Well... why not use th WM8805?? I could ask: Why not use CS8416 😉
Case is, that Crystal Semiconductors have been the marked leader in SPDF receivers since early 1990's, and with good reason. The simply perform very good.
Jitter performance is nice, and correctly implemented way better than specifications. CS has made these specs very conservative.
Also tha CS chips always lock perfectly to signal. To be honest, we have never had any reason to consider anything else.
One reason to use the WM8805 is the multi-spdif pass thru. If you need that.
Also much lower cost than CS.
Also much lower cost than CS.
I would guess with a short cable.
The risk of mains hum modulated jitter with a galvanically connected, single ended interface is great .
I don't see the use of multi-SPDIF pass thru.... For what applications are you thinking?
About cost: The CS8416 is priced at 9,5 USD Audio Processing | Digi-Key
I don't know WTB the WM8805 at a lower price. Please let me know....
It was terminated as 75 Ohm and with a short cable.
Transformers do have hysteresis no matter what transformer you choose, and that corrupts the waveform with both rounding and overshoot imposing jitter.
But of course the transformer can be necessary, but actually I think in one end only. I do not know any gear utilizing transformers in the SPDIF input, it has always been output PT´s.
If I had to choose one though, I´d go for scientific conversion.
There is no PLL in the DAC itself, PLL is found in the reciever, therefor the DAC is pretty sensitive to jitter allthough Crystal says it is not 😀.
Intermodulation of course is an issue which should not be ignored, because one never knows if it will occur, but in this case it is easy, just choose another clock.
I don´t remember werewolf saying such a thing, because eliminating jitter is the main purpose of upsampling.
The AD1896 and SRC4192 are both ASRC´s with internal buffers, intended for incoming jitter cancellation.
Thus the ASRC´s intrinsic jitter, depends on the stability of the MCLK. If the MCLK is precise down to the 3psec they are specified to, then the DAC itself will have an almost jitterfree incoming signal if not poluted on its way.
This is an important matter, since jitter becomes more and more malign with higher sampling frequency, and we do want high sampling frequency because of resolution.
But I agree with you that a clean and as jitterfree SPDIF signal as possible is preferable. My own DAC, which is the earlier project, has all theese features, but still the incoming signal quality is very important to it, which it should not be in theory.
But looking at the signal from the SPDIF source, then extensive jitter and waveform corruption naturally can lead to misinterpretation of the data by the reciever. So a good clean SPDIF signal is crousial, and if possible the 75 Ohm impedance should be regarded as precisely as possible.
But I do not feel confident that transformers is the right way to move in this regard. Everyone i saw until now, did really corrupt the waveform seriously, but you have to hold that up against huge ground potentials and noise.
Yes, if you terminate both TX end RX end with 75 Ohm resistor, then put transformer between, there will be degradation, no doubt, I agree with you.
But id you make TX end...say..300 Ohm, then put transformer between and make right termination on RX end + driver circuit infront of CS8416 (or H-pad to attenuate reflexions), then I disagree.
Square waveform corruption, rounding or overshoot is usually caused by impedance mismatch or reflexion.
National have nice high speed things for line driver, Linear have nice documents how to proper decouple high speed things, Salas designed nice shunt to power supply those things.
Newava have some nice transformers and TDR is telling me about proper termination and how long should be digital cables. Much longer than me 🙂
Last edited:
Hi,
Hmmm, laudable idea indeed. Let's make a really good DIY DAC everyone can build.
I find this interesting! May I enquire what where the DAC's tested for comparison?
Have you compared EVERY DAC available in the market? Maybe I did something wrong, but I was left unimpressed by the CS4398 when compared to just about ANY other choice with similar outline spec, never mind broadening the horizon to less common options.
Market Leader in what? Producing ridiculously high levels of intrinsic jitter? Okay, they appear to have finally fixed that in recent revisions, if you set the operating mode right it is almost as good as the ancient 8414. Oh pshaw...
