Sonic differences of DI DACs
Hi protos,
Thanks for your reply [post #1239]
> Ok, this is very difficult, but I will try to describe differences in perceived sound quality between both DI DACs, mainly based on reactions of the people who already listened to both DACs.
In general, when compared to standard (oversampled) DACs, DI DACs respond very fast, provide extreme detail and phase accuracy, the trebles are smooth and crystal clear, echo's of voices and instruments slowly fade into complete silence. The midrange is warm and sounds realistic / natural. Bass reproduction is accurate and tight. In general the DI DAC's reveal much more information about the sonic properties of musical instruments and voices. Also information about acoustic space is reproduced very accurately, this results in a incredible sound depth.
The perceived difference is huge, it's almost like listening to a live performance. These improvements are the result of many experiments, listening tests, and systematically modifying the design for optimal results.
Differences between the 4 DI DAC configurations:
Basically the DI 8 and DI 8M have the clearest sound due to lower distortion of the TDA1541A used, the DI 8M has a more transparent and slightly warmer sound.
The DI 8 main properties are clarity and neutrality
The DI 8M main properties are warmth, clarity, neutrality and transparency
The DI 16 and DI 16M sound very smooth and forgiving, due to the higher resolution (16X interpolation), the midrange sounds even warmer than the DI 8 / DI 8M. Similar to the DI 8M, the DI 16M has a more transparent and slightly warmer sound.
The DI 16 main properties are warmth and smoothness.
The DI 16M main properties are warmth, smoothness and transparency
So it's a matter of taste, some people like slightly smoother warmer sound, others prefer slightly clearer and more neutral sound. Tubes add transparency and more warmth, making the sound very natural. The DI 8M is my personal favorite.
> Yes there have been many modifications since the start of this thread. As soon as I got time, I will make a detailed description of the DI system. I plan to publish it on my website.
Hi protos,
Thanks for your reply [post #1239]
> Ok, this is very difficult, but I will try to describe differences in perceived sound quality between both DI DACs, mainly based on reactions of the people who already listened to both DACs.
In general, when compared to standard (oversampled) DACs, DI DACs respond very fast, provide extreme detail and phase accuracy, the trebles are smooth and crystal clear, echo's of voices and instruments slowly fade into complete silence. The midrange is warm and sounds realistic / natural. Bass reproduction is accurate and tight. In general the DI DAC's reveal much more information about the sonic properties of musical instruments and voices. Also information about acoustic space is reproduced very accurately, this results in a incredible sound depth.
The perceived difference is huge, it's almost like listening to a live performance. These improvements are the result of many experiments, listening tests, and systematically modifying the design for optimal results.
Differences between the 4 DI DAC configurations:
Basically the DI 8 and DI 8M have the clearest sound due to lower distortion of the TDA1541A used, the DI 8M has a more transparent and slightly warmer sound.
The DI 8 main properties are clarity and neutrality
The DI 8M main properties are warmth, clarity, neutrality and transparency
The DI 16 and DI 16M sound very smooth and forgiving, due to the higher resolution (16X interpolation), the midrange sounds even warmer than the DI 8 / DI 8M. Similar to the DI 8M, the DI 16M has a more transparent and slightly warmer sound.
The DI 16 main properties are warmth and smoothness.
The DI 16M main properties are warmth, smoothness and transparency
So it's a matter of taste, some people like slightly smoother warmer sound, others prefer slightly clearer and more neutral sound. Tubes add transparency and more warmth, making the sound very natural. The DI 8M is my personal favorite.
> Yes there have been many modifications since the start of this thread. As soon as I got time, I will make a detailed description of the DI system. I plan to publish it on my website.
Hi EC 
Confess that you have a complete staff of EE behind you
(I got exhausted simply from reading all your detailed work )
So, no direct 48BCK capability for the moment...In the future I will ask for help to DIY such timing chain. (yeah, I'm stuborn...I need to listen by myself )
Keep on the good work, man.
