Hi everybody,
I hope someone can help me with my lack of digital knowlegde.
I'm looking for a method to make a balanced out DAC with 4 PCM1702 chips with the option for NOS/8xOS.
I've looked at the Passlabs D1 schematic and this link: http://www.geocities.com/yury_g/dac.htm
Can somebody explain me how it works and whether i can use something similar for my project.
Many greetings,
Thijs
I hope someone can help me with my lack of digital knowlegde.
I'm looking for a method to make a balanced out DAC with 4 PCM1702 chips with the option for NOS/8xOS.
I've looked at the Passlabs D1 schematic and this link: http://www.geocities.com/yury_g/dac.htm
Can somebody explain me how it works and whether i can use something similar for my project.
Many greetings,
Thijs
Balanced DAC
Hi Thijs,
Have a look at this thread.
http://db.audioasylum.com/cgi/m.pl?forum=tweaks&n=60815&highlight=elso+tda1541&r=&session=
Hi Thijs,
Have a look at this thread.
http://db.audioasylum.com/cgi/m.pl?forum=tweaks&n=60815&highlight=elso+tda1541&r=&session=
Thanks... nice link.. but
I spend the whole weekend up to now to read the TDA1541, TDA1543, DF1706, PCM1702 and CS8414 datasheets... and (nearly) all relevant threads on this forum and... it helps but I still don't understand everything...
Can I conclude that:
1] by simply inverting the TTL SDATA infront of a PCM1702 in effect inverts the analog out... this would give me a balanced DAC Like Passlabs D1? Why use four inverters?
2] Making a balanced pout from a TDA1541 is more difficult (see guido's thread on this forum)
3] to make the PCM1702 NonOverSampling could be done like in this link: http://www.geocities.com/yury_g/dac.htm if you use mode 5 for the CS8414
greetings,
Thijs
I spend the whole weekend up to now to read the TDA1541, TDA1543, DF1706, PCM1702 and CS8414 datasheets... and (nearly) all relevant threads on this forum and... it helps but I still don't understand everything...
Can I conclude that:
1] by simply inverting the TTL SDATA infront of a PCM1702 in effect inverts the analog out... this would give me a balanced DAC Like Passlabs D1? Why use four inverters?
2] Making a balanced pout from a TDA1541 is more difficult (see guido's thread on this forum)
3] to make the PCM1702 NonOverSampling could be done like in this link: http://www.geocities.com/yury_g/dac.htm if you use mode 5 for the CS8414
greetings,
Thijs
Balanced DAC
Hi Thijs,
Yes I admit it is a bit confusing. But schoolteacher as I am I wanted you first to read the thread on the Audio Asylum.
As Guido Tent pointed out for two's complements signal it works by inverting the DATA signal before entering the DAC to get a opposite phase in the analog domain. To be precise it works for all the bits except the least significant bit. This is documented in Horowitz (the Art of Electronics, book)page 476 in my second edition.
In the Pass DAC the inverting process is done with NOR or OR gates. I don't remember exactly. The elegancy of the scheme by Pass is no time difference beteeen the inverted and non-inverted digital signal. I am not sure it is important. I always used the inverted signal obtained by a 74HC04. No new developments on my front as I have given up the balanced DAC idea. Slightly smeared highs to my ears.
PCM1704 is a special case as you can, with the voltage at pin 10, arrange for inverted or non-inverted output.
TDA1541 is not difficult only use one chip for the normal signal (L& R) and one chip for the inverted signal (L & R). Of course this only works well if all dacs are equal. They are not in real world, that is a problem with these balanced DACs. I have been using two TDA1541AS1 in a balanced configuration.
PCM1702 on a CS8414 NON-OS mode 5?
I would use mode 6 (18 bits outputmode ) and add two zero's to get 20 bits format for this 20 bit DAC.
This can be done with two D-type flip-flops. I can dig up the schematic. If you don't insert "zero's" you will get sound output but the volume will be much too low.
Why am I doing this???
Hi Thijs,
Yes I admit it is a bit confusing. But schoolteacher as I am I wanted you first to read the thread on the Audio Asylum.
As Guido Tent pointed out for two's complements signal it works by inverting the DATA signal before entering the DAC to get a opposite phase in the analog domain. To be precise it works for all the bits except the least significant bit. This is documented in Horowitz (the Art of Electronics, book)page 476 in my second edition.
In the Pass DAC the inverting process is done with NOR or OR gates. I don't remember exactly. The elegancy of the scheme by Pass is no time difference beteeen the inverted and non-inverted digital signal. I am not sure it is important. I always used the inverted signal obtained by a 74HC04. No new developments on my front as I have given up the balanced DAC idea. Slightly smeared highs to my ears.