What is the Sample Rate Converter for? I never saw (or heard) the point, not even using the dCS ASRC. Most DAC's perform better at lower sample rates than at higher ones anyway and ASRC's can never improve any signal presented to them, at best they cause minimal harm, of course things are rarely at "best" status.
The kind of design you have set out to make available as PCB Kit for 200 Bucks - well you can buy the same sort of e-bay for the same money, but fully assembled and with a nice case, made in China. And the Wima "Foil" capacitors they use are likely the same fakes you would end up buying.
Zhaolu D2.5
So why waste time replicating this kind of thing very laboriously and at greater cost?
IF I where to go out on a limb to actually give you some advise and if sound quality is of ANY concern at all (instead of matching marketing propaganda), I'd say:
Use a better DAC (almost ANY DAC will be better, one with a "slow rolloff" filter option is preferred). Have a read here at DIYA to see what others feel works well, for starters. Even better, why not make the DAC a plugin module with PCB options for a range of popular chips as "breakoff" parts on the side of the main PCB?
Use a better Receiver (almost ANY other available Receiver will be better, including Cirrus Logic 8414). Implement the receiver input correctly. This usually means either brutally direct or with some really clever circuitry (Pat "Jocko Homo" DiGiaccomo discussed and shared some here before he had enough and buggered off). From experience, correctly done the "clever circuitry" works better than "brutally simple", but "brutally simple" works much better than pretty much anything else.
Do not use an ASRC, if you must remove jitter, do it the correct way, with a secondary PLL.
There are several usable published examples, but nothing has ever been implemented that is directly accessible for the average DIY'er. Tent Labs has nice VCXO's for the Job, low noise power supplies and low noise PLL loop filters can be a challenge to implement, but if done right jitter can be quite effectively locked out.
This could be a plug in module or maybe implemented as something people can leave unpopulated if the expense is an issue, so DIY'er have the option.
Use better powersupplies than 317/337 per application note, they are so 1980's. Here on DIYAudio there are many published options that do better and do not cost appreciably more, I am kind of partial to something called "Teddyreg" for ease/performance ratio, but there are many others.
Do not use Op-Amp output Stages, try something like this: http://www.diyaudio.com/forums/attachments/digital-line-level/141182d1253190252-few-questions-pcm1793-based-usb-dac-sacdenhancer_evo3.pdf costs less and sounds better. Using J-Fets might get even better, but starts costing real money, one could provide footprints allowing a cheap Bipolar version and a more expensive J-Fet option.
Anyway, those are my not particularly humble opinions on what would produce a DAC that would give the DIY(A) community something worth building, rather than to instead just browse e-bay for an already build unit for less.
And if you can make it so that this "worth building DAC" can be build for 200 Bucks in parts cost and PCB set cost together, all the better.
Ciao T
Hmmm, laudable idea indeed. Let's make a really good DIY DAC everyone can build.
I find this interesting! May I enquire what where the DAC's tested for comparison?
Have you compared EVERY DAC available in the market? Maybe I did something wrong, but I was left unimpressed by the CS4398 when compared to just about ANY other choice with similar outline spec, never mind broadening the horizon to less common options.
Market Leader in what? Producing ridiculously high levels of intrinsic jitter? Okay, they appear to have finally fixed that in recent revisions, if you set the operating mode right it is almost as good as the ancient 8414. Oh pshaw...
What is the Sample Rate Converter for? I never saw (or heard) the point, not even using the dCS ASRC. Most DAC's perform better at lower sample rates than at higher ones anyway and ASRC's can never improve any signal presented to them, at best they cause minimal harm, of course things are rarely at "best" status.
The kind of design you have set out to make available as PCB Kit for 200 Bucks - well you can buy the same sort of e-bay for the same money, but fully assembled and with a nice case, made in China. And the Wima "Foil" capacitors they use are likely the same fakes you would end up buying.
Zhaolu D2.5
So why waste time replicating this kind of thing very laboriously and at greater cost?
