@nautibuoy
Could you please clarify your signal path for the readers in the present context? Is it I2SoverUSB with two power supplies into Marcel's dac-3, or something else? If Simple DSD Converter is used, where does it fit in? If not using it is it out of the circuit? How is it powered?
Reason I ask all the questions is that there may or may not be something different in the hardware in addition to any differences in conversion algorithms. Not clear enough about the hardware part of it to fully understand.
Thanks!
Could you please clarify your signal path for the readers in the present context? Is it I2SoverUSB with two power supplies into Marcel's dac-3, or something else? If Simple DSD Converter is used, where does it fit in? If not using it is it out of the circuit? How is it powered?
Reason I ask all the questions is that there may or may not be something different in the hardware in addition to any differences in conversion algorithms. Not clear enough about the hardware part of it to fully understand.
Thanks!
Discovered a problem with the setup this morning when I was scoping out signals to get DSD256 working. It turns out the Andrea I2S inverter board works fine to produce inverted I2S signals if the FIFO buffer is set to operate in non-RTZ mode. However, when the FIFO board is producing RTZ waveforms for the non-inverting input signals that go directly to the shift register inputs, then everything is good. However, if those non-inverting RTZ I2S signals are inverted and sent to the shift register inverting inputs, it turns out the resulting waveforms from the inversion operation amount to something more like "return to 5v" instead of "return to zero." That' doesn't work right.
The above having been said, there isn't much practical effect at the moment since I am only using the non-inverting analog outputs into passive filtering.
What I decided to do in the meantime was remove the drive signals from the inverting shift register inputs so that power supply noise generated by incorrect inverting signals would not modulate shift register Vref power. Hopefully that will help keep non-inverting SE analog outputs cleaner than they might otherwise be under the present conditions.
With that done, and after letting things settle-in most of the day in DSD256 operation, I will try to get some people over tomorrow to listen. My own impression at this time is that I am deep enough down into the heart of this dac to notice what I suspect could be sound of the D/A conversion resistors. Did have one other person listen so far. We both notice some remaining blurring of the sound as compared to Andrea dac output array sound in SE output mode. Don't know if I want to get into resistor swapping at this point since I already did that experiment with Andrea's dac board (might make an exception though if Acko wants to invest into some smd metal foil resistors to try on his Marcel RTZ dac board currently on loan here 🙂 ).
Other than that, will plan on reporting back when I have some more listening impressions to share.
The above having been said, there isn't much practical effect at the moment since I am only using the non-inverting analog outputs into passive filtering.
What I decided to do in the meantime was remove the drive signals from the inverting shift register inputs so that power supply noise generated by incorrect inverting signals would not modulate shift register Vref power. Hopefully that will help keep non-inverting SE analog outputs cleaner than they might otherwise be under the present conditions.
With that done, and after letting things settle-in most of the day in DSD256 operation, I will try to get some people over tomorrow to listen. My own impression at this time is that I am deep enough down into the heart of this dac to notice what I suspect could be sound of the D/A conversion resistors. Did have one other person listen so far. We both notice some remaining blurring of the sound as compared to Andrea dac output array sound in SE output mode. Don't know if I want to get into resistor swapping at this point since I already did that experiment with Andrea's dac board (might make an exception though if Acko wants to invest into some smd metal foil resistors to try on his Marcel RTZ dac board currently on loan here 🙂 ).
Other than that, will plan on reporting back when I have some more listening impressions to share.
Summarizing:
1. The distortion cancellation between the positive and negative sides has been removed by going single-ended and slowing down the common-mode loop
2. Then the suppression of out-of-band quantization noise has been much reduced and some slight roll-off below 20 kHz has been introduced by using a simpler filter
3. I don't remember the order, but at some moment a reclocker and isolation board has been added
4. The reference decoupling capacitors have been replaced with more expensive models that are (practically) linear and non-microphonic
5. The reference current has been made much more data-dependent by using incorrect bit patterns, leading to a less clean reference, which can cause extra down conversion of out-of-band rubbish and extra distortion
and every time it supposedly sounds better.
