I really can not follow you... You are very focused on telling how non-interesting this DAC is, because it's based on CS8416 and CS4398, which is also used in some cheap DAC's on EBAY. And then again... There are lots of TDA1541, TDA1543, AD1853, PCM1794, WM8740, AK4394/5/6 around 200 USD. Does this mean, that it will be completely uninteresting to use any of these chips??? If so, please let us know what DAC-chip is interesting.....
I really do not understand!
High performance audio is not just about picking the "best" components. It is much more about tuning the different parts into a perfect design. Yes, you can find lots of bad CS4398 designs, as you can also find very nice CS4398 designs.... But that also goes for any other DAC chip. If you do not agree, please let me know a DAC-chip, that can not end up in a bad performing DAC.
What we are focusing on, is NOT to pick some chips, that in theory will be the best. We aim to make the best sonic performer. And as far as we have come, CS4398 has kicked some serious butt, on everything we have tried from TI, AD, AKM and Wolfson.
I aree with you, that the Wolfson SPDIF receiver may be a good choice, and we may end up picking that based on your recommendations. But it has to perform in a sonic performance test.
By focusing only on sonic performance, we will not end up with a typical "EBAY junk style" as you name it.
BTW: Why is it, that you mention that the technology in CS4398 is that bad?? I showed THD at -107dB in 2003. Wolfson are only at -104dB 6 years later. Why is it, that Wolfson can not outperform the old CS4398, if teh technology is that much better??
Hi,
Actually, there are few e-bay DAC's that come with Case and fully assembled for 200 Bucks, but many that do use CS4398.
My point is, why build laboriously at much higher cost (as your 200 Buchs comes without a good looking case) a DAC you can buy finished (3-Pin regulators, Op-Amp output stage, CS8416, CS4398)?
It is like trying to build a copy of Yugo car at higher cost.
I hope so. Your original proposal at the beginning of the thread sounded more like it would end up the other way.
THD measured with many averaged cycles is one of the least useful indicators for sonic performance in about anything, including/especially DAC's.
I look for example at the digital filter technology, the amount of out of band noise and the "instantaneous" distortion long before I even consider DNR & THD+N.
Ciao T
Actually, there are few e-bay DAC's that come with Case and fully assembled for 200 Bucks, but many that do use CS4398.
My point is, why build laboriously at much higher cost (as your 200 Buchs comes without a good looking case) a DAC you can buy finished (3-Pin regulators, Op-Amp output stage, CS8416, CS4398)?
It is like trying to build a copy of Yugo car at higher cost.
I hope so. Your original proposal at the beginning of the thread sounded more like it would end up the other way.
THD measured with many averaged cycles is one of the least useful indicators for sonic performance in about anything, including/especially DAC's.
I look for example at the digital filter technology, the amount of out of band noise and the "instantaneous" distortion long before I even consider DNR & THD+N.
Ciao T
Digital filter technology? Curious about how the technology rather than the performance is of interest to you. Also what's "instantaneous" distortion and can you tell what its going to be from the datasheet or do you need some special measuring equipment?
Hi,
Technology as in what particular coefficients are used and how they relate to subjective performance etc...
You can measure "instantanious" distortion by sending a known complex signal into a DAC and then recording the output with a suitable dynamic range and sample rate ADC (must be much better than the DAC) and comparing the two. You cannot measure it with normal Audio Stuff (e.g. AP Two), the bandwidth needed means you need pretty serious RF capable gear.
Using traditional DAC's (multibit) you get a close correlation. Modern noise shaping DAC's have a large amount of deviation at any given sample point, if the ADC samples at a sufficiently high frequency.
It is of course inherent to the underlying principles of operation.
Ciao T
Technology as in what particular coefficients are used and how they relate to subjective performance etc...
You can measure "instantanious" distortion by sending a known complex signal into a DAC and then recording the output with a suitable dynamic range and sample rate ADC (must be much better than the DAC) and comparing the two. You cannot measure it with normal Audio Stuff (e.g. AP Two), the bandwidth needed means you need pretty serious RF capable gear.
Using traditional DAC's (multibit) you get a close correlation. Modern noise shaping DAC's have a large amount of deviation at any given sample point, if the ADC samples at a sufficiently high frequency.
It is of course inherent to the underlying principles of operation.