IF I where to go out on a limb to actually give you some advise and if sound quality is of ANY concern at all (instead of matching marketing propaganda), I'd say:
Use a better DAC (almost ANY DAC will be better, one with a "slow rolloff" filter option is preferred). Have a read here at DIYA to see what others feel works well, for starters. Even better, why not make the DAC a plugin module with PCB options for a range of popular chips as "breakoff" parts on the side of the main PCB?
Use a better Receiver (almost ANY other available Receiver will be better, including Cirrus Logic 8414). Implement the receiver input correctly. This usually means either brutally direct or with some really clever circuitry (Pat "Jocko Homo" DiGiaccomo discussed and shared some here before he had enough and buggered off). From experience, correctly done the "clever circuitry" works better than "brutally simple", but "brutally simple" works much better than pretty much anything else.
Do not use an ASRC, if you must remove jitter, do it the correct way, with a secondary PLL.
There are several usable published examples, but nothing has ever been implemented that is directly accessible for the average DIY'er. Tent Labs has nice VCXO's for the Job, low noise power supplies and low noise PLL loop filters can be a challenge to implement, but if done right jitter can be quite effectively locked out.
This could be a plug in module or maybe implemented as something people can leave unpopulated if the expense is an issue, so DIY'er have the option.
Use better powersupplies than 317/337 per application note, they are so 1980's. Here on DIYAudio there are many published options that do better and do not cost appreciably more, I am kind of partial to something called "Teddyreg" for ease/performance ratio, but there are many others.
Do not use Op-Amp output Stages, try something like this: http://www.diyaudio.com/forums/attachments/digital-line-level/141182d1253190252-few-questions-pcm1793-based-usb-dac-sacdenhancer_evo3.pdf costs less and sounds better. Using J-Fets might get even better, but starts costing real money, one could provide footprints allowing a cheap Bipolar version and a more expensive J-Fet option.
Anyway, those are my not particularly humble opinions on what would produce a DAC that would give the DIY(A) community something worth building, rather than to instead just browse e-bay for an already build unit for less.
And if you can make it so that this "worth building DAC" can be build for 200 Bucks in parts cost and PCB set cost together, all the better.
Ciao T
As I am responsible for some of the choices made for this project, I´ll try to comment on your points.
@1
Yes let´s do that.
@2
At my place with state of the art gear.
Send me a PB with your mail adress and I´ll send you pictures of the set-up.
And of course it was also tested elsewhere.
@3
We tested AD, BB and CS.
After choosing, we´ve tried out more makes.
We still find CS pretty compettitive.
@4
CS8416 does have better specs than 8414, especially after 2004 revision.
@5
If you do not see the point in upsampling, I really think this project is of no interest to you. It is somewhat of topic of this thread, to explain why we chose upsampling, but I recommend that you read werewolfs thread about upsampling. He explains why upsampling is attractive at all.
Even if you do not belive that upsampling in any way can improve a digital signal, it still might be good reading for you😀
@6
The point about "any DAC is better than CS4398" I will not comment on.
@7
The same goes for the reciever
@8
And the same goes for "not using an ASRC"
@9
Long ago we tried "discrete high precission clocks"
They were discarded, and we think we spend time enough on this kind of stuff.
@10 and on
We did that already in the earlier design.
We did already take it to the limits, but a project like that is not suitable for DIY, because of compact layout and complex design.
This project is meant to take it as far as it gets for 200$ with all parts available from 1 vendor and with everything on one board.
Separating it on more boards is IMHO a very bad idea, seen from both a sonic and an electronic point of view.
We might change the analog stage from IC to discrete though.
As Kurt von Kubik says, we have tested a lot of DAC chips. I see that many users pick the PCM1794A. I think I know why... Basically because it has the best specifications. That's also the reason we tried it out.... But in sonic performance, it really failed against the CS4398.
The CS4398 really surpriced us (Mostly KvK, since I have been working with CS chips almost 15 years now). KvK has always liked the multibit DAC's from BB. That's why we tested against PCM1704K, which also failed in sonic performance against CS4398.