From a technical point of view, I agree that 3 is an improvement. The lack of microphonics of the capacitors of 4 is also an improvement, provided the new capacitors have small enough inductance, which they probably have. In fact, I deliberately avoided ceramic class 2 capacitors at higher-impedance nodes in the reference circuitry, which are much more sensitive to microphonic behaviour of the capacitors.
The other points don't seem logical to me.
1. The distortion cancellation between the positive and negative sides has been removed by going single-ended and slowing down the common-mode loop
2. Then the suppression of out-of-band quantization noise has been much reduced and some slight roll-off below 20 kHz has been introduced by using a simpler filter
3. I don't remember the order, but at some moment a reclocker and isolation board has been added
4. The reference decoupling capacitors have been replaced with more expensive models that are (practically) linear and non-microphonic
5. The reference current has been made much more data-dependent by using incorrect bit patterns, leading to a less clean reference, which can cause extra down conversion of out-of-band rubbish and extra distortion
and every time it supposedly sounds better.
From a technical point of view, I agree that 3 is an improvement. The lack of microphonics of the capacitors of 4 is also an improvement, provided the new capacitors have small enough inductance, which they probably have. In fact, I deliberately avoided ceramic class 2 capacitors at higher-impedance nodes in the reference circuitry, which are much more sensitive to microphonic behaviour of the capacitors.
The other points don't seem logical to me.
Last edited:
Mark,
So one channel still RTZ and while simply inverted on the Se to Diff board, the other channel is now RT5V, true ?
That means you should remove the Se to Diff board, go back on the Fifo to Bclk instead of RTZ and let Marcel’s board do the RTZ and clock doubling to get a correct sound impression.
Hans
P.S. This was exactly why I was strugling with the 2*BCLK in my diagram
So one channel still RTZ and while simply inverted on the Se to Diff board, the other channel is now RT5V, true ?
That means you should remove the Se to Diff board, go back on the Fifo to Bclk instead of RTZ and let Marcel’s board do the RTZ and clock doubling to get a correct sound impression.
Hans
P.S. This was exactly why I was strugling with the 2*BCLK in my diagram
I will just take #3 and #4 only at this stage. The rest does not make sense to me to consider…
Last edited:
I've never listened to RTZ without an isolator/reclocker arrangement (JLSounds USB module) when using my HQP 'source'. The only exception in my RTZ listening experience was the use of Pavel's DSD'it for the build I did for a friend, which received PCM data from Roon - DSD'it reclocks it's output IIRC.
Last edited:
@Nautibuoy, are you able to test with output filter board replaced by a suitable transformer type, like Lundhall etc? This will be more like DSC2.x arrangement. Mark has proposed something similar and another member here has also reported some benefits.
Sorry @acko, no can do - I am currently sans RTZ DAC until I can assemble myself a couple of new boards, hopefully in a 3-4weeks time, nor do I have any spare transformers. I have Lundahl transformer output arrangements in both my Valve DAC (Marcels 3rd order passive filter) and DSC2 but I'm not prepared to dismantle those fine sounding, and regularly used decoders, for this exercise. I do still want to experiment again with the 'CD Enhancer' transformer/valve-based output arrangement I posted a while back, for which I have most of the parts.
Is this a surprise? With high expectation bias and lack of AB testing no other result should be expected. Same thing has been repeated time after time in threads involving tweaking and modifications.
Next step will be claiming that some expensive super-ultra-low-noise resistor will sound better although LIGO resistor test already shows that e.g. the current noise of Susumu RG is practically at Johnson noise.
