because some people claim that SPDIF is the root of all evil.
Most people don't know how to design decent SPDIF circuits either. Driving any kind of cable length with I2s will be just as hard as doing SPDIF well. I have head my Scott Nixon Non OS DAC trounce a very tweaked CD player with a low noise clock circuit. The CD player wasn't even close. So much for blaming everything on the SPDIF interface. This seems to be another one of those audio urban legends.............
Most people don't know how to design decent SPDIF circuits either. Driving any kind of cable length with I2s will be just as hard as doing SPDIF well. I have head my Scott Nixon Non OS DAC trounce a very tweaked CD player with a low noise clock circuit. The CD player wasn't even close. So much for blaming everything on the SPDIF interface. This seems to be another one of those audio urban legends.............
just transmit the I2S signal directly with modern signalling techniques.
Knock your self out.......... I am sure that's enough imformation for Peter to go implement the interface, right? Oh and be sure to do it at logic levels for lots of RFI. There is nother to designing low noise digital interfaces anybody should be able to figure it out. Oh and god forbide you should use an interface that is standard enough to use with different transports and DACs that would make too much sense.
Knock your self out.......... I am sure that's enough imformation for Peter to go implement the interface, right? Oh and be sure to do it at logic levels for lots of RFI. There is nother to designing low noise digital interfaces anybody should be able to figure it out. Oh and god forbide you should use an interface that is standard enough to use with different transports and DACs that would make too much sense.
It must be that all the heavy metals in the ground water down there have turned you into an abusive *****, but I won't let that excuse you.
My scheme for transmitting I2S over an LVDS link has been discussed here already. It is simpler and cheaper than the Crystal S/PDIF receivers, and it uses industry standard ICs, cables, and connectors.
NOTE: The asterisks(*) denote that this word has been censored. IN this specific case, the meaning has been determined to be vulgar, and the statement malicous. Refer to SinBin. -CHRIS8
My scheme for transmitting I2S over an LVDS link has been discussed here already. It is simpler and cheaper than the Crystal S/PDIF receivers, and it uses industry standard ICs, cables, and connectors.
NOTE: The asterisks(*) denote that this word has been censored. IN this specific case, the meaning has been determined to be vulgar, and the statement malicous. Refer to SinBin. -CHRIS8
Hey......I drink the same water! Is that why I am a horse's butt, too?
I think Phred just wants to get me stirred up because he knows how mad I am today. And why.
Go for it, buddy. You have my e-mail.......we'll work out a better solution. Anything to get rid of those Crystal chips and stinking RCA connectors is worth all the effort it takes.
Bye, Phred. Say "hi" to the chunky girls at the pizza place for me.
Jocko
I think Phred just wants to get me stirred up because he knows how mad I am today. And why.
Go for it, buddy. You have my e-mail.......we'll work out a better solution. Anything to get rid of those Crystal chips and stinking RCA connectors is worth all the effort it takes.
Bye, Phred. Say "hi" to the chunky girls at the pizza place for me.
Jocko
Different Types of Interpolation
While both use FIR filters to perform a type of interpolation, the AD1896 and the DF1704 have significantly different functions.
The DF1704 performs 8X oversampling. The primary goal of this is to move the noise spectrum coming out of the DAC to a higher frequency. This allows the analog low-pass reconstruction filter to have it's knee way above the audio spectrum and to have a shallower slope. This in turn helps keep the filter from interfering with the phase response of the audio spectrum.
The DF1704 is a synchronous device. It uses the same master clock for both the input and the output. It doesn't have any inherent jitter reduction. It takes a 96KHz stereo bitstream (2-32bit frames) and produces two 768KHz mono (24 bit frame) bitstreams. If fed a 44.1KHz stereo bitstream, it outputs two 352.8KHz mono bitstreams.
The AD1896 performs 'curve fitting' between PCM data points. This allows it to create 'smoother' data than the original, with the possibility of increased output sampling frequency and word size. In a way, it's analogous to taking the data into a DAC feeding an ADC to produce new data. The important difference is that this is done completely in the digital domain instead of using the analog domain. This keeps the process immune from noise and timing irregularities. (There's a very good explanation of the interpolation and resampling process in the AD1896 datasheet.)
The AD1896 is an asynchronous device. The input clock is independent of the output clock. There is a ratio estimation circuit that constantly compares the frequencies of the two clocks. A side benefit of this device is that the output is reclocked. This gives significant jitter reduction. If fed 44.1KHz 16 bit data, it can produce 96KHz 24bit data. If fed 96KHz 24bit data, the output would still be 96KHz 24bit, but it would be reclocked with reduced jitter. (It could also reduce the data, but in your case it wouldn't be desirable.)