However... CS4398 can be a little picky about PSU, decoupling and layout. When we got that right in our last project, it really performs great.
If you ever come to Denmark, feel free to pay us a visit an enjoy this DAC. Except a small group of "Tube Fanatics", anyone who listened to it, has been amazed.
About using op-amps....: We both agree that this will not lead to a state of ar DAC. We choose the op-amp approach due to the DIY-friendly-ness... But we decided to go for 100% discrete No NFB design. And we still have a target price of 199 USD all included!
But but but.... If you want to help on this project, why don't you tell us which SPDIF receiver and DAC chip to use, instead of just saying that we don't use the best. Tell us what to use, and we will do some listening-tests to verify the performance.
Hi all,
First about the SPDIF issues: there are a few good threads around, like this one http://www.diyaudio.com/forums/digital-source/67247-s-pdif-digital-output.html. Especially useful are the links provided to the DiyHifi.org forum. The Newawa (22083) seems to be optimized for 2MHz rates while the SC transformers are...well, backed by marketing rules 🙄.
One proposition: what if a TX section as shown there is also included on the pcb (with a V-score around it which would make it easily detachable) so that those with not so good transports will not be at the mercy of a cr@ppy link. It may give more consistent results provided that care is taken by each one when choosing the cables/BNC connectors,...heck these might even be 'included' in the kit. This would certainly set the project apart from the chinese offerings
For the analog stage I vote for the idea of having a by-passable OPA1632 for those who want a balanced output. Then a well designed analog filter is a must. For the final result, IMHO, it doesn't matter if an op-amp or a discrete design is used as much as the filter order, characteristic and cut-off freq. I would even wager on the idea that more than 50% of the sonic quality depends on these last three parameters. If needed I could be of help here...
First about the SPDIF issues: there are a few good threads around, like this one http://www.diyaudio.com/forums/digital-source/67247-s-pdif-digital-output.html. Especially useful are the links provided to the DiyHifi.org forum. The Newawa (22083) seems to be optimized for 2MHz rates while the SC transformers are...well, backed by marketing rules 🙄.
One proposition: what if a TX section as shown there is also included on the pcb (with a V-score around it which would make it easily detachable) so that those with not so good transports will not be at the mercy of a cr@ppy link. It may give more consistent results provided that care is taken by each one when choosing the cables/BNC connectors,...heck these might even be 'included' in the kit. This would certainly set the project apart from the chinese offerings
For the analog stage I vote for the idea of having a by-passable OPA1632 for those who want a balanced output. Then a well designed analog filter is a must. For the final result, IMHO, it doesn't matter if an op-amp or a discrete design is used as much as the filter order, characteristic and cut-off freq. I would even wager on the idea that more than 50% of the sonic quality depends on these last three parameters. If needed I could be of help here...
We did try op-amp vs discrete in our last DAC, with the same cut off freq. ANd still with a HUGE difference pointing towards discrete design.
But feel free to post ideas on the filter.... Anything will be taken into account...
But feel free to post ideas on the filter.... Anything will be taken into account...
Thorsten, if you don't like the idea of this thread, then please don't participate. This kind of attitude spoils the forum in my opinion.
Kurt & Hurtig,
When you evaluated the DACs, did you use vendor evaluation boards with all other ancillary equipment the same?
I ask because there is a lot of disagreement about op-amps and I suspect it is because each op-amp needs to be specifically designed for its task. The idea that you can just swap op-amps in a standard circuit and pick the best doesn't necessarily lead to the best performance in a particular design.
Using the eval boards would at least level the playing field as each vendor optimizes their eval kits to show the product in it's best light.
When you evaluated the DACs, did you use vendor evaluation boards with all other ancillary equipment the same?
I ask because there is a lot of disagreement about op-amps and I suspect it is because each op-amp needs to be specifically designed for its task. The idea that you can just swap op-amps in a standard circuit and pick the best doesn't necessarily lead to the best performance in a particular design.