Ciao T
Instantaneous distortion measurement
Wow, sounds like you must have some serious kit available to play with. To better today's best DAC's performance would mean, say, a 20+ bit ADC running above 200kHz sample rate? Yet the AP (admitted it was AP1 when I last used one) could do 200kHz (with ****-poor THD though). So are we looking at 20bits and 1+MHz sampling here to get useful results?
Does the ADC need to be locked to a multiple of the DAC sample rate? Or is purely asynchronous operation going to be OK?
Wow, sounds like you must have some serious kit available to play with. To better today's best DAC's performance would mean, say, a 20+ bit ADC running above 200kHz sample rate? Yet the AP (admitted it was AP1 when I last used one) could do 200kHz (with ****-poor THD though). So are we looking at 20bits and 1+MHz sampling here to get useful results?
Does the ADC need to be locked to a multiple of the DAC sample rate? Or is purely asynchronous operation going to be OK?
Hi,
The RF gear I need to go to some other people's labs... My own park is limited to various Pro Sound Cards, an AP Two and some nice 'scope's including some digital ones with nice FFT build in.
You can use a 12 Bit ADC and look at full scale signals. The issues are obvious enough.
I suspect locking to multiples would work very well, I never managed to spend enough time to really test the various options. Typhically it is "hook up the Eval PCB, play the sequence and record", then repeat many 100's of times and then do stat's on the results plus compare some individual traces to averages and the original signal.
I am sure one can do better than that. Prior to doing this I was limited to sending in comb-filtered pink noise and see what "grass" crops in the forks of the comb using short FFT sequences.
Ciao T
The RF gear I need to go to some other people's labs... My own park is limited to various Pro Sound Cards, an AP Two and some nice 'scope's including some digital ones with nice FFT build in.
You can use a 12 Bit ADC and look at full scale signals. The issues are obvious enough.
I suspect locking to multiples would work very well, I never managed to spend enough time to really test the various options. Typhically it is "hook up the Eval PCB, play the sequence and record", then repeat many 100's of times and then do stat's on the results plus compare some individual traces to averages and the original signal.
I am sure one can do better than that. Prior to doing this I was limited to sending in comb-filtered pink noise and see what "grass" crops in the forks of the comb using short FFT sequences.
Ciao T
WM8805 oscillator and master clock
The WM8805 has a built-in crystal oscillator and can provide a system clock, from the PLL or crystal. We were going to use a DIP xtal oscillator but... ( from Wolfson datasheet):
"Alternatively, an external CMOS compatible clock signal can be applied to the XIN pin in the absence of a crystal, although this is not recommended when using the PLL as the PLL requires a jitter-free OSCCLK signal for optimum performance."
It doesn't make sense to me, as the oscillator is fed to the PLL to get rid of jitter!? Wolfson is saying a Pierce oscillator w/crystal is better? Strange.
I'd use a quartz crystal on the WM8805 and the master clock output (not from the PLL) for the SRC and the DAC.
The WM8805 has a built-in crystal oscillator and can provide a system clock, from the PLL or crystal. We were going to use a DIP xtal oscillator but... ( from Wolfson datasheet):
"Alternatively, an external CMOS compatible clock signal can be applied to the XIN pin in the absence of a crystal, although this is not recommended when using the PLL as the PLL requires a jitter-free OSCCLK signal for optimum performance."
It doesn't make sense to me, as the oscillator is fed to the PLL to get rid of jitter!? Wolfson is saying a Pierce oscillator w/crystal is better? Strange.
I'd use a quartz crystal on the WM8805 and the master clock output (not from the PLL) for the SRC and the DAC.
Instantaneous distortion or something else?
Attached an output waveform from a NS DAC I'm considering buying. Obvious even with the 8 bit ADC in the 'scope that its not a pure sinewave. Yet I take the roughness to be out of band noise. Are you saying that 'instantaneous distortion' is not out of band noise by another name?
Attached an output waveform from a NS DAC I'm considering buying. Obvious even with the 8 bit ADC in the 'scope that its not a pure sinewave. Yet I take the roughness to be out of band noise. Are you saying that 'instantaneous distortion' is not out of band noise by another name?
Attachments
Hi,
Yup. The on chip generator is okay, but far from great.
I think the concern is that in system design some designers may share the oscillator between multiple loads and maybe place it far from the WM880X, adding excessive ground bounce etc.
Using the CLKOUT pin to drive multiple loads and to isolate the oscillator is good practice. From my experience designing the 880X into an OEM Customers product suggests that using a high quality oscillator with a suitably low noise supply perform markedly better than the on board pierce oscillator.