So if you want to use the PCM1704, it would be best to use either the DF1704 or DF1706 oversampling filter. (You also might want to consider isolators between the filter and the DACs to keep the analog section quieter, as suggested in the application notes.)
The AD1896 ASRC does wonders for improving sound quality (IMO), especially with 16bit 44.1KHz CDs. If you want to use it with the PCM1704, it needs to have the DF1704 in between.
I bought some from Pioneer. $19.60 in single quantities. They quoted me an eight week lead time, but they arrived in four. Their website shows that they currently have these in stock.
******
Regarding the DIR1703:
In PLL mode (which is what I assume you would want), the data clock is generated by a PLL synched to the incoming S/PDIF stream. It is completely independent of the crystal or oscillator feeding the DIR1703, which is used to clock other internal circuits.
Regards,
Brian.
jwb said:
While both use FIR filters to perform a type of interpolation, the AD1896 and the DF1704 have significantly different functions.
The DF1704 performs 8X oversampling. The primary goal of this is to move the noise spectrum coming out of the DAC to a higher frequency. This allows the analog low-pass reconstruction filter to have it's knee way above the audio spectrum and to have a shallower slope. This in turn helps keep the filter from interfering with the phase response of the audio spectrum.
The DF1704 is a synchronous device. It uses the same master clock for both the input and the output. It doesn't have any inherent jitter reduction. It takes a 96KHz stereo bitstream (2-32bit frames) and produces two 768KHz mono (24 bit frame) bitstreams. If fed a 44.1KHz stereo bitstream, it outputs two 352.8KHz mono bitstreams.
The AD1896 performs 'curve fitting' between PCM data points. This allows it to create 'smoother' data than the original, with the possibility of increased output sampling frequency and word size. In a way, it's analogous to taking the data into a DAC feeding an ADC to produce new data. The important difference is that this is done completely in the digital domain instead of using the analog domain. This keeps the process immune from noise and timing irregularities. (There's a very good explanation of the interpolation and resampling process in the AD1896 datasheet.)
The AD1896 is an asynchronous device. The input clock is independent of the output clock. There is a ratio estimation circuit that constantly compares the frequencies of the two clocks. A side benefit of this device is that the output is reclocked. This gives significant jitter reduction. If fed 44.1KHz 16 bit data, it can produce 96KHz 24bit data. If fed 96KHz 24bit data, the output would still be 96KHz 24bit, but it would be reclocked with reduced jitter. (It could also reduce the data, but in your case it wouldn't be desirable.)
So if you want to use the PCM1704, it would be best to use either the DF1704 or DF1706 oversampling filter. (You also might want to consider isolators between the filter and the DACs to keep the analog section quieter, as suggested in the application notes.)
The AD1896 ASRC does wonders for improving sound quality (IMO), especially with 16bit 44.1KHz CDs. If you want to use it with the PCM1704, it needs to have the DF1704 in between.
I bought some from Pioneer. $19.60 in single quantities. They quoted me an eight week lead time, but they arrived in four. Their website shows that they currently have these in stock.
******
Regarding the DIR1703:
In PLL mode (which is what I assume you would want), the data clock is generated by a PLL synched to the incoming S/PDIF stream. It is completely independent of the crystal or oscillator feeding the DIR1703, which is used to clock other internal circuits.
Regards,
Brian.
As I wrote in an other thread, I am planning a similar DAC, with 4 PCM1704 for balanced output with Passlabs D1 output stage. Either with CS8420 receiver/Async Sample Rate converter, and DF1704 upsampler, or with DIR1703 receiver, AD1896 ASRC and DF1706 upsampler.
I put some drawings here:
http://www.neunelfer.de/jpg/PCM1704-Output.jpg
http://www.neunelfer.de/jpg/CS8420-Input.jpg
http://www.neunelfer.de/jpg/DIR1703-Input.jpg
I have trouble understanding the DIR1703 (and in partikular the AD19896) data sheets, regarding clock, so I took up the CS8420 as an option again. But the AD part seems to be of higher quality. If I get some help or review on the schematics (its my first DAC), this should be a nice project to learn from for many on this forum,and give a nice sounding DAC also. I will do a layout on the design, trying to keep it as compact as possible, using SMT parts for resistors and small caps. Should not get separate regulatores for each chip, only good decoupling with caps mounted near all chips. TI suggest even to place the SMC caps on the top side of the board not to get the additional inductance from the vias. This is one reason to go for SMD caps.
(not for output signal coupling, of course).
I would do the output stage with Mosfets, but maybe we could get footprints for both on the PCB?
And even put both (DIR1703/AD1896 and CS8240) input stages on borad, jumper selectable, so someone could compare, and others could place only the parts of the desired version on the PCB.