M
Confess that you have a complete staff of EE behind you
(I got exhausted simply from reading all your detailed work
)So, no direct 48BCK capability for the moment...In the future I will ask for help to DIY such timing chain. (yeah, I'm stuborn...I need to listen by myself
)Keep on the good work, man.
M
Hi Terry Demol,
thanks for your reply [post#1243]
The shiftregister reclocker (part of the USB/DI2S interface) is clocked by a separate 48 MHz crystal oscillator and is located inside the DI DAC close to the DAC chips.
The external universal I2S interface box has both, a synchronous and a shiftregister reclocker. One of these can be selected depending on the application.
- The universal I2S interface (synchronous reclocker option) uses the transport's master clock (11.28, 16.93 or 33.86 MHz) for reclocking. Reclocking is synchronous here because both BCK and the transport's master clock run in sync, now BCK clock sampling takes place when BCK has settled completely (not on the continuously phase shifting BCK transients). This requires placing the universal I2S interface relatively close to the transport.
- The universal I2S interface (shiftregister option) uses a separate 48MHz or 56.488MHz crystal oscillator. Since no transport's master clock is needed, the universal I2S interface can now be placed close to the DAC.
- When using the DI DAC DI2S input, there is no clock / reclocking. The DI2S input acts as a straight forward input to the DAC chips.
The universal I2S interface is basically a small box with both a coax and Toslink SPDIF input, a differential master clock input (RS422), and a differential (DI2S) output. It can be powered by the transport or by a adapter.
thanks for your reply [post#1243]
The shiftregister reclocker (part of the USB/DI2S interface) is clocked by a separate 48 MHz crystal oscillator and is located inside the DI DAC close to the DAC chips.
The external universal I2S interface box has both, a synchronous and a shiftregister reclocker. One of these can be selected depending on the application.
- The universal I2S interface (synchronous reclocker option) uses the transport's master clock (11.28, 16.93 or 33.86 MHz) for reclocking. Reclocking is synchronous here because both BCK and the transport's master clock run in sync, now BCK clock sampling takes place when BCK has settled completely (not on the continuously phase shifting BCK transients). This requires placing the universal I2S interface relatively close to the transport.
- The universal I2S interface (shiftregister option) uses a separate 48MHz or 56.488MHz crystal oscillator. Since no transport's master clock is needed, the universal I2S interface can now be placed close to the DAC.
- When using the DI DAC DI2S input, there is no clock / reclocking. The DI2S input acts as a straight forward input to the DAC chips.
The universal I2S interface is basically a small box with both a coax and Toslink SPDIF input, a differential master clock input (RS422), and a differential (DI2S) output. It can be powered by the transport or by a adapter.
Shiftregister reclocker
Hi anbello,
The shiftregister reclocker is different from the Kusunoki reclocker:
1) It uses the right reclock frequency:
48.000MHz / 2.8224MHz = 17.006 (well almost)
56.488MHz / 2.8224MHz = 20
Multiples of the exact BCK frequency are used (17 and 20).
The Kusunoki reclocker uses reclock frequencies that are not a multiple of the BCK frequency. This results in the typical phase shifted squarewave shape.
2) The shiftregister reclocker is basically a synchronous digital one-shot circuit, the reclocked BCK is tapped from the shiftregister's second D flip-flop output. The digital one-shot locks two subsequent transients by resetting the shiftregister just before the negative going edge of BCK, also resetting the second D flip-flop output to zero (reclock output).
The kusunoki reclocker doesn't.
Hi anbello,
The shiftregister reclocker is different from the Kusunoki reclocker:
1) It uses the right reclock frequency:
48.000MHz / 2.8224MHz = 17.006 (well almost)
56.488MHz / 2.8224MHz = 20
Multiples of the exact BCK frequency are used (17 and 20).
The Kusunoki reclocker uses reclock frequencies that are not a multiple of the BCK frequency. This results in the typical phase shifted squarewave shape.
2) The shiftregister reclocker is basically a synchronous digital one-shot circuit, the reclocked BCK is tapped from the shiftregister's second D flip-flop output. The digital one-shot locks two subsequent transients by resetting the shiftregister just before the negative going edge of BCK, also resetting the second D flip-flop output to zero (reclock output).
