Terry, good point for Thorp re close-in phase noise issues with Si clocks. There are maybe some other things that could be explained as well. I am trying for an asynchronous dac that sounds as good or better than the typical synchronous implementations with PCM audio. I also want to be able to accept SPDIF inputs (which are always asynchronous), yet with exceptionally good sound qualily too, and without introducing delays such as FIFOs may bring. There are ways to do such things, but they are not all known among diy'ers. Only know of a small number of proprietary dac designs that have solved the problems. (Even for Benchmark DAC-3, I recently found a low-level shortcoming at least one other person has noticed, although it sounds so good in most respects most people would probably not notice, and I haven't heard any other talk of it.)
@Thorp,
Congratulations on the good results and listening satisfaction you report for your K2M dac. Also, appreciate the additional software description beyond what is explained by the posted schematic. I hope you will choose to hang out around here with us at least some of the time as we continue on our dac modding journey. You seem like a well-spoken and thoughtful designer, and I can understand feeling good when hearing good results for the first time.
Congratulations on the good results and listening satisfaction you report for your K2M dac. Also, appreciate the additional software description beyond what is explained by the posted schematic. I hope you will choose to hang out around here with us at least some of the time as we continue on our dac modding journey. You seem like a well-spoken and thoughtful designer, and I can understand feeling good when hearing good results for the first time.
@Terry
I agree, both jitter and phase noise of the oscillator should be as low as possible, and also they are lower as the output frequency is lower for the same type of chip.
The reasons why I have chosen Si549:
- for a synchronous clock I need two output MCLK frequencies (for 44.1kHz and 48kHz music files)
- if I would chose 2 osc. chips with insane low jitter and phase noise, it might be possible that required hardware for switching between them to be much noisier and jitter generator than a single low specs. osc. chip.
- I am not so sure about what you said about the Si549 chip, but its jitter is specified as low as 200 fs (typical) for CMOS version that I use (95 fs for differential ones). It is the lowest jitter and phase noise I2C adjustable oscillator I have found. Do you have more information about it, other than in its datasheet?
- even 200-300 fs jitter osc. chip requires a very professional designed, manufactured and verified PCB in order to get its timing precision. More than that, there are some decoupling caps (i.e. respecting some distances for the projected PCB traces) which cannot be soldered as a diy-er at home: 0805 smd size it's OK, 0603 is tricky, but smaller is almost impossible to solder at home.
@Mark
Thank you for your congratulations and also thank you for the valuable information you have been sharing with us. I will also try to share the information I will find, as often as I can.
Regarding the SPDIF inputs, I do not have so much knowledge, but maybe in the future I will start to learn about them. Anyway, my XMOS USB to I2S adapter is SPDIF out ready, so at least I have a test source.
Returning to K2M project:
I don't have any doubts, the best sound is achieved when the sync_mode registry bit is set to 1 - it enables quick lock when fs and MCLK are synchronous. In this case, adjusting the DPLL has no effect since it is by-passed.
I had to modify my code because it used the K2M registry to read the incoming signal sample rate for display it, but when sync_mode is 1, the 4 registries who store the incoming sample rate are empty. At this moment, XMOS board not only switch the MCLK from 90.xxx to 98.xxx, but also it provides the information for sample rate display.
The first issue I have found is that the 1.3V power supply didn't worked. Reading the DAC's registry were tricky for 96+ kHz incoming signal and it took me a while until I figured out. Re-soldering LT3045 solved that problem.
The second issue is that the K2M chips don't like MCLK change (from 90xxx to 98xxx). From the read-only registries I get sometimes strange values, from both dac chips, sometimes from only one dac chip, sometimes from the other. Only when the sync_mode is 1, all the registries have the expected values.
I am still working on that. After I will get a "free of bugs" code, I will try different timing delays and/or hardware reset, software reset and other registry settings to find what is going wrong.
I agree, both jitter and phase noise of the oscillator should be as low as possible, and also they are lower as the output frequency is lower for the same type of chip.
The reasons why I have chosen Si549:
- for a synchronous clock I need two output MCLK frequencies (for 44.1kHz and 48kHz music files)
- if I would chose 2 osc. chips with insane low jitter and phase noise, it might be possible that required hardware for switching between them to be much noisier and jitter generator than a single low specs. osc. chip.