Ciao T
Yup. The on chip generator is okay, but far from great.
I think the concern is that in system design some designers may share the oscillator between multiple loads and maybe place it far from the WM880X, adding excessive ground bounce etc.
Using the CLKOUT pin to drive multiple loads and to isolate the oscillator is good practice. From my experience designing the 880X into an OEM Customers product suggests that using a high quality oscillator with a suitably low noise supply perform markedly better than the on board pierce oscillator.
Ciao T
Hi,
Hmm, so, if we call it "out of band noise" it stops being distortion. Clever.
Maybe I stop including > 20KHz harmonics in amplifer THD measurements because they are "out of band noise" so that my class AB amplifiers no longer show a rise in THD with rising frequency.
Forgive me, but I personally prefer to characterise digital systems in the way I do with analogue ones. In which case we have a bad case of noise modulation or maybe LFD's much vaunted "fuzzy distortion" at our hands (and yes, I understand that this is how DS works inherently).
But of course it can be made inaudible by calling it "OOBN" instead of "Distortion"...
Or maybe not.
Ciao T
PS, if you can see this fuzz with a 12-Bit ADC it means it's pretty bad, no? Definitely not the kind of THD claimed in the datasheet (of course, the Datasheet specifies that THD is measures with a high order 20KHz lowpass, somewhere in small print few people read.
You might also want to try the comb filtered noise, just for fun.
Hmm, so, if we call it "out of band noise" it stops being distortion. Clever.
Maybe I stop including > 20KHz harmonics in amplifer THD measurements because they are "out of band noise" so that my class AB amplifiers no longer show a rise in THD with rising frequency.
Forgive me, but I personally prefer to characterise digital systems in the way I do with analogue ones. In which case we have a bad case of noise modulation or maybe LFD's much vaunted "fuzzy distortion" at our hands (and yes, I understand that this is how DS works inherently
). But of course it can be made inaudible by calling it "OOBN" instead of "Distortion"...
Or maybe not.
Ciao T
PS, if you can see this fuzz with a 12-Bit ADC it means it's pretty bad, no? Definitely not the kind of THD claimed in the datasheet (of course, the Datasheet specifies that THD is measures with a high order 20KHz lowpass, somewhere in small print few people read.
You might also want to try the comb filtered noise, just for fun.
Hi,
Hmm, so, if we call it "out of band noise" it stops being distortion. Clever.
Maybe I stop including > 20KHz harmonics in amplifer THD measurements because they are "out of band noise" so that my class AB amplifiers no longer show a rise in THD with rising frequency.
Forgive me, but I personally prefer to characterise digital systems in the way I do with analogue ones. In which case we have a bad case of noise modulation or maybe LFD's much vaunted "fuzzy distortion" at our hands (and yes, I understand that this is how DS works inherently
But of course it can be made inaudible by calling it "OOBN" instead of "Distortion"...
Or maybe not.
Ciao T
PS, if you can see this fuzz with a 12-Bit ADC it means it's pretty bad, no? Definitely not the kind of THD claimed in the datasheet (of course, the Datasheet specifies that THD is measures with a high order 20KHz lowpass, somewhere in small print few people read.
You might also want to try the comb filtered noise, just for fun.
I do recall you saying, that if your amp is a wide band one, then why filter the DAC output at all. Why not let it pass through?
To me that is probably the worst thing I can imagine, because I´ve heard what it does to even stable and well designed gear.
It might be due to IMD or something, but I do not need any measurements or theoretical basics to get the grasp of the severety of this kind of signals.
The best formulation of an axiom considering this might be " be capable of signal transparancy up to at least 3-5 times what you need, and then don´t do it anyway".
Musical signals are very subtle above just 4-6 KHz. The highest basic tone on a piano or a violin is just around 4 KHz, the rest is very subtle overtones loosing level with every octave rising.
Digital circuits however, lets out super sonic signals at much higher levels, i.e. -5-20 dB FS @ i.e. 3-400 KHz.
No rocket science there - it parallels the traditional 'THD+N' figure. Distortion is signal-related error and noise is all the rest. Are you thinking there's a good opportunity for marketing ** here?
Maybe I stop including > 20KHz harmonics in amplifer THD measurements because they are "out of band noise" so that my class AB amplifiers no longer show a rise in THD with rising frequency.