Using the eval boards would at least level the playing field as each vendor optimizes their eval kits to show the product in it's best light.
Hi,
Given how narrow your field for testing was I am unsurprised that you found the CS performance to be competive. There are some ACTUAL MARKET LEADERS when it comes to performance who you excluded.
Unlikely, I'm afraid, plus I am likely one of the "small group of "Tube Fanatics".
However, my point stands, why not give the user the choice of differential voltage output DAC with a plugin Module? And provide small module PCB's for the CS Chip and the other alternatives, including those you never tested?
But DESPITE providing better "paper" specs it still MEASURES much more jitter than anything else out there.
I am familiar with the arguments. I have been listening again and again to upsampling. I have helped design commercial CD-Players that include (selectable) upsampling. I find that ususally the sound is better without.
Here is why:
1) Basic Information Theory - there is no way to add real information, only distortion. ANY ASRC must add "misinformation". So why would I want to reduce and distort the actual information from my CD/DVD etc?
2) Filter Quality - as a rule, the digital filter in the ASRC becomes dominant in the system in which it works. This means even if the DAC's have more advanced digital filters that provide better sound quality their internal filters will be downgraded to what is in the ASRC.
For example, if you tested all different DAC"s with the ASRC in place I would expect many of the really significant sonic differences to be ereased.
3) Jitter - ASRC's cannot remove jitter, they can only translate it from one domain into another. How this effects audibility is debatable, but given that it is relatively easy to solve the issue of jitter, why use ASRC's?
This is not what I am advocating.
What I am suggesting is reclocking using a secondary PLL.
For example, the CS8416 has an INTRINSIC jitter exceeding 500ps unless it's "new" Phase Detecter Mode is invoked, in which case it has an intrinsic Jitter of around 120 - 200pS depending on sample rate.
The CS8416 does only reject incomming jitter above 10KHz. (BTW, the "market leading" monolithic receiver has 50pS intrinsic jitter and rejects jitter above 100Hz, meaning a 40db or 100 fold improvement in jitter rejection at 10KHz over the Cirrus part).
So the jitter presented to the DAC is a combination of incoming jitter (which may be in the region of several nanoseconds) plus the intrinsic jitter (the jitter present even with a jitter free source).
This jitter needs to be removed if we do not whish to reduce the performance of our DAC (be it CS4398 or whatever I favour).
One way is to use a secondary PLL with a turnover frequency that is really low (fraction of Hz). This is what I was recommending.
Of course, the elements within this secondary PLL need to have very good jitter performance, which generally translates into very low noise in all areas. It is not about "super clocks" or the like.
But 200 bucks buy assembled DAC's using the same chips and basic design features in a nice case. Where is the point making the same thing in DIY? I can understand using DIY to make a home-made "super-car" or "racing" car that is pretty much unavailable or unaffordable commercially.
Byut why try making a KIA from scratch for more money than it costs to buy a commercially made Kia?
From an electronic point of view only, using a small PCB to hold the DAC and local PSU decoupling can give most of the benefits of 4-layer PCB's using (much cheaper) 2-Layer PCB's. So I do not think the issue is as clearcut as you suggest.
Now there is a good idea to at least elevate the performance of this project above cheaper to buy e-bay DAC's...
On the receiver side, do you have the neccesary test gear to measure the jitter performance? There are a range of receivers from many sources.
The "best" would mean to implement the SPDIF receiver in a CPLD or FPGA and using VCXO's for providing the clocks required for receiver. Such a design would be in essence jitter-free, to the limits of the VCXO jitter, which is mainly determined by the power supplies.
Equally, the best DAC would most likely be FPGA based and combine a super fast multibit core with PDM/PWM for the lower bits.
FPGA's are also a good way to implement Digital Filters with responses that are, shall we say, less than common but beneficial to the sonic performance.
You are welcome to try it that way. It would truely be a "DIY" DAC and likely an exceptional one... ;-)
Ciao T
Given how narrow your field for testing was I am unsurprised that you found the CS performance to be competive. There are some ACTUAL MARKET LEADERS when it comes to performance who you excluded.