- I am not so sure about what you said about the Si549 chip, but its jitter is specified as low as 200 fs (typical) for CMOS version that I use (95 fs for differential ones). It is the lowest jitter and phase noise I2C adjustable oscillator I have found. Do you have more information about it, other than in its datasheet?
- even 200-300 fs jitter osc. chip requires a very professional designed, manufactured and verified PCB in order to get its timing precision. More than that, there are some decoupling caps (i.e. respecting some distances for the projected PCB traces) which cannot be soldered as a diy-er at home: 0805 smd size it's OK, 0603 is tricky, but smaller is almost impossible to solder at home.
@Mark
Thank you for your congratulations and also thank you for the valuable information you have been sharing with us. I will also try to share the information I will find, as often as I can.
Regarding the SPDIF inputs, I do not have so much knowledge, but maybe in the future I will start to learn about them. Anyway, my XMOS USB to I2S adapter is SPDIF out ready, so at least I have a test source.
Returning to K2M project:
I don't have any doubts, the best sound is achieved when the sync_mode registry bit is set to 1 - it enables quick lock when fs and MCLK are synchronous. In this case, adjusting the DPLL has no effect since it is by-passed.
I had to modify my code because it used the K2M registry to read the incoming signal sample rate for display it, but when sync_mode is 1, the 4 registries who store the incoming sample rate are empty. At this moment, XMOS board not only switch the MCLK from 90.xxx to 98.xxx, but also it provides the information for sample rate display.
The first issue I have found is that the 1.3V power supply didn't worked. Reading the DAC's registry were tricky for 96+ kHz incoming signal and it took me a while until I figured out. Re-soldering LT3045 solved that problem.
The second issue is that the K2M chips don't like MCLK change (from 90xxx to 98xxx). From the read-only registries I get sometimes strange values, from both dac chips, sometimes from only one dac chip, sometimes from the other. Only when the sync_mode is 1, all the registries have the expected values.
I am still working on that. After I will get a "free of bugs" code, I will try different timing delays and/or hardware reset, software reset and other registry settings to find what is going wrong.
No I don't have specific phase noise plots or Allan Variance for this oscillator.
Si Labs did publish phase noise plots for some of their other programmable oscillators years ago and their performance, from memory was slightly inferior to 549, however where they did fall down was low frequency phase noise, being significantly worse than a really good XO.
Also the jitter spec of 200 / 95fsec is above 12kHz, which is not important for audio. You want very good LF to MF jitter and this is where these clocks are not comparable to fixed freq crystal XO's. However, they are certainly very good and getting a lot better.
You can use a fairly simple switching arrangement with 2 chips for Crystek / NDK oscillators. Check the FIFO thread, it is a clock driver chip with OP disable. The jitter is really low and wont degrade the XO to any degree.
The advantage is you can try other clocks.
The best approach is to slave the DAC directly off USB bridge clock(s). This is a much simpler arrangement and way superior to any SPDIF setup except a FIFO with memory buffer.
This is definitely the way to go.
If you can get an XMOS board with 45 / 49MHz clocks, this is the best solution. Most XMOS boards provide a switched clock OP that is there to use.
T
Nobody said good SPDIF is easy, but it is a requirement for some applications.
Regarding FIFO approaches, they are good for for music player only dacs. Not so good for editing or playing video in sync with sound or for some other applications.
Another easy way that works for hobbyists to get better sound quality is to subscribe to Rune and use their very-high-tap-number filters to produce very good quality PCM to DSD conversion. Takes care of the PCM interpolation filtering sound quality issues, which always remains as a sound quality limitation if jitter is the only thing that is addressed.
If one wants to make a good general purpose dac, good for SPDIF, good for video, and that doesn't require pre-processing one's entire ripped library into DSD, then other approaches have to be developed. One can't expect diyaudio-hobbist-music-player dacs to accurately represent the market for commercial dacs. But, for some limited use cases I would agree that well-understood approaches to dac design are already available.
Yes - I come from pro audio background, I know it well.
When asynchronous USB came along it was a great time for digital audio.
A lot of things were solved, then it was just a matter of correct implementation.
True. Just installed a couple of FIFO type arrangements for customers. They sound really good but the delay is something you can or can't live with.