Hate to break it to you, but it'll still show up if you include a THD vs freq plot for your babies. As in some opamp datasheets I've seen - the distortion rises to around 7kHz, then starts to fall. An indication they're using the AP's 22kHz low pass filter
Forgive me, but I personally prefer to characterise digital systems in the way I do with analogue ones. In which case we have a bad case of noise modulation or maybe LFD's much vaunted "fuzzy distortion" at our hands (and yes, I understand that this is how DS works inherently
I see no reason to change the rules for digital when compared to analog either. So in your stereo FM tuner, you include any 38kHz carrier feed-through in the figures? Its only noise modulation if its signal dependent. Otherwise its just plain old noise. I agree that noise modulation is probably the elephant in the room where digital systems are concerned. Rich Cabot proposed a test way back in the 90's, Stereophile used it for a while and found it was a fairly hopeless test as almost everything passed with glowing colours. So they stopped testing for it. But I think there's a need for a good noise modulation test in digital systems.
But of course it can be made inaudible by calling it "OOBN" instead of "Distortion"...
Or maybe not.
Its a good question. I'm reminded of Douglas Adam's 'SEP' - in the rather conservative world of hifi, it becomes the pre-amplifier (or power amp, or speaker) designer's problem. That's one reason I'm a systems designer, not a component designer - must remember to update my sig
I included the pic to demonstrate that your 12 bit ADC is by no means necessary to see it. That scope uses a mere 8 bit ADC, not too invisible is it?
I don't have the equipment myself. What did you find when you did it?
Hi,
Well, you can produce an amp that has flat THD & N up to KHz (without excessive 2nd HD or Noise). Then it will not show a rise...
Good question. Thanfully FM radio is so poor that I never really bothered... ;-)
My biggest issue was that the pilot tone used to drive me mad, despite the huge levels of attenuation in the MPX stereo system, the pilot tone filter and speakers. I still could hear it. Nowadays my HF response in my ears has tailed off enough that I am likley no longer bothered, but there is nothing on radio worth listening to (cue in Queen - Radio Gaga).
Same here, I like do things as whole systems.
The test only needs a decent FFT (even AP2 will do) and a suitable signal (something broadly similar can be concocted in the AP2 as well).
I tested a DAC by a company that I shall not mention to avoid complaints that I am "bashing" their products. With the combfiltered noise having peaks at 0dbfs (but not above) we observed a noisefloor between -40...-50dB filling in the the deep notches of the signal. This noisefloor was signal dependent and quite wideband. Op-Amp changes on their eval board managed to reduce levels to around -60dB, still much more than I find acceptable.
Ciao T
Hate to break it to you, but it'll still show up if you include a THD vs freq plot for your babies. As in some opamp datasheets I've seen - the distortion rises to around 7kHz, then starts to fall. An indication they're using the AP's 22kHz low pass filter
Well, you can produce an amp that has flat THD & N up to KHz (without excessive 2nd HD or Noise). Then it will not show a rise...
Good question. Thanfully FM radio is so poor that I never really bothered... ;-)
My biggest issue was that the pilot tone used to drive me mad, despite the huge levels of attenuation in the MPX stereo system, the pilot tone filter and speakers. I still could hear it. Nowadays my HF response in my ears has tailed off enough that I am likley no longer bothered, but there is nothing on radio worth listening to (cue in Queen - Radio Gaga).
Its a good question. I'm reminded of Douglas Adam's 'SEP' - in the rather conservative world of hifi, it becomes the pre-amplifier (or power amp, or speaker) designer's problem. That's one reason I'm a systems designer, not a component designer - must remember to update my sig
Same here, I like do things as whole systems.
The test only needs a decent FFT (even AP2 will do) and a suitable signal (something broadly similar can be concocted in the AP2 as well).
I tested a DAC by a company that I shall not mention to avoid complaints that I am "bashing" their products. With the combfiltered noise having peaks at 0dbfs (but not above) we observed a noisefloor between -40...-50dB filling in the the deep notches of the signal. This noisefloor was signal dependent and quite wideband. Op-Amp changes on their eval board managed to reduce levels to around -60dB, still much more than I find acceptable.
Ciao T
Hi,
That is not QUITE what I said.
I said:
"If your amplifier has a power bandwidth of XKHz, why should there be any problem presenting it with a signal within that power bandwidth all the way to full power."
