Clocks on an Amanero board are buffered by the CPLD, thus jitter is not well minimized. Also, the clock power supplies might could be better.
Might it have been changing out the last two remaining ceramic bypass caps (for MCLK signal path ICs) and replacing them with MU caps?
Also, just wondering if you tried the Iancanada clock in the new V15 board yet? IIUC it helped significantly for the previous version board?
Yes, it is possible that the MU caps had an effect. Plus I'd removed the inverter that was previously in the clock line to the D flip-flips
I have not yet had chance to switch the Ian Canada scPure XO over to the v15 board - I have a few parallel activities going on so juggling things to get all run-in and ready for a less chaotic comparison.
However, I have kept my DAC1 unit going continuously without touching it since the scPure was installed about 2 weeks ago. This is my primary source of music all day in my home office.
I can honestly say the scPure is a big step up on the Crystek. At first switch on it sounded similar to the Crystek but I knew the recommendation from Ian Canada was not to listen for 3 days, however I ignored that. 😏 Fast forward to around 2-3 days and at that point I noticed a step-change. A quite obvious improvement. From that point it continued to improve but with more subtlety.
Sound-stage has gone wider and deeper (although I didn't think it could!) Everything just seems more realistic and more holographic. I'm thinking that both the HF and bass are still improving.
Once I get the v15, scPure and the "correct" filter board together in one place I am expecting great things! At that point I can also experiment with splitting the v15 and moving the clock part away from the USB and possibly screening it, etc.
I have not yet had chance to switch the Ian Canada scPure XO over to the v15 board - I have a few parallel activities going on so juggling things to get all run-in and ready for a less chaotic comparison.
However, I have kept my DAC1 unit going continuously without touching it since the scPure was installed about 2 weeks ago. This is my primary source of music all day in my home office.
I can honestly say the scPure is a big step up on the Crystek. At first switch on it sounded similar to the Crystek but I knew the recommendation from Ian Canada was not to listen for 3 days, however I ignored that. 😏 Fast forward to around 2-3 days and at that point I noticed a step-change. A quite obvious improvement. From that point it continued to improve but with more subtlety.
Sound-stage has gone wider and deeper (although I didn't think it could!) Everything just seems more realistic and more holographic. I'm thinking that both the HF and bass are still improving.
Once I get the v15, scPure and the "correct" filter board together in one place I am expecting great things! At that point I can also experiment with splitting the v15 and moving the clock part away from the USB and possibly screening it, etc.
Only one thing I would comment on with respect to the impressive changes with SCPure, holographic sound stage, a big step up, etc., is that there is another whole big step up possible with Acko oscillators and better squaring circuitry. Yet again more holographic than someone accustomed to SC Pure would probably imagine to be possible. This is part of what makes me think these @MarcelvdG RTZ DSD dacs have the potential to compete with some of the best dacs in the world.
I have been requested to summarize the dac setup shown in the 2nd pic as can be seen in the first post of this thread.
Basically, it involves linking to three or four different threads which are involved in the overall dac project.
1. First part is the dac board itself. Its in a long thread but the first post explains where to find schematics and Gerbers: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-6847540
2. Second part is the FPGA-based PCM2DSD converter: https://www.diyaudio.com/community/threads/simple-dsd-modulator-for-dsc2.370177/post-6593785
3. Third part includes the open source boards of this thread, or else the simplified and compact version at: https://www.diyaudio.com/community/...2soverusb-pcm2dsd-rtz-dac.423401/post-7917601
I would suggest to look over those threads to see what is involved. Obviously there is a lot of SMD soldering, so that may be one thing we would need to talk about.
Another thing would be the overall architecture, how things are connected together, and anything not explained in detail in the above threads, such as power supplies, USB board, etc.
Last thing is still a work in progress which is the output stage for the dac. There are some existing designs already, but more are being investigated. Will update on that as new information comes in.
Basically, it involves linking to three or four different threads which are involved in the overall dac project.