YES:::
As I suggested a long time ago Mark, HQplayer is a super HQ format changer / upsampler. The general consensus ATM is Roon is definitely getting much better but HQplayer is probably best out there. There are now other very good ones but I'm slightly out of the loop lately. It depends. If you need SPDIF to play CD's, play them in PC CD player or rip them to HDD.
The single choice of SPDIF compatibility adds potentially big compromises.
If you can at all do without it, the whole game changes and in a good way.
Terry
Omitting SPDIF still leaves open issues with audio for video playback, and for video and audio editing. If you can't hear the audio accurately, how can you edit (e.g. mix & master) to best effect?
Also, you can't expect most consumers who want great sound quality to fool around with ripping, music servers, etc., if they don't feel comfortable with that tech-y kind of stuff. Some people will very understandably want really good dacs that just work, and that cost less than Chord Dave. How to make inroads there? I still think more work needs to be done, and it needs to be not solely proprietary.
Thank you for clarification and for your advices.
At this moment I use some XMOS USBtoI2S adapters which I designed a couple of years ago. I had to modify them in order to work as slave and for accepting 90.xxxx/98.xxxx frequencies (divided by 2) from Si549. The oscillator/s on the USB board are useless now.
Attachments
Don't know about HQPlayer. Read somewhere Rune recently introduced some very long FIR filters to emulate what is in Chord Dave. Some people seem to be liking that. Don't know if HQPlayer keeps up or is static. Then there is the old Weiss Saracon offline SRC. Pretty hard to beat too. Distortion is low enough that it should be inaudible at -160dB, or something like that. Costs a lot more than HQPlayer or Rune, however.
HQplayer has been actively developed for many years and continues to be so.
Here's a discussion on the M scaler. HQplayers designer's name on the forum is 'Miska'. He's quite a genius when it comes to audio software. There are other threads, even one on HQplayer itself.
Chords New M -Scaler - DAC - Digital to Analog Conversion - Audiophile Style
T
Also, I have to wonder what people do who want to use active digital cross-overs for their speakers. Obviously, if one is working towards good sound quality, one is probably not going to pre-cross-over and then pre-DSD convert music for every driver then send it out perfectly coherently to every power amp located locally at every speaker. As a practical matter we need good, low-cost dacs for people who want to use active speakers. Maybe they can use usb to some extent or maybe it will have to be SPDIF or AES. What then?
Last edited:
Hello Mark!
I want to ask you for advice. Is TPS7A4700 низкая Шум RF Питание модуль 3 В 3,3 В 5 В 12 В 15 В 1A регулируемый купить на AliExpress instead of LT3042 PSRR RF Linear Regulator Power Board Module Ultra low High Noise Board Reverse Battery Current Protection 20V 200mA-in Integrated Circuits from Electronic Components & Supplies on Aliexpress.com | Alibaba Group suitable for powering chip es9038q2m down the line AVCC? You recommended that LT3045 is not the best option. And according to datashita TPS7A4700 is more suitable for powering the DAC. Or do food at LT1763?
I want to ask you for advice. Is TPS7A4700 низкая Шум RF Питание модуль 3 В 3,3 В 5 В 12 В 15 В 1A регулируемый купить на AliExpress instead of LT3042 PSRR RF Linear Regulator Power Board Module Ultra low High Noise Board Reverse Battery Current Protection 20V 200mA-in Integrated Circuits from Electronic Components & Supplies on Aliexpress.com | Alibaba Group suitable for powering chip es9038q2m down the line AVCC? You recommended that LT3045 is not the best option. And according to datashita TPS7A4700 is more suitable for powering the DAC. Or do food at LT1763?
Hi Mark, i made AVVC opamp regulator 2x3.3V, which is supplied by lt3042, i must admit that now i hear better separation in high spectrum, for DVCC and rest DAC board supply 3.3V i channged board ams 1117 regulator for TPS7A4700.
Overall another step in SQ by these two regulators…. (both supply by separate power supply)
Thanks for tip with opamp regulation….
Next step will be to try new opamp current mode output stage according your schema .
Now i have my HDAM+Diamond out buffer (many times better than voltage board one…) but wanna
to try opamp solution fully in current mode…
Only question, which opamps you suggest to use, because i have at home some muses 8920, ad797, some opa but only one piece each…(these were used for testing on original poor voltage mode output stage),
I could use „better“ ones above (muses, opa, ad) for comon sum part left and right channel. For differential parts can i use „normal“ lme49720 ?
