are increased (correlated errors).
The scrambler circuit will prevent this form happening by swapping L/R channel data within the DAC chip. This is achieved by manipulating the I2S data for each DAC chip (group).
When paralleling 4 chips, I2S data would need to be as follows:
DAC #1, L/R (this is the standard sequence)
> L output = L output, R output = R output
DAC #2, R/L (channels reversed on the DAC chip)
> L output = R output, R output = L output
DAC #3, L/R
> L output = L output, R output = R output
DAC #4, R/L
> L output = R output, R output = L output
-ecdesigns- said:
Now you hope that those two typical errors ( left and right channel ) will cancel each other.
TDA needs too many bugfixes for my taste.
As there are:
DEM reclocking *
Input logic signal shaping *
1 mA output current offset *
Stereo DAC chip *
Inherent correlated LL errors, no MSB adjust **
Large footprint **
Bad output voltage compliance **
What else ?
* could fix
** can not fix
Bernhard said:
Even if one was able to fix all those issues, the TDA 1541 couldn't touch the AKM AK4396.
John I am convinced that with your expertise and empirical approach you will finally extract the last bit of performance out of
the TDA chips but they will still not be able to reach a decently implemented 4396.
I strongly suggest that anyone who wants to build something that
deserves the word "ultimate" give this Dac a try.
Until I discovered this chip I thought NOS was definitely the way to
go. After my lightly modified cheap Soundcard blew my DDDac with all those expensive caps and resistors out of the water I decided not to waste any time with any other chip anymore. Alex Peychev
of APL says that there is no other Dac on the planet that can touch
it. This is the guy that build players that are reportedly better than any Wadia.
I was so proud of the sound of my DDDac. It had the natural organic sound I was missing with other digital sources.
But when the soundcard with the passive out mod sounded even more natural I knew that oversampling by itself can not the problem. AKM seems to have found a way to overcome the
shortcomings of oversampling. What you get on top of the natural presentation is a wealth of detail and a 3D soundstage that is way above my DDDac.
John, I did not write this to criticize your work but to encourage you
to try something new.
Here is a quote from Alex Peychev about the superiority of the AK 4396:
"07-26-06: Aplhifi
....AKM is one of the largest suppliers of DACs and ADCs for the Professional and Broadcast quality products. There is no ADC like their AK5394A and no DAC like their AK4396, IMO
"I have AKM DACs" don't really mean anything. You really need to have the AK4396 because this DAC, being the AKM flagship, is much more advanced than any other AKM DAC. What they have done in the AK4396 is revolutionary in terms of Digital Filtering, Delta Sigma modulator and output stages. This DAC has up to 10 times less out-of-band noise compared to anything similar on the market today, as well as the capability to accept 216KHz sample rate while keeping the same digital filter oversampling rate and the same speed of the modulator.
I feel that Slim Devices made a very good choice selecting AK4396 DAC for their new "Transporter".
Note: The AK4396 DAC sounds lean when brand new. It gains bass, richness and fullness in about 200 hours, so be patient.
Regards, Alex"
Greets,
Klaus
The Bluegears Be-inspirer can be purchased for about $90 and it has
four 4396 Dacs on it.
I got five samples from a German distributor, but I really had to talk
the guy into it as they usually don't sell under a hundred chips.
At the moment I got a EMU 0404 USB with the passive out mod running and this is much better than my modded PCI card. The
combination with the asynchronous PCM transfer and the 4396 seems
to be a dream combination. This is still with the cheap 5V power
supply and no other mods then the mentioned passive out.
The next step will be paralleling four dacs as I got a run of boards
with two TeddyRegs one for the digital and one for the analog side
done.
To see what I did with th PCI card you can look at my picasa web album here:
http://picasaweb.google.de/Klaus.Freis/Audio
The last three pics describe the mod.
Sorry John for this off topic adventure.
Greets,
Klaus
four 4396 Dacs on it.
I got five samples from a German distributor, but I really had to talk
the guy into it as they usually don't sell under a hundred chips.
At the moment I got a EMU 0404 USB with the passive out mod running and this is much better than my modded PCI card. The
combination with the asynchronous PCM transfer and the 4396 seems
to be a dream combination. This is still with the cheap 5V power
supply and no other mods then the mentioned passive out.
The next step will be paralleling four dacs as I got a run of boards
with two TeddyRegs one for the digital and one for the analog side
done.
To see what I did with th PCI card you can look at my picasa web album here:
http://picasaweb.google.de/Klaus.Freis/Audio
The last three pics describe the mod.
Sorry John for this off topic adventure.
Greets,
Klaus
Bernhard said:
Coupling caps ?
Hey, don't pull a Bernhard on me
Coupling caps are history for me AK4396 2.5V DC bias
Tripath chips 2.5V DC bias
Thats one more advantage of my Tripath amps besides that they are
better than my Accuphase E303 or my AKSA 55n.
But this is getting even more OT, so I will shut up now.
Greets,
Klaus
Hi Klaus,
i tested Tripath amps like AMP3 (41Hz.com) or T-amp (high modded)...
and ended with that, i couldnt live with them. Not good sound to me.
Till now best is solid state (MOS-FET) from EC-Designs. It is modified of course: no input capacitor, no global feedback, only local feedback.
Natural, transparent as tube amps, no special coloring the sound.
Actualy i have now all components from John (EC-design).
Also (what i want to tell) DI4T DAC.
It is hard to believe, what could CD could sound good with DI4T.
No edginess, no sterile..., very analog sound, transparent, clean, LIVE.
Slowly i respect, how much time and effort give to improve DAC. As i contact him, i understand his goal is to make ultimate natural sounding DAC. I think he made is near maximum what is possible get from CD media.