If there is interest in this, I could order some more PCBs then to keep the costs down for all.
jwb: Do You plan to build Your design? Or is it only meant as a proposal? You seem ta have a lot of experience with DACs, and I would like to get some comments on my design from You.
Maybe there is enough interest on this project on the forum, should get running if we can get at least 10 pc PCB ordered. And some review is needed before making the PCBs, should work without prototyping first.
Peter
I put some drawings here:
http://www.neunelfer.de/jpg/PCM1704-Output.jpg
http://www.neunelfer.de/jpg/CS8420-Input.jpg
http://www.neunelfer.de/jpg/DIR1703-Input.jpg
I have trouble understanding the DIR1703 (and in partikular the AD19896) data sheets, regarding clock, so I took up the CS8420 as an option again. But the AD part seems to be of higher quality. If I get some help or review on the schematics (its my first DAC), this should be a nice project to learn from for many on this forum,and give a nice sounding DAC also. I will do a layout on the design, trying to keep it as compact as possible, using SMT parts for resistors and small caps. Should not get separate regulatores for each chip, only good decoupling with caps mounted near all chips. TI suggest even to place the SMC caps on the top side of the board not to get the additional inductance from the vias. This is one reason to go for SMD caps.
(not for output signal coupling, of course).
I would do the output stage with Mosfets, but maybe we could get footprints for both on the PCB?
And even put both (DIR1703/AD1896 and CS8240) input stages on borad, jumper selectable, so someone could compare, and others could place only the parts of the desired version on the PCB.
If there is interest in this, I could order some more PCBs then to keep the costs down for all.
jwb: Do You plan to build Your design? Or is it only meant as a proposal? You seem ta have a lot of experience with DACs, and I would like to get some comments on my design from You.
Maybe there is enough interest on this project on the forum, should get running if we can get at least 10 pc PCB ordered. And some review is needed before making the PCBs, should work without prototyping first.
Peter
Hi,
> To make sure this DAC is actually different from the previous
> efforts, I will not use an asynchronous sample rate convereter.
Use the AD1896 but make sure it operates SYNCRONOUS with an option to switch to asyncronous operation and to have a 1:2 and 1:4 ratio between I/O frequency. You can generate the appropriate FSYNC frequency via a divider from the MCK of the receiver. This way you have a much shorter digital filter than most commercial availabel ones with more flexibility, the option to use no oversampling or 2 times or 4 times oversampling etc....
The next step is simple. If you use enough PCM1704 in parallel (say 16pcs) and differential (so 32 pcs PCM1704 per channel) you can use a passive I/V conversion with an output transformer to get the outputlevel up to the right level.
With 32 pcs PCM 1704 per channel you get a reasonably low output impedance with this scheme and a much improved dynamic range (15db more dynamic range) and with no active stage to muck up this you should be able to make the DAC with widest dynamic range/highest resolution for audio on the planet, with > 125db dynamic range guranteed by design.
I have been thinking or something like this for a while but have neither time nor money to progress this (getting hitched)....
BTW, consider a RAM buffer in the DAC between the input and the rest. In such a case if the buffer is made long enough (like holding a minimum of three samples) you can make your own basic digital filter in the same go. I looked at that but so far think the AD1896 would be an excellent alternative with loads of flexibility.
Sayonara
> To make sure this DAC is actually different from the previous
> efforts, I will not use an asynchronous sample rate convereter.
Use the AD1896 but make sure it operates SYNCRONOUS with an option to switch to asyncronous operation and to have a 1:2 and 1:4 ratio between I/O frequency. You can generate the appropriate FSYNC frequency via a divider from the MCK of the receiver. This way you have a much shorter digital filter than most commercial availabel ones with more flexibility, the option to use no oversampling or 2 times or 4 times oversampling etc....
The next step is simple. If you use enough PCM1704 in parallel (say 16pcs) and differential (so 32 pcs PCM1704 per channel) you can use a passive I/V conversion with an output transformer to get the outputlevel up to the right level.
With 32 pcs PCM 1704 per channel you get a reasonably low output impedance with this scheme and a much improved dynamic range (15db more dynamic range) and with no active stage to muck up this you should be able to make the DAC with widest dynamic range/highest resolution for audio on the planet, with > 125db dynamic range guranteed by design.
I have been thinking or something like this for a while but have neither time nor money to progress this (getting hitched)....
BTW, consider a RAM buffer in the DAC between the input and the rest. In such a case if the buffer is made long enough (like holding a minimum of three samples) you can make your own basic digital filter in the same go. I looked at that but so far think the AD1896 would be an excellent alternative with loads of flexibility.