PCM1704 is a special case as you can, with the voltage at pin 10, arrange for inverted or non-inverted output.
TDA1541 is not difficult only use one chip for the normal signal (L& R) and one chip for the inverted signal (L & R). Of course this only works well if all dacs are equal. They are not in real world, that is a problem with these balanced DACs. I have been using two TDA1541AS1 in a balanced configuration.
PCM1702 on a CS8414 NON-OS mode 5?
I would use mode 6 (18 bits outputmode ) and add two zero's to get 20 bits format for this 20 bit DAC.
This can be done with two D-type flip-flops. I can dig up the schematic. If you don't insert "zero's" you will get sound output but the volume will be much too low.
Why am I doing this???
Great Thanks again!!
When I said: "I don't understand everting", you did realize that that was an understatement?
Inverting:
I've read about the time delay but I don't think it matters. In NOS mode every single bit delay equals something like 3mm propagation delay: as I never callibrate my speaker distances to my ears in millimeters, we don't have to worrie about that. In OS mode the delay is actually even less. Still I think that Waynes' idear (he desinged the digital part of the D ?) is elegant.
I didn't think about the low output when driving a PCM1702 directly from a CS8414. But I understand you lose a lot of dynamic range when only 16bits are used. So my NOS PCM1702 project is out of the window: enter DF1706 (got 3 samples from lovely TI, want one?)
The balanced idea still appeals to me. I have dowloaded the BurrBrown application bullitin mentioned in the Audio Asylum thread. Going to read it now.
Now don't dispair but ......
I have planned three projects:
1] a NOS DAC build of 32 paralel TDA1543 (1.5A current draw
); could be seperated a 16 parallel balanced NON DAC. This should give about 15dB improvement of THD and Noise specs in a perfect world.... can this be done? Inverting the I2S data?
2] a NOS DAC build of 8 or 9 parrallel TDA1541A. This should improve low level specs towards TDA1541S1 and S2 levels... maybe 4 parralel balanced out if possible
3] a DF1706/PCM1702 2 parralel/channel DAC or balanced... It depends if it is easy to make a balanced version of the DACs
All will get seperate power suplies, dedicated seperate inpout receivers with their dedicated power supplies: it's a long trem project...
I have build a CS3443, NOS TDA1543 and a NOS TDA1541 so far, but I lack real electrionic education. I have desinged a discrete I/V stage that should perform very well..
What would you think is the most easy project of those three to start with?
Many greetings,
Thijs
Thanks again!! When I said: "I don't understand everting", you did realize that that was an understatement?
Inverting:
I've read about the time delay but I don't think it matters. In NOS mode every single bit delay equals something like 3mm propagation delay: as I never callibrate my speaker distances to my ears in millimeters, we don't have to worrie about that. In OS mode the delay is actually even less. Still I think that Waynes' idear (he desinged the digital part of the D ?) is elegant.
I didn't think about the low output when driving a PCM1702 directly from a CS8414. But I understand you lose a lot of dynamic range when only 16bits are used. So my NOS PCM1702 project is out of the window: enter DF1706 (got 3 samples from lovely TI, want one?)
The balanced idea still appeals to me. I have dowloaded the BurrBrown application bullitin mentioned in the Audio Asylum thread. Going to read it now.
Now don't dispair but ......
I have planned three projects:
1] a NOS DAC build of 32 paralel TDA1543 (1.5A current draw
); could be seperated a 16 parallel balanced NON DAC. This should give about 15dB improvement of THD and Noise specs in a perfect world.... can this be done? Inverting the I2S data?2] a NOS DAC build of 8 or 9 parrallel TDA1541A. This should improve low level specs towards TDA1541S1 and S2 levels... maybe 4 parralel balanced out if possible
3] a DF1706/PCM1702 2 parralel/channel DAC or balanced... It depends if it is easy to make a balanced version of the DACs
All will get seperate power suplies, dedicated seperate inpout receivers with their dedicated power supplies: it's a long trem project...
I have build a CS3443, NOS TDA1543 and a NOS TDA1541 so far, but I lack real electrionic education. I have desinged a discrete I/V stage that should perform very well..
What would you think is the most easy project of those three to start with?
Many greetings,
Thijs
Goal?
Hi Thijs,
Before I answer to your questions, what is the goal or objective of your project(s)? Lower signal to noise ratio? Is this important ? My single TDA1543 is dead quiet. More bass? By means of balanced mode? Parallel DAC's . Which aspect of the sound will improve?