Actually, that is my point. When done with stable and correctly designed equipment it does NOTHING.
For fun, including the Sinc funtion that is unavoidable and using a 50KHz first order filter in conjunction with a non-oversampling DAC at 44.1KHz we find that the first image of 4KHz appears with a level of 21dB below the originating signal. The first Image of 1KHz is around -33.5dB below the originating signal and the first image of 100Hz will 54dB below the originating signal.
The second set of images are again at -54dB for the 100Hz image, -34dB for 1KHz and -24dB for the image of 4Khz. For images of 1KHz around the 200KHz mark the levels for images are around 54dB below the originating signal. At around 300KHz we get around -66dB for the images of 1KHz.
No amplifier (or speaker) I have ever designed would have had problems with me applying such a signal.
Ciao T
That is not QUITE what I said.
I said:
"If your amplifier has a power bandwidth of XKHz, why should there be any problem presenting it with a signal within that power bandwidth all the way to full power."
Actually, that is my point. When done with stable and correctly designed equipment it does NOTHING.
For fun, including the Sinc funtion that is unavoidable and using a 50KHz first order filter in conjunction with a non-oversampling DAC at 44.1KHz we find that the first image of 4KHz appears with a level of 21dB below the originating signal. The first Image of 1KHz is around -33.5dB below the originating signal and the first image of 100Hz will 54dB below the originating signal.
The second set of images are again at -54dB for the 100Hz image, -34dB for 1KHz and -24dB for the image of 4Khz. For images of 1KHz around the 200KHz mark the levels for images are around 54dB below the originating signal. At around 300KHz we get around -66dB for the images of 1KHz.
No amplifier (or speaker) I have ever designed would have had problems with me applying such a signal.
Ciao T
My scope's 'math' function includes an FFT, not in slightest bit 'decent' though. So I guess I might have to build something myself - AP is very poor 'bang for the buck' for the things I'm interested in playing with.
Oh, now I want to guess who this might be. The company that started out making switched-cap instrumentation A/Ds in the 1980s?
Hi,
My 'alternate" is aEMU 1616M, but it's ADC is not good for this kind of measurements.
Could be, could be.
Ciao T
My 'alternate" is aEMU 1616M, but it's ADC is not good for this kind of measurements.
Could be, could be.
Ciao T
I´ve seen mirror measurements of TDA chips much worse than what you say, but never mind, I´ve also listened to them.
It seems to me, that our ideas of good electronics, are completely different from yours, which really makes it hard to get a common view on the proportions of design issues, and thus the rank of priorities.
Of course one should do every step as good as possible, but "as good as possible" will be completely different, when one of the originators like tube gear, FETs, op-amps, computer based digital sources and NOS DACs, and the other ones do not.
This I think is pretty much the current issue.
I think it would be a good idea to try out the Wolfson receiver, not that I believe it will do much difference, but smaller improvements also should be considered.
Choosing a DAC chip is of course also important, and I think BBPCM1704K might still be the best on market, but I still did not ever hear any gear equipped with that chip, get anywhere near my own CS4398 style DAC.
I still think it is mostly about implementation of the chips, more than choosing X or Y of what ever reason, and most important, a very transparent analog stage is needed, if you really want to hear your records. This also goes for the rest of the audio chain, and in exactly that matter I feel pretty confident.
It will be a matter of experiments, I do believe in your experience as a designer, but I do not believe in your priorities though.
But I am thankfull for the input regarding the input receiver, and it will be regarded.
Hi,
Then perhaps the analog stages where remiss, or the chips involved where not actually 16 Bit capable (eg. TDA1543 used, not TDA1541 with DEM synchronisation). The numbers i listed do expect competent implementation.
Note, I did not suggest using Non-OS and so on. To me Non-OS, Tubes etc. are one way of doing things and I did not promote them for use in your project.
I have only taken exception where you make untrue statements as to what is happening with Non-OS designs (and tubes etc.).
You are lumping things together in a "us and them" fashion, where everyting that you have developed some form of aversion to (justified or not) is in the "them" part.
For example, correctly Implemented PC based sources can have an SPDIF output with levels of Jitter comparable to the very best CD transports out there and match or exceed the subjective sonic results produced by any CD-Transport.
To lump tubes, op-amps and fets into one is even more extreme than the islamic dietary laws, and more to the point, it accepts other items that have severe issues of their own (talk about straining the gnat and swallowing camels).