Sayonara
have turned you into an abusive *****
Ah nothing like reasoned discourse from my esteemed college...... Yes you can an I2S interface with three cable interfaces and the resultant termation nd RFI issues. Or like I said..... you can do one well done SPDIF interface and have the flexibility of using different DACs and or transports interchangably as many developers, DIYers, and just about every one else in the industry does. There is also the potential to use readily available pulse transformers and analog diffential interfaces for reduction of RFI and common mode noise on the interface. Also advantages in decoupling noise from the transport from getting into the analog circuits through the DAC dgital inputs.
I guess it is easier to rant with hysterics about the SPDIF interface than to go do the work of improving it. My experience and that of a many well reguarded DAC designers is that this interface has plenty of potential and is not ready to be thrown out yet. The Audio Alchemy and Muse digital interfaces seem to have dissappeared. The I2S cable that I built for the AA stuff worked best in very short lengths(15") due to the lack of proper termination by the way.
http://www.diyvideo.com/forums/showthread.php?s=&postid=90854#post90854
Fred
P.S. It is nice to know that the forum has such hands off moderators so I can be called an "abusive *****" for discussing technical issues that I have three years of professional experience in. Makes me really want to be a contributer here.
Post edited(word has been censored), as result of offender's(not user Fred Dieckmann) usage of it with malicious intent has been censored. Your complaint has been heard, and action has been taken as result. Please refer to SinBin for result. -CHRIS8
Ah nothing like reasoned discourse from my esteemed college...... Yes you can an I2S interface with three cable interfaces and the resultant termation nd RFI issues. Or like I said..... you can do one well done SPDIF interface and have the flexibility of using different DACs and or transports interchangably as many developers, DIYers, and just about every one else in the industry does. There is also the potential to use readily available pulse transformers and analog diffential interfaces for reduction of RFI and common mode noise on the interface. Also advantages in decoupling noise from the transport from getting into the analog circuits through the DAC dgital inputs.
I guess it is easier to rant with hysterics about the SPDIF interface than to go do the work of improving it. My experience and that of a many well reguarded DAC designers is that this interface has plenty of potential and is not ready to be thrown out yet. The Audio Alchemy and Muse digital interfaces seem to have dissappeared. The I2S cable that I built for the AA stuff worked best in very short lengths(15") due to the lack of proper termination by the way.
http://www.diyvideo.com/forums/showthread.php?s=&postid=90854#post90854
Fred
P.S. It is nice to know that the forum has such hands off moderators so I can be called an "abusive *****" for discussing technical issues that I have three years of professional experience in. Makes me really want to be a contributer here.
Post edited(word has been censored), as result of offender's(not user Fred Dieckmann) usage of it with malicious intent has been censored. Your complaint has been heard, and action has been taken as result. Please refer to SinBin for result. -CHRIS8
Re: have turned you into an abusive prick
In all that fascinating technical discussion, I didn't even noticed that
. I thought your threshold for abuse was set up higher. In the light of recent events and the fact that moderators are under the fire, I will not even think about proposing anything like sin bin sentence.
jwb, you used unnaceptable language and offended a respected member of this forum. I think that apology is in order. BTW, do you still consider me to be harmless?
Fred Dieckmann said:
In all that fascinating technical discussion, I didn't even noticed that
. I thought your threshold for abuse was set up higher. In the light of recent events and the fact that moderators are under the fire, I will not even think about proposing anything like sin bin sentence. jwb, you used unnaceptable language and offended a respected member of this forum. I think that apology is in order. BTW, do you still consider me to be harmless?
I2S
I have found that I2S needs to be as tweaky as possible to get the best results. Most I2S cables and connectors suck. If you can, hardwire the cables at both ends and if you must buffer, try it at one end only. Of course, the wire and dialectric used will make a big difference. At one point I was using an Audio Magic I2S cable that was connected between a modified AA jitter/bit expander and a custom DAC. I had already hardwired the cable to the DAC pins but the other end was using the usual mini-din plug into the AA piece. Then I made my own I2S cable using 6 nines copper wire in air and thin teflon and hardwired it at both ends. Way, way better sound. So, there is no free lunch. Everything has to be done to the max to get the best sound.
Ric Schultz
I have found that I2S needs to be as tweaky as possible to get the best results. Most I2S cables and connectors suck. If you can, hardwire the cables at both ends and if you must buffer, try it at one end only. Of course, the wire and dialectric used will make a big difference. At one point I was using an Audio Magic I2S cable that was connected between a modified AA jitter/bit expander and a custom DAC. I had already hardwired the cable to the DAC pins but the other end was using the usual mini-din plug into the AA piece. Then I made my own I2S cable using 6 nines copper wire in air and thin teflon and hardwired it at both ends. Way, way better sound. So, there is no free lunch. Everything has to be done to the max to get the best sound.
Ric Schultz
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.