Also there should be some decoupling capacitors on each opamp 100nf on each leg 4 and 8 ?
A also have 2x(5x22uf) wima which i will use for common decoupling of input +-15V…
Thanks
Vita
Overall another step in SQ by these two regulators…. (both supply by separate power supply)
Thanks for tip with opamp regulation….
Next step will be to try new opamp current mode output stage according your schema .
Now i have my HDAM+Diamond out buffer (many times better than voltage board one…) but wanna
to try opamp solution fully in current mode…
Only question, which opamps you suggest to use, because i have at home some muses 8920, ad797, some opa but only one piece each…(these were used for testing on original poor voltage mode output stage),
I could use „better“ ones above (muses, opa, ad) for comon sum part left and right channel. For differential parts can i use „normal“ lme49720 ?
Also there should be some decoupling capacitors on each opamp 100nf on each leg 4 and 8 ?
A also have 2x(5x22uf) wima which i will use for common decoupling of input +-15V…
Thanks
Vita
Hi 888777,
Sorry, I have not tried LDO regulators for AVCC, but I have heard other people say that they like the sound of opamp supplies for AVCC better than they like the sound quality with LT3045 or LT3042. One dual opamp can used for AVCC using the schematic and tips ESS provides in the following document: http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
Also, we have a schematic that we like to use for AVCC that can be found in this thread in post #3003 https://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-301.html#post5577605
Another option some people like that may work well although I haven't tried it is ADM7150 or ADM7154, such as the AVCC-SR in the following page: Trident-SR
Sorry, I have not tried LDO regulators for AVCC, but I have heard other people say that they like the sound of opamp supplies for AVCC better than they like the sound quality with LT3045 or LT3042. One dual opamp can used for AVCC using the schematic and tips ESS provides in the following document: http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
Also, we have a schematic that we like to use for AVCC that can be found in this thread in post #3003 https://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-301.html#post5577605
Another option some people like that may work well although I haven't tried it is ADM7150 or ADM7154, such as the AVCC-SR in the following page: Trident-SR
Last edited:
Hi Vita,
I originally used LME49720 for the output stage and for AVCC. I think they can work well if the +-15v power supply is clean and there is not too much RF around them. They can sound a little bright which can be some distortion from local RF fields, such as from DECT wireless phone base stations. So long as not too much RF around I would probably try them first. I used dip sockets for the opamps in the 2nd dac board I modded to make it easy to see how I liked different opamps, which might be something to consider. Also, normally we use .1uf X7R ceramic and 10uf tantalum at each IC power pin for decoupling. Then I also usually find that about 100uf of parallel film caps can help too from each of the +-15v rails to ground near where output stage power enters the dac board.
In addition to the above, I use an external 5v power supply for the low voltage circuits, and put secondary regulators on the dac board to drop the 5v power down to +3.3v. I am using one LT1763 +3.3v regulator for each of DVCC, VCCA, and the clock.
We normally also upgrade the clock module as well.
I haven't posted a list for awhile, so I will attach below a list of posts you might want to look at. Some of them have pictures and explanations of what I have been working on over time with my two modded dacs.
Hope it is okay that I wrote a little more than what you asked about, since other people may be reading too and wanting to catch up on things.
I originally used LME49720 for the output stage and for AVCC. I think they can work well if the +-15v power supply is clean and there is not too much RF around them. They can sound a little bright which can be some distortion from local RF fields, such as from DECT wireless phone base stations. So long as not too much RF around I would probably try them first. I used dip sockets for the opamps in the 2nd dac board I modded to make it easy to see how I liked different opamps, which might be something to consider. Also, normally we use .1uf X7R ceramic and 10uf tantalum at each IC power pin for decoupling. Then I also usually find that about 100uf of parallel film caps can help too from each of the +-15v rails to ground near where output stage power enters the dac board.
In addition to the above, I use an external 5v power supply for the low voltage circuits, and put secondary regulators on the dac board to drop the 5v power down to +3.3v. I am using one LT1763 +3.3v regulator for each of DVCC, VCCA, and the clock.
We normally also upgrade the clock module as well.
I haven't posted a list for awhile, so I will attach below a list of posts you might want to look at. Some of them have pictures and explanations of what I have been working on over time with my two modded dacs.