You see I can ask questions too!
Now for one:
1) 32 parallel TDA1543 or 16 parallel in balanced configuration. I don't see a immediate problem. Inverting I2S data or Sony/BB does not make any diiference. Both are two's complement. Only TDA1543A uses offset binary. CS8412 and CS8414 can only output two's complement.
2) same as above.
3) digital filter DF1706 and PCM1702. The balanced idea should also work here. Why not combine the DF1706 with the PCM 1704??
Project number one appeals most to me because I like the TDA1543 the most. Most simple?
Just a sidenote:
With a digital interface and digital receiver you never get it better than a one box player. I have made an effort to improve the interface leaving the silly AES/UBU standard. And also by using a Asynchronous Reclocker after the digital inputreceiver. And using Wildmonkeysects loopfilter, and low noise supply for the PLL section of the CS8412. But the one box solution clearly wins.
Hi Thijs,
Before I answer to your questions, what is the goal or objective of your project(s)? Lower signal to noise ratio? Is this important ? My single TDA1543 is dead quiet. More bass? By means of balanced mode? Parallel DAC's . Which aspect of the sound will improve?
You see I can ask questions too!
Now for one:
1) 32 parallel TDA1543 or 16 parallel in balanced configuration. I don't see a immediate problem. Inverting I2S data or Sony/BB does not make any diiference. Both are two's complement. Only TDA1543A uses offset binary. CS8412 and CS8414 can only output two's complement.
2) same as above.
3) digital filter DF1706 and PCM1702. The balanced idea should also work here. Why not combine the DF1706 with the PCM 1704??
Project number one appeals most to me because I like the TDA1543 the most. Most simple?
Just a sidenote:
With a digital interface and digital receiver you never get it better than a one box player. I have made an effort to improve the interface leaving the silly AES/UBU standard. And also by using a Asynchronous Reclocker after the digital inputreceiver. And using Wildmonkeysects loopfilter, and low noise supply for the PLL section of the CS8412. But the one box solution clearly wins.
Yeah... I know... it's all a bit ridiculus.. It's more the fun of building the stuff than actually taking the time to review the results.. I'm not saying these project will indeed give supperior perfomance.. but it looks great, all those chips on a euro board, in a nice alluminium CONRAD case....
.. my TDA1543 by the way is not dead-quiet .. it has significant background noise (it is the simple DAC kit with passive I/V) ... I'll let you know how thing progress, thanks for all the help..
greetings,
Thijs
.. my TDA1543 by the way is not dead-quiet .. it has significant background noise (it is the simple DAC kit with passive I/V) ... I'll let you know how thing progress, thanks for all the help..
greetings,
Thijs
Thijs, Elso,
Guido in the AA thread is not G Tent. It is the other Guido...
Btw R2D2 stands next to my monitor.
As for a differential TDA1541. You probably know i got that working. Not by just feeding the inverted data signal to one, but by splitting the i2s in a left and right i2s signal. Read the 2*TDA1541 threat from a while back. This could be used for other DAC's too.
As explained in the cd7 threat below, i don't have much time at the moment... Elso, i am interested in the output you used for the diff TDA1541. Can you share some info?
Greetings,
Guido B
Guido in the AA thread is not G Tent. It is the other Guido...
Btw R2D2 stands next to my monitor.
As for a differential TDA1541. You probably know i got that working. Not by just feeding the inverted data signal to one, but by splitting the i2s in a left and right i2s signal. Read the 2*TDA1541 threat from a while back. This could be used for other DAC's too.
As explained in the cd7 threat below, i don't have much time at the moment... Elso, i am interested in the output you used for the diff TDA1541. Can you share some info?
Greetings,
Guido B
Hi Thijs, can you explain why you want to parallel 32 TDA1543's ?
Only for noise reasons ? My DAC is quiet too with only one TDA. Only problem is that loads have influence on soundquality. A buffer at the output will make the output-signal independent of the load within some borders.
Please realize a CS8412 will not be capable to "drive" 32 chips, you will probably have to buffer things and my guess is that these extra electronics will waste the effect of the paralled DAC chips.
Extremity for the extremity will not automatically lead to good results but you can try of course.
Only for noise reasons ? My DAC is quiet too with only one TDA. Only problem is that loads have influence on soundquality. A buffer at the output will make the output-signal independent of the load within some borders.
Please realize a CS8412 will not be capable to "drive" 32 chips, you will probably have to buffer things and my guess is that these extra electronics will waste the effect of the paralled DAC chips.
Extremity for the extremity will not automatically lead to good results but you can try of course.