The kusunoki reclocker doesn't.
Hi John
I want to add a good (diy) spdif cable between DVD player and surround receiver. Now i use a 1 Eu 49 cable from the Action!
You make your cables yourself, with twisted pairs. Do you have some tips for me to twist my own cable, or will Oehlbach NF113 do the job better?
Oehlbach uses twisted pairs also:
http://www.oehlbachkabel.de/other/nl/oehlbach_physik.php
Thanks
I want to add a good (diy) spdif cable between DVD player and surround receiver. Now i use a 1 Eu 49 cable from the Action!
You make your cables yourself, with twisted pairs. Do you have some tips for me to twist my own cable, or will Oehlbach NF113 do the job better?
Oehlbach uses twisted pairs also:
http://www.oehlbachkabel.de/other/nl/oehlbach_physik.php
Thanks
DI DAC progress
Hi kstlfido,
Thanks for your compliments [post #1249]
Sorry for the late reply, but I have been very busy as usual.
If you think these designs look amazing, you should hear the DI 8 now after the latest modifications, and fixing a major fault. In the hurry I managed to mix-up the tubes in the tube diff amp
. Instead of using the ECC83 I fitted a ECC81, and instead of the ECC 81 a ECC 82 was mounted. This was the reason I continued optimizing the DI 8, and waited with ordering the PCBs, not knowing the swapped tubes were causing the problems.
This fault must have been in the DI 8 since assembly. Well after putting the correct tubes back in, it became clear I had overdone it quite a bit with modifications. The DI 8 sound is unbelievable now, it even puts the Octal DI DAC completely in the shadow.
So now both the DI 8 and DI 16 designs are final, and the PCB layout data can be prepared for PCB manufacturing.
Hi kstlfido,
Thanks for your compliments [post #1249]
Sorry for the late reply, but I have been very busy as usual.
If you think these designs look amazing, you should hear the DI 8 now after the latest modifications, and fixing a major fault. In the hurry I managed to mix-up the tubes in the tube diff amp
. Instead of using the ECC83 I fitted a ECC81, and instead of the ECC 81 a ECC 82 was mounted. This was the reason I continued optimizing the DI 8, and waited with ordering the PCBs, not knowing the swapped tubes were causing the problems.This fault must have been in the DI 8 since assembly. Well after putting the correct tubes back in, it became clear I had overdone it quite a bit with modifications. The DI 8 sound is unbelievable now, it even puts the Octal DI DAC completely in the shadow.
So now both the DI 8 and DI 16 designs are final, and the PCB layout data can be prepared for PCB manufacturing.
SPFIF interlink
Hi tubee,
Thanks for your reply [post #1250]
After all experimenting I have done with interfaces, I can only tell you that even the best SPDIF interlink will only bring marginal improvements since most audio interface receiver clock generators are specified at approx. 200ps jitter. This is already way too high, even when using a perfect interlink with optimal loading.
I used a differential SPDIF interlink during experimenting. I suppose you have a coax connection on your DVD player, this has to be converted to a differential signal first in order to use a twisted pair interlink. So it's not very practical.
Hi tubee,
Thanks for your reply [post #1250]
After all experimenting I have done with interfaces, I can only tell you that even the best SPDIF interlink will only bring marginal improvements since most audio interface receiver clock generators are specified at approx. 200ps jitter. This is already way too high, even when using a perfect interlink with optimal loading.
I used a differential SPDIF interlink during experimenting. I suppose you have a coax connection on your DVD player, this has to be converted to a differential signal first in order to use a twisted pair interlink. So it's not very practical.
Ok thanks John for your comments.
Yes the dvd p and Marantz amp uses coax.
Will do some clock mods on them (dvdp plays SACD & DVD A/V) but will problably not upgrade the Spdif to differential, too much work. Better invest this in other projects. Have been searching around: for coax Belden 1505A looks very good, simple and cheap. Thinking to put a BNC socket in DVD/Receiver.