Unlikely, I'm afraid, plus I am likely one of the "small group of "Tube Fanatics".
However, my point stands, why not give the user the choice of differential voltage output DAC with a plugin Module? And provide small module PCB's for the CS Chip and the other alternatives, including those you never tested?
But DESPITE providing better "paper" specs it still MEASURES much more jitter than anything else out there.
I am familiar with the arguments. I have been listening again and again to upsampling. I have helped design commercial CD-Players that include (selectable) upsampling. I find that ususally the sound is better without.
Here is why:
1) Basic Information Theory - there is no way to add real information, only distortion. ANY ASRC must add "misinformation". So why would I want to reduce and distort the actual information from my CD/DVD etc?
2) Filter Quality - as a rule, the digital filter in the ASRC becomes dominant in the system in which it works. This means even if the DAC's have more advanced digital filters that provide better sound quality their internal filters will be downgraded to what is in the ASRC.
For example, if you tested all different DAC"s with the ASRC in place I would expect many of the really significant sonic differences to be ereased.
3) Jitter - ASRC's cannot remove jitter, they can only translate it from one domain into another. How this effects audibility is debatable, but given that it is relatively easy to solve the issue of jitter, why use ASRC's?
This is not what I am advocating.
What I am suggesting is reclocking using a secondary PLL.
For example, the CS8416 has an INTRINSIC jitter exceeding 500ps unless it's "new" Phase Detecter Mode is invoked, in which case it has an intrinsic Jitter of around 120 - 200pS depending on sample rate.
The CS8416 does only reject incomming jitter above 10KHz. (BTW, the "market leading" monolithic receiver has 50pS intrinsic jitter and rejects jitter above 100Hz, meaning a 40db or 100 fold improvement in jitter rejection at 10KHz over the Cirrus part).
So the jitter presented to the DAC is a combination of incoming jitter (which may be in the region of several nanoseconds) plus the intrinsic jitter (the jitter present even with a jitter free source).
This jitter needs to be removed if we do not whish to reduce the performance of our DAC (be it CS4398 or whatever I favour).
One way is to use a secondary PLL with a turnover frequency that is really low (fraction of Hz). This is what I was recommending.
Of course, the elements within this secondary PLL need to have very good jitter performance, which generally translates into very low noise in all areas. It is not about "super clocks" or the like.
But 200 bucks buy assembled DAC's using the same chips and basic design features in a nice case. Where is the point making the same thing in DIY? I can understand using DIY to make a home-made "super-car" or "racing" car that is pretty much unavailable or unaffordable commercially.
Byut why try making a KIA from scratch for more money than it costs to buy a commercially made Kia?
From an electronic point of view only, using a small PCB to hold the DAC and local PSU decoupling can give most of the benefits of 4-layer PCB's using (much cheaper) 2-Layer PCB's. So I do not think the issue is as clearcut as you suggest.
Now there is a good idea to at least elevate the performance of this project above cheaper to buy e-bay DAC's...
On the receiver side, do you have the neccesary test gear to measure the jitter performance? There are a range of receivers from many sources.
The "best" would mean to implement the SPDIF receiver in a CPLD or FPGA and using VCXO's for providing the clocks required for receiver. Such a design would be in essence jitter-free, to the limits of the VCXO jitter, which is mainly determined by the power supplies.
Equally, the best DAC would most likely be FPGA based and combine a super fast multibit core with PDM/PWM for the lower bits.
FPGA's are also a good way to implement Digital Filters with responses that are, shall we say, less than common but beneficial to the sonic performance.
You are welcome to try it that way. It would truely be a "DIY" DAC and likely an exceptional one... ;-)
Ciao T
Hi,
Actually, I like the idea of this thread, namely "lets make a open source DIY DAC that is really good".
I just feel that the way this going now it is going to end up as something that already exists in a dozend cheap chinese DAC's that can bought ready to run for less than the project would cost.