Hi Guido B (alias R2D2?) and Jean-Paul,
Thanks for the repies. I just read the BB aplication note concerning binairy codes and I finally think I might understand it one beautiful day... somewhere in the future...
The 2*TDA1541 thread has been inspected last week also. Looks like a complicated project, very good that you got it working nicely. Happy listening!
I realized that 32 inputs of 10k (?) and 15pF thats.. ehhhhh ..312 Ohm and 480pF won't be driven very happely by a CS8412.. so buffers are in order.. This might introduce even more jitter, but I hope it will be randomly distributed along the 32 DACs .. I do hope for some results worth listening to, but it doesn't need to be better than my cheapy Philips CD723.. it a fun project
G'night,
Thijs
Thanks for the repies. I just read the BB aplication note concerning binairy codes and I finally think I might understand it one beautiful day... somewhere in the future...
The 2*TDA1541 thread has been inspected last week also. Looks like a complicated project, very good that you got it working nicely. Happy listening!
I realized that 32 inputs of 10k (?) and 15pF thats.. ehhhhh ..312 Ohm and 480pF won't be driven very happely by a CS8412.. so buffers are in order.. This might introduce even more jitter, but I hope it will be randomly distributed along the 32 DACs .. I do hope for some results worth listening to, but it doesn't need to be better than my cheapy Philips CD723.. it a fun project
G'night,
Thijs
Balanced DAC
Hi All,
Guido (R2D2), sorry for confusing you with Guido Tent. It happened before!
For balanced operation I made a scheme with three discrete opamps. Two opamps are for IV-conversion and one is for output buffering and balanced to single ended conversion. I also incorporated a third order Bessel low-pass filter. -3dB point at 10kHz. The first pole is in the IV-converter. I can post the general schematic but will not post details of the discrete opamp.
Thijs, There was a famous Japanese DAC using 16(?) parallel Burr-Brown PCM56. I don't remember the name or brand. A long time ago I had the PCM56 in a balanced configuration in my Sony player. So I had four PCM56 in the working. The PCM56 had very good bass but poorly defined highs. Later I changed to AD1851 that had much better highs but weaker bass.
Hi All,
Guido (R2D2), sorry for confusing you with Guido Tent. It happened before!
For balanced operation I made a scheme with three discrete opamps. Two opamps are for IV-conversion and one is for output buffering and balanced to single ended conversion. I also incorporated a third order Bessel low-pass filter. -3dB point at 10kHz. The first pole is in the IV-converter. I can post the general schematic but will not post details of the discrete opamp.
Thijs, There was a famous Japanese DAC using 16(?) parallel Burr-Brown PCM56. I don't remember the name or brand. A long time ago I had the PCM56 in a balanced configuration in my Sony player. So I had four PCM56 in the working. The PCM56 had very good bass but poorly defined highs. Later I changed to AD1851 that had much better highs but weaker bass.
Balanced IV-converter
Hi Guido B,
Attached the schematic of a balanced IV-converter and outputbuffer. I got the idea from a Crystal datasheet. The whole filter is calculated with Burr-Browns FilterPro(R). A nice side effect is that any offset caused by the DACs is cancelled as long as the offset is equal for both DACs. This eliminates the ouput cap.
I similated both inputs with Microcap. It simulated good and measurements confirm that. No overshoot at a 1kHz sqaurewave.
As a experiment I changed the filterresponse to Butterworth and a 15kHz -3dB point. The squarewave showed the slight overshoot exactly as in the FilterPro papers. Sound was worse more "digital" so I quickly changed it back to Bessel filter response. This result withheld me from trying higher order Butterworth filters. Kusunoki states that the bad sound is not due to oversampling per se but due to the digital filter. I don't agree as you will understand.
http://www.tnt-audio.com/intervis/kusunoki_e.html
Hi Guido B,
Attached the schematic of a balanced IV-converter and outputbuffer. I got the idea from a Crystal datasheet. The whole filter is calculated with Burr-Browns FilterPro(R). A nice side effect is that any offset caused by the DACs is cancelled as long as the offset is equal for both DACs. This eliminates the ouput cap.
I similated both inputs with Microcap. It simulated good and measurements confirm that. No overshoot at a 1kHz sqaurewave.
As a experiment I changed the filterresponse to Butterworth and a 15kHz -3dB point. The squarewave showed the slight overshoot exactly as in the FilterPro papers. Sound was worse more "digital" so I quickly changed it back to Bessel filter response. This result withheld me from trying higher order Butterworth filters. Kusunoki states that the bad sound is not due to oversampling per se but due to the digital filter. I don't agree as you will understand.
http://www.tnt-audio.com/intervis/kusunoki_e.html
-3dB Point at10kHz
It is all about choices!