But still: the 1505A is RG59 (so 75ohm), and the cheapo Action cable: RG59!
Yes the dvd p and Marantz amp uses coax.
Will do some clock mods on them (dvdp plays SACD & DVD A/V) but will problably not upgrade the Spdif to differential, too much work. Better invest this in other projects. Have been searching around: for coax Belden 1505A looks very good, simple and cheap. Thinking to put a BNC socket in DVD/Receiver.
But still: the 1505A is RG59 (so 75ohm), and the cheapo Action cable: RG59!
D-I 16 schematic
Hello
maybe I missed something in this long thread - but I could not find the schematic of the D-I 16 analog part. See post #1111.
How are the 16 TDA1543 used ? 2,4 or 8 times analog interpolating. 2 or more in parallel ..? Output stage with OPA627 - even differential used ?
Anyhow I build a DDDAC which uses 12 TDA1543 in parallel with a simple resistor as I/V stage. The sound is good but the high freq. are too hard for my taste - and I have the strong feeling that filtering is necessary but the approach here is more elegant.
thanks
awesome work from ecdesign !
gebi
Hello
maybe I missed something in this long thread - but I could not find the schematic of the D-I 16 analog part. See post #1111.
How are the 16 TDA1543 used ? 2,4 or 8 times analog interpolating. 2 or more in parallel ..? Output stage with OPA627 - even differential used ?
Anyhow I build a DDDAC which uses 12 TDA1543 in parallel with a simple resistor as I/V stage. The sound is good but the high freq. are too hard for my taste - and I have the strong feeling that filtering is necessary but the approach here is more elegant.
thanks
awesome work from ecdesign !
gebi
DI 16
Hi gebi,
Thanks for the compliments [post #1254]
I will post the final DI 16 schematics and silkscreen plot soon.
The DI 16, provides 16X interpolation, there are 2 groups of 8 X TDA1543 DAC chips with the outputs in parallel. There are two I/V converters (LM4562)/ channel R=470 Ohm, C=100pF.
One group is inverted, the other is non-inverted. This enables the use of a differential output stage (LM4562) that's DC coupled (no coupling capacitor necessary). Each individual TDA1543 reference pin is connected to the corresponding I/V OP-amp non-inverting input trough a 4K7 resistor), the non-inverting inputs of the I/V OP-amps are decoupled to analog ground. The TDA1543 25mV ac output compliance is met, ensuring low linearity errors.
I use LM4562 exclusively now, it makes the OPA627 sound muddy in comparison. Due to the fact that the DI system already provides a wonderful warm midrange, it matches the slightly "thinner" sound of the LM4562 perfectly, resulting in a very well balanced sound.
NOS DAC's provide some problems with higher frequencies due to the low resolution among other things (you are basically listening to square wave-like signals at higher frequencies). But also the lower frequencies suffer from lack of resolution. Just an example with sine waves:
500 Hz NOS DAC: 88 samples, DI8: 706 samples, DI 16: 1411 samples
1 KHz NOS DAC: 44 samples, DI8: 353 samples, DI 16: 706 samples
5 KHz NOS DAC: 9 samples, DI8: 70 samples, DI 16: 140 samples
10 KHz NOS DAC: 4 samples, DI8: 32 samples, DI 16: 64 samples
Because the DI system interpolates both, timing and amplitude, the analog resolution also increases proportional to the timing resolution. The following examples are theoretical:
NOS DAC: 65536 possible analog levels 16 bit resolution
DI 8 DAC: 524288 possible analog levels 19 bit resolution
DI 16 DAC: 1048576 possible analog levels 20 bit resolution
This results in the following theoretical resolution enhancement:
NOS DAC: 1X
DI 8 DAC: timing resolution (8X) * analog resolution (8X) = 64X
DI 16 DAC: timing resolution (16X) * analog resolution (16X) = 256X
The DI system enhances resolution (8X or 16X interpolation), due to the technique used, there will be a natural HF roll-off. Therefore no corrective filters are necessary. This is a big advantage as phase linearity errors can be kept very low this way. The DI system can be expanded to 32X or 64X interpolation if desired.