Hope it is okay that I wrote a little more than what you asked about, since other people may be reading too and wanting to catch up on things.
Last edited:
thank you, Mark.
Schemes from the ECC or according to the post 3003 I can not realize yet due to the fact that I live in the village and we do not have shops with radio parts. Although this option is most interesting to me. Later I will have the opportunity, but this later ...
But I understand you are using LT1763? Your photos in the post 3006, they clearly show that it is they. They are also ldo. But at the same time you do not recommend them?
I would like to hear from you at least a brief explanation of why.
By the way, the DAC came to me from the Chinese. In spite of everything, opening the parcel, I saw exactly SM PCB v. 1.07. With display. The sound is better than the DAC PCM 5102) One feels that food plays an important role. Unfortunately, I made a mistake and burned the DAC chip. Already ordered another.
Schemes from the ECC or according to the post 3003 I can not realize yet due to the fact that I live in the village and we do not have shops with radio parts. Although this option is most interesting to me. Later I will have the opportunity, but this later ...
But I understand you are using LT1763? Your photos in the post 3006, they clearly show that it is they. They are also ldo. But at the same time you do not recommend them?
I would like to hear from you at least a brief explanation of why.
By the way, the DAC came to me from the Chinese. In spite of everything, opening the parcel, I saw exactly SM PCB v. 1.07. With display. The sound is better than the DAC PCM 5102) One feels that food plays an important role. Unfortunately, I made a mistake and burned the DAC chip. Already ordered another.
Hi 888777,
The reason people say is because the sound quality seems a bit flat and lifeless with LDOs for AVCC, but I don't know why. I have not tried to investigate it. I would suggest then to try maybe TPS7A4700 and see how you like it. I think it will be better than the way it now without any dedicated AVCC regulator. Ideally, each AVCC channel should have its own regulator rather then using one regulator for both, but again, it might sound better with one shared regulator than the way it is now.
Regarding LT1763, I have only been using those for non-AVCC 3.3v regulators and I always use them with a .01uf C0G cap as described in the data sheet for noise reduction. Probably lower noise is best for AVCC, if possible.
The reason people say is because the sound quality seems a bit flat and lifeless with LDOs for AVCC, but I don't know why. I have not tried to investigate it. I would suggest then to try maybe TPS7A4700 and see how you like it. I think it will be better than the way it now without any dedicated AVCC regulator. Ideally, each AVCC channel should have its own regulator rather then using one regulator for both, but again, it might sound better with one shared regulator than the way it is now.
Regarding LT1763, I have only been using those for non-AVCC 3.3v regulators and I always use them with a .01uf C0G cap as described in the data sheet for noise reduction. Probably lower noise is best for AVCC, if possible.
Last edited:
so this is the problem ldo - a kind of lifeless sound ...
I still try tps7a4700. Given the fact that he is positioned on the datasheet and to power the DAC. Fast and clean. Unlike lt 3042/45, which, according to the data sheet, there is no word on the power supply of the DAC.
There is an aliexpress based on ne5534. HiFi двойной мощность выход Ультра низкий уровень шума линейный регулятор мощность класс A В 12 В 5 в Core питание для усилитель цифро аналогового преобразователя усилители домашние купить на AliExpress But there is no scheme and you can’t see what's what on the photo ... So I’m not risking buying.
I still try tps7a4700. Given the fact that he is positioned on the datasheet and to power the DAC. Fast and clean. Unlike lt 3042/45, which, according to the data sheet, there is no word on the power supply of the DAC.
There is an aliexpress based on ne5534. HiFi двойной мощность выход Ультра низкий уровень шума линейный регулятор мощность класс A В 12 В 5 в Core питание для усилитель цифро аналогового преобразователя усилители домашние купить на AliExpress But there is no scheme and you can’t see what's what on the photo ... So I’m not risking buying.
And one more question. Does it make sense to buy something on lm317 / lm337 to power the analog part? Or until the output is redone, can you get by with 7815/7915?
or am I not looking there at all? and you have to initially think about the implementation of power on an operational amplifier? NE 5534, LME 49720 and so on? And do not possibly make unnecessary purchases?
or am I not looking there at all? and you have to initially think about the implementation of power on an operational amplifier? NE 5534, LME 49720 and so on? And do not possibly make unnecessary purchases?