I want a smooth 3150 sine wave.
I can put the -3dB point at 20kHz. This will not remove the staircase on the sinewave.
It does not sound dull as I feared.
NON-OS sounds very bright to my ears. Maybe this explains a bit.
I have been critisized a lot for the low crossover point. That's why I tried the 15kHz filter. It is a dilemma we will never solve in a NON-OS DAC. If you want a steep, brickwall filter you get ringing and overshoot and bad sound. If you choose a filtertype without ringing like the Bessel characteristic you don't get a steep filter. Especially the soft knee or shoulder of the Bessel filter can not be avoided. A higher order Bessel filter at a higher crossover will not solve the problem. I can not bend the laws of physics. Steep filtering is not possible without ringing. If you have a scheme that accomplishes steep filtering without ringing I would be most interested. As we all know the bandwith of the redbook CD is choosen too low. This was due to the constricts at the time of introduction of the CD.
I have also experimented with notch filters like in the Zanden DAC. Notch filters give even more ringing and overshoot. The result is a very weird square wave.
Feel free to use a higher crossoverpoint f.a. at 20kHz. It will not remove the staircase signal, only round it off a little.
Hi Jean-Paul,jean-paul said:
It is all about choices!
I want a smooth 3150 sine wave.
I can put the -3dB point at 20kHz. This will not remove the staircase on the sinewave.
It does not sound dull as I feared.
NON-OS sounds very bright to my ears. Maybe this explains a bit.
I have been critisized a lot for the low crossover point. That's why I tried the 15kHz filter. It is a dilemma we will never solve in a NON-OS DAC. If you want a steep, brickwall filter you get ringing and overshoot and bad sound. If you choose a filtertype without ringing like the Bessel characteristic you don't get a steep filter. Especially the soft knee or shoulder of the Bessel filter can not be avoided. A higher order Bessel filter at a higher crossover will not solve the problem. I can not bend the laws of physics. Steep filtering is not possible without ringing. If you have a scheme that accomplishes steep filtering without ringing I would be most interested. As we all know the bandwith of the redbook CD is choosen too low. This was due to the constricts at the time of introduction of the CD.
I have also experimented with notch filters like in the Zanden DAC. Notch filters give even more ringing and overshoot. The result is a very weird square wave.
Feel free to use a higher crossoverpoint f.a. at 20kHz. It will not remove the staircase signal, only round it off a little.
Re: famous dac...
Hi t,
You mean this one?
They did not even bother to mention the type of the DAC chip.
http://www.audionote.co.jp/digital/content.htm
tbla said:the dac you talk about, is the HIBARI from audio note japan.......
Hi t,
You mean this one?
They did not even bother to mention the type of the DAC chip.
http://www.audionote.co.jp/digital/content.htm
Thijs,
If the CS8412 gives jitter and you buffer it with one buffer, you will feed your 32 dacs with the same signal. Then they all suffer from the jitter. Maybe if you use separate buffers to feed the dacs you might have different timing at the dac input. But if it gives an improvement?
There are three options to reduce jitter i have seen in projects on the net:
one box solution: use an old player to tap i2s. You need an old player, newer models have the oversampling done inside the decoder. I took a 1994 philips apart the other day, this already has this (SAA7345&TDA1545, no i2s anyway). If you go for this, you need to upgrade the cdp's clock to reduce jitter.
async overclocking, like Elso uses.
reclocking in dac with fs and feedback of the clock signal to the cd player. The way i am building my dac. It is not ready yet, so no impressions yet on sound. Not really an universal solution i must admit.
use a pll to create a new stable clock, which follows the incoming clock.
Greetings,
GuidoB
If the CS8412 gives jitter and you buffer it with one buffer, you will feed your 32 dacs with the same signal. Then they all suffer from the jitter. Maybe if you use separate buffers to feed the dacs you might have different timing at the dac input. But if it gives an improvement?
There are three options to reduce jitter i have seen in projects on the net:
one box solution: use an old player to tap i2s. You need an old player, newer models have the oversampling done inside the decoder. I took a 1994 philips apart the other day, this already has this (SAA7345&TDA1545, no i2s anyway). If you go for this, you need to upgrade the cdp's clock to reduce jitter.
async overclocking, like Elso uses.
reclocking in dac with fs and feedback of the clock signal to the cd player. The way i am building my dac. It is not ready yet, so no impressions yet on sound. Not really an universal solution i must admit.
use a pll to create a new stable clock, which follows the incoming clock.
Greetings,
GuidoB
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.