1. First part is the dac board itself. Its in a long thread but the first post explains where to find schematics and Gerbers: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-6847540
2. Second part is the FPGA-based PCM2DSD converter: https://www.diyaudio.com/community/threads/simple-dsd-modulator-for-dsc2.370177/post-6593785
3. Third part includes the open source boards of this thread, or else the simplified and compact version at: https://www.diyaudio.com/community/...2soverusb-pcm2dsd-rtz-dac.423401/post-7917601
I would suggest to look over those threads to see what is involved. Obviously there is a lot of SMD soldering, so that may be one thing we would need to talk about.
Another thing would be the overall architecture, how things are connected together, and anything not explained in detail in the above threads, such as power supplies, USB board, etc.
Last thing is still a work in progress which is the output stage for the dac. There are some existing designs already, but more are being investigated. Will update on that as new information comes in.
Last edited:
Forgot at add a link to the Acko clocks: https://sites.google.com/site/ackodac/modules-and-kits/clocks-and-oscillators Those are the best, truly SOA.
Please let me know if any questions about any or all of the above. If none, it would be helpful to know why not.
General questions would be fine too, like maybe, "I heard a DSD dac before and I didn't like the sound of it; is this just more of that?" To be very clear, no, its not its not just more of that.
Maybe interesting to know, Bruno Putzeys designed a dac called Mola Mola. It sells for around $12,000 for two channels. Most people consider Bruno to be a pretty serious engineer. Mola Mola is a DSD dac. In some ways its a lot like the diy one here I would recommend.
Please let me know if any questions about any or all of the above. If none, it would be helpful to know why not.
General questions would be fine too, like maybe, "I heard a DSD dac before and I didn't like the sound of it; is this just more of that?" To be very clear, no, its not its not just more of that.
Maybe interesting to know, Bruno Putzeys designed a dac called Mola Mola. It sells for around $12,000 for two channels. Most people consider Bruno to be a pretty serious engineer. Mola Mola is a DSD dac. In some ways its a lot like the diy one here I would recommend.
I have a lot of reading and research to do before I could have any questions. Just what I was looking for. Thanks!
It is called the Tamabqui. Mola Mola is the company.
Really? It has three ADSP 21489 processors.
Well, there are clear differences which is why I didn't say they were almost exactly alike in every way, or something like that.
Clearly they are a heck of a lot more alike than either of them would be compared to an ESS or an AKM chip dac. Although IIUC, Mola Mola Tambaqui no longer measures better than the latest consumer dacs in terms of SINAD. Especially in terms of SINAD/$. I would expect it still sounds better though.
Clearly they are a heck of a lot more alike than either of them would be compared to an ESS or an AKM chip dac. Although IIUC, Mola Mola Tambaqui no longer measures better than the latest consumer dacs in terms of SINAD. Especially in terms of SINAD/$. I would expect it still sounds better though.
Last edited:
Comparing Mola Mola Tambaqui to this version of Marcel's RTZ DAC is far-fetched to say the least.
Mola Mola Tambaqui was designed to be the best performing DAC at its time. Here is what Bruno Putzeys said in an interview:
"It is my experience -- confirmed by every new thing I do -- that when you get into really high measured performance, really low distortion, superlow noise, then the ultimate subjective sound quality starts improving and continues to improve in step with the measurements. At some point you will find that a product that measures absolutely perfectly under an extensive battery of tests will sound a lot better than a product with more typical high-end audio performance that has been tuned by ear for years."
(from https://www.soundstageultra.com/ind...s-of-mola-mola-hypex-and-grimm-audio-part-one)
While Marcel's RTZ DAC is a commendable design the combination with PCM2DSD has clear issues which have not been solved to date despite many efforts. There is level dependent out-of-band noise (see attachment) which, even though it looks like it, is not distortion. Marcel's RTZ DAC works much better with some HQPlayer quasi-multibit filters but even then it can hardly be compared to Mola Mola Tambaqui.