Regarding my quick comparison of PCM2DSD, yesterday:
HQPlayer > DSD256 data > JLSounds > NoDAC RC filter > Noir HPA > HiFiMan Sundara HPs
HQPlayer > PCM data > JLSounds > PCM2DSD > NoDAC RC filter > Noir HPA > HiFiMan Sundara HPs
This was a workbench exercise but I did use good quality power supplies and no compoimising of the JLSounds isolation.
I also have an Amanero clone that I have now reconnected to the PCM2DSD board and will give that a listen later - the nice thing with that is it is very easy to playback from my smartphone using the USB Audio Pro app and a USB OTG connector - pretty good sound quality in my experience even when not outputting via USB -I regularly use it with my Sennheiser Bluetooth HPs when out and about - https://www.extreamsd.com/index.php/products/usb-audio-player-pro
HQPlayer > DSD256 data > JLSounds > NoDAC RC filter > Noir HPA > HiFiMan Sundara HPs
HQPlayer > PCM data > JLSounds > PCM2DSD > NoDAC RC filter > Noir HPA > HiFiMan Sundara HPs
This was a workbench exercise but I did use good quality power supplies and no compoimising of the JLSounds isolation.
I also have an Amanero clone that I have now reconnected to the PCM2DSD board and will give that a listen later - the nice thing with that is it is very easy to playback from my smartphone using the USB Audio Pro app and a USB OTG connector - pretty good sound quality in my experience even when not outputting via USB -I regularly use it with my Sennheiser Bluetooth HPs when out and about - https://www.extreamsd.com/index.php/products/usb-audio-player-pro
Ok, understood. Happy to order one more assembled board, this time for you to test - at your earliest convenience. Also, to help out with Lundahl or this Bisesik trafos , don’t know which is better and plus an output board to match the RTZ board. All previous discretes have transformer outputs but this RTZ DAC has electronic type. I am curious. Any proposals for something optimised, if interested?
@acko I'm going to build myself the two RTZ boards I originally set out for myself and that will be it on asembling RTZ DAC boards - assembling the boards is tedious and time consuming and I'm not prepared to commit to supplying any more for others, particularly when I see the crude hacking being inflicted on the one I supplied to you. I'm sure in time I'll try some different things when I get my RTZ DACs working but any outcomes will be entirely subjective and by no means definitive.
In terms of a transformer based output arrangement, I think the best person to consider that option is Marcel, the RTZ is his design...
In terms of a transformer based output arrangement, I think the best person to consider that option is Marcel, the RTZ is his design...
Hi Marcel, thank you for the questions. This is probably a good time to go over your list of concerns.
Regarding distortion cancellation, an unfortunate side effect of trying to cancel distortion was that the sound was blurred and low level musical details were missing. At the time I suggested and still suspect likely that element matching limitations due to resistor tolerances and switch resistance variations results in reproduction of small musical details that are not perfectly equal and opposite between phases. There are both unwanted distortion and some low level wanted musical information which are being cancelled. The typical presumption is there is nothing worse than harmonic distortion. Well, I disagree in this case. The loss of low level musical information is worse than a little low level distortion. Besides, everyone knows real music doesn't sound blurred and missing details. That simply can't be right. So its a matter of pick your poison. I will pick to allow a little more distortion while I dig down to find out how to reduce the distortion in the first place. Resistor excess noise is a type of distortion too, and in this case it is somewhat objectionable (suspected to be, that is), more so than a tiny bit of harmonic distortion.
The op amp filter sounded blurry. I agree there is a need for a better filter. Either a discrete type, or else a passive one with inductors. In the meantime I need to reduce the blur to hear what the actual dac is doing. The blur of the op amp filter board masks the dac array sound. Whether or not the opamp blur shows up well on an FFT is immaterial to me. It still makes it hard to hear all of the low level musical detail, and hard to hear what the remaining low level problems are.
Yes. It does make the sound more precise, and its helps some with detail and stereo imaging.
The sound changed and distortion caused by non-linear capacitance is now less. That fact that you tried to minimize ripple currents in the capacitors was a good idea to help, but better caps help some too.