So I am suggesting to the authors of this "open source" project that it may be worth to raise the profile to something that you will be still able to build for a few 100 Bucks but which you would never find for the same money "well done" on E-Pay.
Do you really think the world needs ANOTHER CS8416/SRC4192/CS4398/Op-Amp DAC? Less time than I spend on this thread so far can buy this for very little cash.
Ciao T
Actually, I like the idea of this thread, namely "lets make a open source DIY DAC that is really good".
I just feel that the way this going now it is going to end up as something that already exists in a dozend cheap chinese DAC's that can bought ready to run for less than the project would cost.
So I am suggesting to the authors of this "open source" project that it may be worth to raise the profile to something that you will be still able to build for a few 100 Bucks but which you would never find for the same money "well done" on E-Pay.
Do you really think the world needs ANOTHER CS8416/SRC4192/CS4398/Op-Amp DAC? Less time than I spend on this thread so far can buy this for very little cash.
Ciao T
Hi Kurt,
Please have a look how you use the quotes, it does not seem to work as you intended... I hope I picked out your points correctly.
While I personally would not exclude Wolfson so dismissively, some excellent vendors you have not discussed:
Nippon Precision Circuits
Niigata Seimitsu
ESS
Others exist.
Well, "unbroken" signal path are not such a big deal. If we are talking the digital signals, it is by far more relevant to control reflections on the lines and ground bounce, than any "unbroken path".
Power lines in fact SHOULD be broken with suitable inductances and have their low noise and impedance established locally suing good bypassing practice, the opposite is likely to cause all sorts of funny business (just poke a 'scope with a few 100MHz bandwidth around the powerlines to see).
For the analog side, if your analogue stage has a very high input impedance any current flow on the signal lines is minimised and any "signal path breaks" are reduced in magnitude accordingly.
It is all a tradeoff and many ways exist to get what is desired.
Actually, I do not.
I was quoting the measurements from the exact article you reference. If you where to look at the actual data, instead only at the advertising pamphlet writing that selectively derives conclusions from the data it would be obvious.
As all DS DAC's fold back jitter above the baseband into the baseband the jitter peaking of the 8416 above 40KHz cannot be ignored, as is done in the paper.
Further, if the default setting for the Phase detector (PDUR = 0) is used the jitter is about 4 times (12dB) greater than the 8414. And if you set PDUR = 1 in hardware you loose the > 96KHz sample rates.
So you need to provide a MCU that handles sample rate detection and sets PDUR = 1 explicitly for single and double speed sample rates. Even then as noted, the jitter of the 8416 remains larger than the 8414, as illustrated in the reduction of the dynamic range with the 8416.
So, I do not need to disprove the measurements at all, instead maybe you should look a LITTLE closer at the measurements you cite (and the datasheet).
Exactly what I said. In the Time Domain jitter is removed. However, it has not exactly gone away? How could it? So the question is where the jitter has ended up. It is illuminating to use the AP Two's "generate jitter" function and to then measure the results...
The values are in the paper you referenced. They appear if the CS8416 is used in "default" configuration and are marked as "PDUR = 0". Unless you explicitly reconfigure the CS8416 in software or with a hardware pull up on the TX pin the chip will be in PDUR = 0 and hence jitter will be high.
How so? Writing the code for a FPGA and soldering it to a PCB is surely "DIY".
Ciao T
Please have a look how you use the quotes, it does not seem to work as you intended... I hope I picked out your points correctly.
While I personally would not exclude Wolfson so dismissively, some excellent vendors you have not discussed:
Nippon Precision Circuits
Niigata Seimitsu
ESS
Others exist.
Well, "unbroken" signal path are not such a big deal. If we are talking the digital signals, it is by far more relevant to control reflections on the lines and ground bounce, than any "unbroken path".
Power lines in fact SHOULD be broken with suitable inductances and have their low noise and impedance established locally suing good bypassing practice, the opposite is likely to cause all sorts of funny business (just poke a 'scope with a few 100MHz bandwidth around the powerlines to see).