Note that a common digital filter with 8X interpolation and 16 bit output only increases resolution by 8 times. A DAC with digital filter and 20 bit output increases resolution by 32 times, that's only half of the resolution of the DI 8!
Your "strong feeling" about the need for filtering a NOS DAC output is correct, in fact that was the first modification I made to my old twin DAC, adding a 8th order Bessel filter. But this resulted in loss of detail, laid-back sound and phase linearity errors.
Often coupling caps need to be used (DC offset) so these capacitors will have a effect on sound quality too (like dddac mentioned), but the effect is mostly sound coloring. Auricap are among some of the best (most neutral sounding) coupling capacitors. Coupling caps can be used to "shape" sound quality, make the sound warmer or more transparent, however they don't correct a low resolution signal.
Yesterday a musician (plays guitar in a band) came over to have his guitar effect amplifier repaired. Next he listened to the DI 8. I selected a guitar solo (Sara k. and Chris Jones live in concert / All my love), he was simply stunned by the DI 8 DAC performance. Judging by his reaction the produced sound had to be pretty close to how a guitar actually sounds. He just wondered about a strange noise he heard in the recording, I was already worrying that I missed something, tuning the DI DACs.
Turned out he actually heard the mechanical thump of the piano pedals, something most DACs won't reveal this clearly.
Hi gebi,
Thanks for the compliments [post #1254]
I will post the final DI 16 schematics and silkscreen plot soon.
The DI 16, provides 16X interpolation, there are 2 groups of 8 X TDA1543 DAC chips with the outputs in parallel. There are two I/V converters (LM4562)/ channel R=470 Ohm, C=100pF.
One group is inverted, the other is non-inverted. This enables the use of a differential output stage (LM4562) that's DC coupled (no coupling capacitor necessary). Each individual TDA1543 reference pin is connected to the corresponding I/V OP-amp non-inverting input trough a 4K7 resistor), the non-inverting inputs of the I/V OP-amps are decoupled to analog ground. The TDA1543 25mV ac output compliance is met, ensuring low linearity errors.
I use LM4562 exclusively now, it makes the OPA627 sound muddy in comparison. Due to the fact that the DI system already provides a wonderful warm midrange, it matches the slightly "thinner" sound of the LM4562 perfectly, resulting in a very well balanced sound.
NOS DAC's provide some problems with higher frequencies due to the low resolution among other things (you are basically listening to square wave-like signals at higher frequencies). But also the lower frequencies suffer from lack of resolution. Just an example with sine waves:
500 Hz NOS DAC: 88 samples, DI8: 706 samples, DI 16: 1411 samples
1 KHz NOS DAC: 44 samples, DI8: 353 samples, DI 16: 706 samples
5 KHz NOS DAC: 9 samples, DI8: 70 samples, DI 16: 140 samples
10 KHz NOS DAC: 4 samples, DI8: 32 samples, DI 16: 64 samples
Because the DI system interpolates both, timing and amplitude, the analog resolution also increases proportional to the timing resolution. The following examples are theoretical:
NOS DAC: 65536 possible analog levels 16 bit resolution
DI 8 DAC: 524288 possible analog levels 19 bit resolution
DI 16 DAC: 1048576 possible analog levels 20 bit resolution
This results in the following theoretical resolution enhancement:
NOS DAC: 1X
DI 8 DAC: timing resolution (8X) * analog resolution (8X) = 64X
DI 16 DAC: timing resolution (16X) * analog resolution (16X) = 256X
The DI system enhances resolution (8X or 16X interpolation), due to the technique used, there will be a natural HF roll-off. Therefore no corrective filters are necessary. This is a big advantage as phase linearity errors can be kept very low this way. The DI system can be expanded to 32X or 64X interpolation if desired.