Correct. It is a problem that will take some time to fix because I don't have the necessary parts on hand. I can get them though if Acko is willing to wait. At that point I would like to go ahead and try dummy loads on the unused inverting analog outputs as we discussed.
I would say its been slowing moving from, what did I say at first, maybe and A- dac to maybe being an A. Don't think its quite in A+ territory yet though, but the potential is still there. To keep it moving in the right directions I will need to go back and fix some things we both agree need fixing (as described above).
I agree that you have used a great deal of logic in your design. However, when it comes to audio I am more of an empiricist. If something works to improve sound and I don't fully know why, then it only means there is something I don't know. It doesn't mean the sound isn't better and more accurate according to human perception. I keep trying to point out that humans are not FFT analyzers. The time domain waveform often takes precedence over the frequency domain view in human perception. Also FFTs of one channel don't tell us if stereo imaging is good or not. Its typically left to chance. Why? Most likely because of the reason Earl Geddes gave in the context of his work with distortion and speakers (which explains a lot more than only about the lack of attention to stereo imaging):
After a serious quest for continued support for more than a year we gave up. Our conclusion; people are satisfied with THD and IMD. It’s like the story of the cop who asks a drunk under a street light what he is doing on his hands and knee’s. The drunk replies “I’m looking for my car keys.” The officer asks “Where did you loose them?” and the drunk replies “Over there by my car.” Baffled, the officer asks “Then why are you looking for them here?” to which the drunk replies, “Because the light is better.” Everyone knows that THD is meaningless, but it’s easy to do and “the light is better.” To add to this situation, we have also found that nonlinear distortion in loudspeakers is, for the most part, not a significant issue. There are, to be sure, subjective distortions that are level dependent and as such are thought to be nonlinear distortion, but they are in fact linear effects that have a nonlinear perception. The testing of this hypothesis is currently underway and the results will probably be available in the future.
Dr. Geddes talks mostly about loudspeakers whereas my interest is in dacs. However IMHO some of the same points hold. HD/IMD/THD is not everything. There are also linear effects that sound like distortion. They just don't look like distortion on an FFT. IMHO we need to get past our hangup with HD as the only thing that matters, a hangup which we have mostly because of the "that's where the light is best" human bias problem.
Last edited:
Mark, at this point in time the concerns of Marcel need to be considered and I am mindful that it is his design and any further work by yourself needs some sort of agreement for the scope going forward. I will await Marcel’s decision
You and Marcel have always had the power to call it quits. I also have that same power. We have only ever proceeded by mutual consent.
Hi Mark, thanks for replying. I didn't actually ask any questions, I just made a summary and gave an opinion.
Ok, Let’s pause work until we can work out a new scope going forward that is agreeable to parties involved.
Last edited:
Close, but not close enough - The CPU loading of my music server does seem to have been reduced somewhat by the upgrades of Audiolinux and HQ Player but not quite enough to be able to play DSD512 with the more demanding modulator/filter settings, which happen to be the ones I like - I'm still getting some dropouts/glitches, albeit at a much reduced level. I may cast around for a video card upgrade to see if some CUDA off-loading helps - my current NVidia card, with a computing capability of 5.0, just falls short of making the grade (5.2) on the Signalyst website. For now I'll stick with my DSD256 preferences.
I'm wondering if different DSD decoder solutions might respond differently to different modulator/filter settings, however, there are so many combinations avaialble in HQ Player it could turn into a lifetime's work.
I'm still surprised that I seem to be the only one who has used the RTZ DAC with HQ Player sourced input data.
I'm wondering if different DSD decoder solutions might respond differently to different modulator/filter settings, however, there are so many combinations avaialble in HQ Player it could turn into a lifetime's work.
I'm still surprised that I seem to be the only one who has used the RTZ DAC with HQ Player sourced input data.