For the analog side, if your analogue stage has a very high input impedance any current flow on the signal lines is minimised and any "signal path breaks" are reduced in magnitude accordingly.
It is all a tradeoff and many ways exist to get what is desired.
Actually, I do not.
I was quoting the measurements from the exact article you reference. If you where to look at the actual data, instead only at the advertising pamphlet writing that selectively derives conclusions from the data it would be obvious.
As all DS DAC's fold back jitter above the baseband into the baseband the jitter peaking of the 8416 above 40KHz cannot be ignored, as is done in the paper.
Further, if the default setting for the Phase detector (PDUR = 0) is used the jitter is about 4 times (12dB) greater than the 8414. And if you set PDUR = 1 in hardware you loose the > 96KHz sample rates.
So you need to provide a MCU that handles sample rate detection and sets PDUR = 1 explicitly for single and double speed sample rates. Even then as noted, the jitter of the 8416 remains larger than the 8414, as illustrated in the reduction of the dynamic range with the 8416.
So, I do not need to disprove the measurements at all, instead maybe you should look a LITTLE closer at the measurements you cite (and the datasheet).
Exactly what I said. In the Time Domain jitter is removed. However, it has not exactly gone away? How could it? So the question is where the jitter has ended up. It is illuminating to use the AP Two's "generate jitter" function and to then measure the results...
The values are in the paper you referenced. They appear if the CS8416 is used in "default" configuration and are marked as "PDUR = 0". Unless you explicitly reconfigure the CS8416 in software or with a hardware pull up on the TX pin the chip will be in PDUR = 0 and hence jitter will be high.
How so? Writing the code for a FPGA and soldering it to a PCB is surely "DIY".
Ciao T
@ ThorstenL
It seems you are more into talking bad about the project, than participating in the making of this project. Please consider, if you have anything to add to the project. If you do, feel free to post real ideas, instead of just telling what you don't like.
If you are one of these people, who no matter how i sounds, will newer like a product without tubes, I guess this project will newer really be the best for you 😉
About DAC chips, we did not feel that NPC had anything relevant for this project. Niigata Seimitsu I do not know... We have a ESS DAC, based on ES9018 and AD797 op-amps.
Still we pick CS4398 as the best sonic performer.
A good sign... Our first prototype of our "Reference DAC" based on CS4398, was lend to a Mark Levinson owner. After a few days, he was ready to change his 10.000 USD ML DAC into our prototype. Guys the CS4398 cannot be that bad...
When that's over, I have to say that the difference between these DAC chip isn't that great. The most of the performance, in achieved in the rest of the design. And that is why I don't see why we should keep on discussing the DAC-chip.
The analog stage will NOT be a standard op-amp design. We will go for a No NFB design!
About those chineese DAC's... If you feel these DAC's will suit you, then maybe you should buy one of them.
It seems you are more into talking bad about the project, than participating in the making of this project. Please consider, if you have anything to add to the project. If you do, feel free to post real ideas, instead of just telling what you don't like.
If you are one of these people, who no matter how i sounds, will newer like a product without tubes, I guess this project will newer really be the best for you 😉
About DAC chips, we did not feel that NPC had anything relevant for this project. Niigata Seimitsu I do not know... We have a ESS DAC, based on ES9018 and AD797 op-amps.
Still we pick CS4398 as the best sonic performer.
A good sign... Our first prototype of our "Reference DAC" based on CS4398, was lend to a Mark Levinson owner. After a few days, he was ready to change his 10.000 USD ML DAC into our prototype. Guys the CS4398 cannot be that bad...
When that's over, I have to say that the difference between these DAC chip isn't that great. The most of the performance, in achieved in the rest of the design. And that is why I don't see why we should keep on discussing the DAC-chip.
The analog stage will NOT be a standard op-amp design. We will go for a No NFB design!
About those chineese DAC's... If you feel these DAC's will suit you, then maybe you should buy one of them.
- Status
- Not open for further replies.