It is different to be objective, but with listenig comparing, measuring, second opinion, could be another story.
My plan was to report impresions of complete EC-designs hifi system, but last month i have difficult with time (adaptation of house..).
I have plan to demonstrate to other audiophiles here and comparison to high prized Burmeister (30K euros) and other stuff. Then reporting.
Anyone who listed till know, was impressed by sound quality (actually we want music not sound), even tube lovers and analog lovers (EMTs, Thorens...).
Just want to share for now.
Best regards,
Bostjan
i tested Tripath amps like AMP3 (41Hz.com) or T-amp (high modded)...
and ended with that, i couldnt live with them. Not good sound to me.
Till now best is solid state (MOS-FET) from EC-Designs. It is modified of course: no input capacitor, no global feedback, only local feedback.
Natural, transparent as tube amps, no special coloring the sound.
Actualy i have now all components from John (EC-design).
Also (what i want to tell) DI4T DAC.
It is hard to believe, what could CD could sound good with DI4T.
No edginess, no sterile..., very analog sound, transparent, clean, LIVE.
Slowly i respect, how much time and effort give to improve DAC. As i contact him, i understand his goal is to make ultimate natural sounding DAC. I think he made is near maximum what is possible get from CD media.
It is different to be objective, but with listenig comparing, measuring, second opinion, could be another story.
My plan was to report impresions of complete EC-designs hifi system, but last month i have difficult with time (adaptation of house..).
I have plan to demonstrate to other audiophiles here and comparison to high prized Burmeister (30K euros) and other stuff. Then reporting.
Anyone who listed till know, was impressed by sound quality (actually we want music not sound), even tube lovers and analog lovers (EMTs, Thorens...).
Just want to share for now.
Best regards,
Bostjan
Attachments
a333bt said:
I understand that you are loyal to Johns designs and after all this
is his thread. But with all respect Bostjan, saying that Tripath does
not sound good, because you tried Amp3 and T-amp is like saying
you tried Tube amps and they don't sound good. The TK2050 chipset
is imho exceptional within the Tripath family. I don't make statements
about sound lightly. I got pleny of high quality amps to compare,
be it tube, solid state or digital.
The same is true for digital sources. I can switch them in a matter
of minutes and I can always check back with my Thorens TD520
turntable with Elac ESG-796 pickup, if I am in doubt. No TDA chip
or any other dac at the moment is a match for the AK4396 dac.
You can build something very good or even excellent with a TDA
1541 but when it comes to digital analog conversion, the word
"ultimate" is reserved for AKM at the moment.
I better keep silent now, before I get in trouble.
Greets,
Klaus
Hi Radian,
Some say that the Sabre DAC is superior .... some prefer selected TI multibit DACs ... I am pretty sure that the next new DAC chip will be "better" too.
You might google for switched-capacitor filters, and required buffer OP-amps, as these components are very likely part of the AK4396 DAC chip.
http://engr.nmsu.edu/~etti/3_3/3_3g.html
I am talking about poor quality on-chip sampling capacitors, electronic switches that suffer from capacitance modulation, and a buffer OP-amp with full feedback (using a OP-amp for unity gain is a bad idea). The buffer OP-amp is likely to have poorer performance than a OPA627, and it's running at only 5V power supply with minimal headroom (OPA627 runs on +/-15V).
I guess everyone has a favorite DAC chip, mine is the TDA1541A. But if you are happy with the AK4396 and a class-T amp, just use it and enjoy the music, as that's the most important thing after all.
I can't comment on AK4396 sound quality, because I didn't hear it perform in a highly optimized circuit yet. In general, I am not very impressed by delta-sigma DACs and digital amplifiers.
Direct comparison has to be done within a matter of seconds (switch) while carefully listening to familiar high-quality test tracks of the same source. Its also important to make sure the volume setting is exactly the same (measure with oscilloscope or suitable multimeter on AC position using a 1 KHz test signal). Using a pause of minutes is too long time period to reliably detect differences as the hearing "adapts" to the silence (pause) between switching sources.
You are also comparing different sources, a record and a CD that were recorded on different equipment.
I appreciate all advice given by diyAudio members, so no problem with me.
Some say that the Sabre DAC is superior .... some prefer selected TI multibit DACs ... I am pretty sure that the next new DAC chip will be "better" too.
You might google for switched-capacitor filters, and required buffer OP-amps, as these components are very likely part of the AK4396 DAC chip.
http://engr.nmsu.edu/~etti/3_3/3_3g.html
I am talking about poor quality on-chip sampling capacitors, electronic switches that suffer from capacitance modulation, and a buffer OP-amp with full feedback (using a OP-amp for unity gain is a bad idea). The buffer OP-amp is likely to have poorer performance than a OPA627, and it's running at only 5V power supply with minimal headroom (OPA627 runs on +/-15V).
I guess everyone has a favorite DAC chip, mine is the TDA1541A. But if you are happy with the AK4396 and a class-T amp, just use it and enjoy the music, as that's the most important thing after all.
I can't comment on AK4396 sound quality, because I didn't hear it perform in a highly optimized circuit yet. In general, I am not very impressed by delta-sigma DACs and digital amplifiers.
Direct comparison has to be done within a matter of seconds (switch) while carefully listening to familiar high-quality test tracks of the same source. Its also important to make sure the volume setting is exactly the same (measure with oscilloscope or suitable multimeter on AC position using a 1 KHz test signal). Using a pause of minutes is too long time period to reliably detect differences as the hearing "adapts" to the silence (pause) between switching sources.
You are also comparing different sources, a record and a CD that were recorded on different equipment.
I appreciate all advice given by diyAudio members, so no problem with me.