Mola Mola Tambaqui was designed to be the best performing DAC at its time. Here is what Bruno Putzeys said in an interview:
"It is my experience -- confirmed by every new thing I do -- that when you get into really high measured performance, really low distortion, superlow noise, then the ultimate subjective sound quality starts improving and continues to improve in step with the measurements. At some point you will find that a product that measures absolutely perfectly under an extensive battery of tests will sound a lot better than a product with more typical high-end audio performance that has been tuned by ear for years."
(from https://www.soundstageultra.com/ind...s-of-mola-mola-hypex-and-grimm-audio-part-one)
While Marcel's RTZ DAC is a commendable design the combination with PCM2DSD has clear issues which have not been solved to date despite many efforts. There is level dependent out-of-band noise (see attachment) which, even though it looks like it, is not distortion. Marcel's RTZ DAC works much better with some HQPlayer quasi-multibit filters but even then it can hardly be compared to Mola Mola Tambaqui.
Attachments
Does the low-level not-quite-harmonic distortion issue (which is really a frequency-modulated idle tone) still occur with my latest configuration file?
https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7895559
https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7895559
I don't know if anybody has measured it with your latest configuration file. Anyhow IIUC Markw4 is not using it for some reason.
Regarding the supposed low-level of this not-quite-harmonic distortion issue it seems to be audible:
https://www.diyaudio.com/community/threads/general-purpose-dac-clock-board.413001/post-7757408
Regarding the supposed low-level of this not-quite-harmonic distortion issue it seems to be audible:
https://www.diyaudio.com/community/threads/general-purpose-dac-clock-board.413001/post-7757408
@PJotr25 has measured it with versions 1.2 and 1.3, so before the DSD remodulating bug was fixed. Using a PCM input signal, he got a second harmonic of about -138 dBFS to -141 dBFS when the signal was -60 dBFS, 1 kHz, and the higher harmonics (or peaks close to harmonics) drowned in his -150 dBFS noise floor. The typical Bessel sideband shape that you saw (as in post #353 of this thread) was nowhere to be found.
Ok. Let's hope somebody measures your latest version with a more performant ADC.
Just as a comparison here is PCM2DSD v03 (i.e. higher level not-quite-harmonic distortions) with AK4493 measured with coherent averaging. Note the level of noise floor. Highest "harmonics" are lower than -160dBFS.
Just as a comparison here is PCM2DSD v03 (i.e. higher level not-quite-harmonic distortions) with AK4493 measured with coherent averaging. Note the level of noise floor. Highest "harmonics" are lower than -160dBFS.
Attachments
There are multiple FPGA firmware versions, including the ones from XX3stksm. The latest version from Marcel is a sigma-delta PWM type modulator, same basic type used in Tambaqui. There is also such a modulator in HQ Player. Thus there are multiple choices the user has for choosing a sound they like.
One difference between the version of Marcels RTZ dac I am using and Tambaqui is the the MCLK frequencies. Tambaqui uses 45/49MHz clocks so that the PWM modulator can output nice measuring DSD about like DSD128 but with less low level artifacts. The problem with that IMHO is that close-in phase noise gets worse by 6dB with each doubling of clock frequencies. Also, no matter which modulator is used in HQ Player, DSD128 just doesn't cut it.
IME DSD doesn't start sounding really good until it gets up to DSD256, which I would rather have than the equivalent of low artifact DSD128. Also, I would expect imaging to suffer with higher MCLK frequencies. Moreover, so far as anyone knows the technology to make clocks performing at the level of Acko clocks didn't exist when Tambaqui was designed. They would have been better than the known SOA at the time.
The problem with using Marcel's sigma-delta modulator with 22/24MHz MCLK frequencies is that it is in some ways like DSD64, which just doesn't sound like good DSD256. The separation between instruments (openness) just isn't there. Neither is the incredible imaging Acko clocks can give with a different modulator.
Another difference is that Tambaqui uses a custom adaptive FIFO-buffered ASRC. Apparently it was hard to fit into the SHARC chips. My philosophy is I don't need it. Isolated asynchronous USB simplifies things greatly and can give equally good sound.
Anyway, we can go on to make a list of differences between Tambaqui and my implementation of Marcel's RTZ dac.