No I actually personally for my own designs agree with the approach that I only take into consideration any outside views and recommendation such as do not seriously conflict with my carefully established, cultivated and strong prejudices.
But I do not try to make "open source" designs that are meant to please, if not the all then at least the many. I try to end up with gear that conforms to my experience of what real music sound like, not some "HiFi Sound", artifice or "better than real" thing, according to my perception (and of those included in listening tests).
You may be surprised, especially if you leave yourself the option to bypass the ASRC so you can hear what the ASRC does when it does not have gobs of jitter from very poor quality receivers to clean up.
If you prefer the sound coming from DAC's that represent music as modulated noise (such a preference is perfectly valid) then the PCM1704 will not be the right choice for you. In addition, like any DAC the PCM1704 needs care in implementation and is subject to several tricks for optimising performance. Who knows, if you spend the effort you did on the CS4398 on a PCM1704 DAC the results may easily surpass the Cirrus Logic chip.
Another way may be to compare a 'non-optimised" CS4398 DAC to a non-optimised PCM1704 DAC with general technology levels equal.
As they say, you can burnish a turd and it may shine even nicer than tarnished gold, but burnish the gold....
I completely agree. In the end the empirical methode is essential here.
So, again we may come back to a project (interim or final) that provides the basics needed:
* pre-regulated power supplies for DAC
* receiver with supplies
* ASRC with bypass and selectable output formats (so we can easily use EIAJ input format DAC's)
* clock circuitry with it's local super low noise regulators (aim for < 5nV|/Hz - the discrete regs I use in my commercial designs manage 1nV|/Hz) etc.
* a reasonably universal non feedback and direct coupled analog stage may be designed so that J-FET or BJT devices can be used in the critical stages and with the option to trade servo's for manual offset adjustment or coupling caps)
* nice low noise shunt regulators for the analog stages
All of the above on one PCB, to which may be docked a smallish PCB that that contains the actual DAC, with it's optimised local regulation, local decoupling and other necessary parts.
Include a few "breakoff PCB Patters including one for your killer CS4398 application and maybe one for PCM1794A (people seem to like that Chip) and the AKM4397 (in case you would like a Cirrus Logic style DAC that actually sounds quite good ) and maybe the latest greatest ESS Chip and a Wolfson DAC.
Voila, you have a project that can be pretty much "all things to all DIY'ers", will likely be humungously popular and you can solicit input for the design of the optimised circuits for the "other" (not Cirrus Logic) Chips from teh community, as you could for other parts of the design (power supplies etc. et al).
In fact, colour me badd, but I would say something along these lines (actually likely split into three different PCB's linked suitably) is more like "open source".
If we had a standard PCB size and set of connections (which can be made NOT using screw terminals ) for a DAC and for a analogue stage others could add their alternative modules to the "open source" design, so that the DIY'er who likes to use PCM1704 and Tube stages can do so, while those who want CS4398 and discrete bipolar analogue stages can use those.
Yes, having these seperate sections has the potential for some minimal degradation compared to a single PCB, but how about SMA/SMB coax to carry digital signals (maybe even the analog connections) and solid core PTFE IDC ribbon cable for analog and of course seperate transformers and supplies for each section (input/receiver/asrc - DAC - Analog).
Ciao T
Then perhaps the analog stages where remiss, or the chips involved where not actually 16 Bit capable (eg. TDA1543 used, not TDA1541 with DEM synchronisation). The numbers i listed do expect competent implementation.
Note, I did not suggest using Non-OS and so on. To me Non-OS, Tubes etc. are one way of doing things and I did not promote them for use in your project.
I have only taken exception where you make untrue statements as to what is happening with Non-OS designs (and tubes etc.).
You are lumping things together in a "us and them" fashion, where everyting that you have developed some form of aversion to (justified or not) is in the "them" part.
For example, correctly Implemented PC based sources can have an SPDIF output with levels of Jitter comparable to the very best CD transports out there and match or exceed the subjective sonic results produced by any CD-Transport.
To lump tubes, op-amps and fets into one is even more extreme than the islamic dietary laws, and more to the point, it accepts other items that have severe issues of their own (talk about straining the gnat and swallowing camels).
No I actually personally for my own designs agree with the approach that I only take into consideration any outside views and recommendation such as do not seriously conflict with my carefully established, cultivated and strong prejudices.