Note that a common digital filter with 8X interpolation and 16 bit output only increases resolution by 8 times. A DAC with digital filter and 20 bit output increases resolution by 32 times, that's only half of the resolution of the DI 8!
Your "strong feeling" about the need for filtering a NOS DAC output is correct, in fact that was the first modification I made to my old twin DAC, adding a 8th order Bessel filter. But this resulted in loss of detail, laid-back sound and phase linearity errors.
Often coupling caps need to be used (DC offset) so these capacitors will have a effect on sound quality too (like dddac mentioned), but the effect is mostly sound coloring. Auricap are among some of the best (most neutral sounding) coupling capacitors. Coupling caps can be used to "shape" sound quality, make the sound warmer or more transparent, however they don't correct a low resolution signal.
Yesterday a musician (plays guitar in a band) came over to have his guitar effect amplifier repaired. Next he listened to the DI 8. I selected a guitar solo (Sara k. and Chris Jones live in concert / All my love), he was simply stunned by the DI 8 DAC performance. Judging by his reaction the produced sound had to be pretty close to how a guitar actually sounds. He just wondered about a strange noise he heard in the recording, I was already worrying that I missed something, tuning the DI DACs.
Turned out he actually heard the mechanical thump of the piano pedals, something most DACs won't reveal this clearly.
DDDAC caps
Hi Doede
nice to hear from you.
Not to be missunderstood - your DAC sounds very good - clearly better than my marantz cd63. but still there is room for improvements.
the outputcaps I'm using are Intertechnik MKP-Q4 10u - nothing special but good
quality.
I have a another question to you - sorry not directly related to this threat -
V_ref is around 2V and the DC audio out is 4V. is this a problem ? in the datasahet a opv is used and therefore both voltages are (should be ?) be in the same range ?
sincerely
Hi Doede
nice to hear from you.
Not to be missunderstood - your DAC sounds very good - clearly better than my marantz cd63. but still there is room for improvements.
the outputcaps I'm using are Intertechnik MKP-Q4 10u - nothing special but good
quality.
I have a another question to you - sorry not directly related to this threat -
V_ref is around 2V and the DC audio out is 4V. is this a problem ? in the datasahet a opv is used and therefore both voltages are (should be ?) be in the same range ?
sincerely
di -16 notes
thanks ecdesign for your clear comments.
As the schematics available I'll build one or order the pcb stuff from you. -))
I read a lot of about how our ear is working. Very interresting was the disscusion about the HTF (head transfer function). In conclusion the phase of the signal is (just for stereo and not in daily live) more important as the absolute freq./amplitude for the localization.
Another note: my cat is always running out of the door listening to my NOS-DOC
sincerely
gebi
thanks ecdesign for your clear comments.
As the schematics available I'll build one or order the pcb stuff from you. -))
I read a lot of about how our ear is working. Very interresting was the disscusion about the HTF (head transfer function). In conclusion the phase of the signal is (just for stereo and not in daily live) more important as the absolute freq./amplitude for the localization.
Another note: my cat is always running out of the door listening to my NOS-DOC
sincerely
gebi
Dutch-> American connection
Right at this moment i am trying to connect I2S to two pcm56 dacs.
Want to try and hear this and find out if harmonic distorsion can be lowered (MSB adjust) compared to 1541.
I2S is from SAA7210, clock provided by a P2P version Kwak 7.
Found logic here on diyaudio. I/V is Pedja's AD844 stage.
The problem is i have digital noise and music, have to check glue logic.
Or clock in must be not 2.8mHz (11.28 / 4)
I don't understand why data has to be shifted 11 clockcycles.
Right at this moment i am trying to connect I2S to two pcm56 dacs.
Want to try and hear this and find out if harmonic distorsion can be lowered (MSB adjust) compared to 1541.
I2S is from SAA7210, clock provided by a P2P version Kwak 7.
Found logic here on diyaudio. I/V is Pedja's AD844 stage.
The problem is i have digital noise and music, have to check glue logic.
Or clock in must be not 2.8mHz (11.28 / 4)
I don't understand why data has to be shifted 11 clockcycles.