Regarding AK4493, it is well known that it can measure quite well yet sound awful. Also, what we typically measure is not full spectrum of measurements Bruno may have made. However, at the time Tambaqui was designed there was no known way to measure the jitter of a dac. A paper describing a way to do is still fairly new, and even it can't measure the close-in phase noise effects of a dac and MCLK combined system. Thus there are some things we are still left to evaluate by ear on a very good reference system located in a well treated room.
So, I will stick with my claim that they are a lot alike, but not exactly alike. They are more similar in some ways dacs fabricated entirely out of CMOS processes. For one big different there is not substrate coupled-noise like most CMOS dacs suffer. Only the latest dac chipset from AKM separates the noisy modulator from the output array. Yes, the noise can be jumbled up so its power is distributed across all FFT bins to hide it, but AKM knows its still a problem in most all CMOS sigma-delta dac chips.
Anyway, Marcel's dac with the right surrounding circuitry can sound far better than any other dac I have ever heard (and I have an ADI with AK4493 that sounds like crap, and I know other people who find the same).
One difference between the version of Marcels RTZ dac I am using and Tambaqui is the the MCLK frequencies. Tambaqui uses 45/49MHz clocks so that the PWM modulator can output nice measuring DSD about like DSD128 but with less low level artifacts. The problem with that IMHO is that close-in phase noise gets worse by 6dB with each doubling of clock frequencies. Also, no matter which modulator is used in HQ Player, DSD128 just doesn't cut it.
IME DSD doesn't start sounding really good until it gets up to DSD256, which I would rather have than the equivalent of low artifact DSD128. Also, I would expect imaging to suffer with higher MCLK frequencies. Moreover, so far as anyone knows the technology to make clocks performing at the level of Acko clocks didn't exist when Tambaqui was designed. They would have been better than the known SOA at the time.
The problem with using Marcel's sigma-delta modulator with 22/24MHz MCLK frequencies is that it is in some ways like DSD64, which just doesn't sound like good DSD256. The separation between instruments (openness) just isn't there. Neither is the incredible imaging Acko clocks can give with a different modulator.
Another difference is that Tambaqui uses a custom adaptive FIFO-buffered ASRC. Apparently it was hard to fit into the SHARC chips. My philosophy is I don't need it. Isolated asynchronous USB simplifies things greatly and can give equally good sound.
Anyway, we can go on to make a list of differences between Tambaqui and my implementation of Marcel's RTZ dac.
Regarding AK4493, it is well known that it can measure quite well yet sound awful. Also, what we typically measure is not full spectrum of measurements Bruno may have made. However, at the time Tambaqui was designed there was no known way to measure the jitter of a dac. A paper describing a way to do is still fairly new, and even it can't measure the close-in phase noise effects of a dac and MCLK combined system. Thus there are some things we are still left to evaluate by ear on a very good reference system located in a well treated room.
So, I will stick with my claim that they are a lot alike, but not exactly alike. They are more similar in some ways dacs fabricated entirely out of CMOS processes. For one big different there is not substrate coupled-noise like most CMOS dacs suffer. Only the latest dac chipset from AKM separates the noisy modulator from the output array. Yes, the noise can be jumbled up so its power is distributed across all FFT bins to hide it, but AKM knows its still a problem in most all CMOS sigma-delta dac chips.
Anyway, Marcel's dac with the right surrounding circuitry can sound far better than any other dac I have ever heard (and I have an ADI with AK4493 that sounds like crap, and I know other people who find the same).
Last edited:
So something not visible can be hidden in FFT but something clearly visible in FFT cannot be hiding even more gory details? Yes, sure.
Any quotes from controlled listening tests for this claim since it is well known? I haven't said anything about the sound of AK4493 but Marcel's RTZ dac has been decommissioned long since as I have better DACs.
The AK4493 based DACs I have heard sound very good to me. This would be SU-1 (within its limits) which I just tried out in a wim which led me to the Kaamos Tech W-DAC XLR (+ their power and clock buffer) which sounds utterly nothing i.e. transparent, solid, clean and very dynamic - i.e. a superb DAC.
//
//