But I do not try to make "open source" designs that are meant to please, if not the all then at least the many. I try to end up with gear that conforms to my experience of what real music sound like, not some "HiFi Sound", artifice or "better than real" thing, according to my perception (and of those included in listening tests).
You may be surprised, especially if you leave yourself the option to bypass the ASRC so you can hear what the ASRC does when it does not have gobs of jitter from very poor quality receivers to clean up.
If you prefer the sound coming from DAC's that represent music as modulated noise (such a preference is perfectly valid) then the PCM1704 will not be the right choice for you. In addition, like any DAC the PCM1704 needs care in implementation and is subject to several tricks for optimising performance. Who knows, if you spend the effort you did on the CS4398 on a PCM1704 DAC the results may easily surpass the Cirrus Logic chip.
Another way may be to compare a 'non-optimised" CS4398 DAC to a non-optimised PCM1704 DAC with general technology levels equal.
As they say, you can burnish a turd and it may shine even nicer than tarnished gold, but burnish the gold....
I completely agree. In the end the empirical methode is essential here.
So, again we may come back to a project (interim or final) that provides the basics needed:
* pre-regulated power supplies for DAC
* receiver with supplies
* ASRC with bypass and selectable output formats (so we can easily use EIAJ input format DAC's)
* clock circuitry with it's local super low noise regulators (aim for < 5nV|/Hz - the discrete regs I use in my commercial designs manage 1nV|/Hz) etc.
* a reasonably universal non feedback and direct coupled analog stage may be designed so that J-FET or BJT devices can be used in the critical stages and with the option to trade servo's for manual offset adjustment or coupling caps)
* nice low noise shunt regulators for the analog stages
All of the above on one PCB, to which may be docked a smallish PCB that that contains the actual DAC, with it's optimised local regulation, local decoupling and other necessary parts.
Include a few "breakoff PCB Patters including one for your killer CS4398 application and maybe one for PCM1794A (people seem to like that Chip) and the AKM4397 (in case you would like a Cirrus Logic style DAC that actually sounds quite good
) and maybe the latest greatest ESS Chip and a Wolfson DAC.Voila, you have a project that can be pretty much "all things to all DIY'ers", will likely be humungously popular and you can solicit input for the design of the optimised circuits for the "other" (not Cirrus Logic) Chips from teh community, as you could for other parts of the design (power supplies etc. et al).
In fact, colour me badd, but I would say something along these lines (actually likely split into three different PCB's linked suitably) is more like "open source".
If we had a standard PCB size and set of connections (which can be made NOT using screw terminals
) for a DAC and for a analogue stage others could add their alternative modules to the "open source" design, so that the DIY'er who likes to use PCM1704 and Tube stages can do so, while those who want CS4398 and discrete bipolar analogue stages can use those.Yes, having these seperate sections has the potential for some minimal degradation compared to a single PCB, but how about SMA/SMB coax to carry digital signals (maybe even the analog connections) and solid core PTFE IDC ribbon cable for analog and of course seperate transformers and supplies for each section (input/receiver/asrc - DAC - Analog).
Ciao T
@ThorstenL
I understand that a modular design is what you want, but I think you might be the one to do a such project then, because that is not what we want.
We do not want to contribute to something that we do not like ourselves.
And it is no longer prejudice, but hard earned experience, when op-amps, tubes, FETs and NOS designs are discarded.
I would follow a modular design with interest though, so I look forward to see such things described.
But this is to be a compact design with everything on one board incl. separate transformers for D and A.
I understand that a modular design is what you want, but I think you might be the one to do a such project then, because that is not what we want.
We do not want to contribute to something that we do not like ourselves.
And it is no longer prejudice, but hard earned experience, when op-amps, tubes, FETs and NOS designs are discarded.
I would follow a modular design with interest though, so I look forward to see such things described.
But this is to be a compact design with everything on one board incl. separate transformers for D and A.
Hi,
Not quite. I think it is what the community would react very favourably to.
For me it's a non-issue, I have tons of prototyping PCBs that combine smd footprints and pads, together with inspired 3d hardwiring accomodate a lot.
Good luck, I hope the results get you what you desire.
Ciao T
Not quite. I think it is what the community would react very favourably to.
For me it's a non-issue, I have tons of prototyping PCBs that combine smd footprints and pads, together with inspired 3d hardwiring accomodate a lot.
Good luck, I hope the results get you what you desire.
Ciao T
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.