Hi Marcel,
Understand about the theory. There appears to be something wrong however, possibly with how the theory is implemented in practice. Hard to say exactly.
Thing is I know there are people who have listened to your dac (an others like it) and hear it the same as I do. Would you be willing to listen yourself, or is there anyone you would trust to listen in your stead?
Its not even clear if a large scale double blind study would convince you to listen and see for yourself?
Mark
Understand about the theory. There appears to be something wrong however, possibly with how the theory is implemented in practice. Hard to say exactly.
Thing is I know there are people who have listened to your dac (an others like it) and hear it the same as I do. Would you be willing to listen yourself, or is there anyone you would trust to listen in your stead?
Its not even clear if a large scale double blind study would convince you to listen and see for yourself?
Mark
I would agree the twists are too tight to be ideal. But I would consider that part of the geometry that defines its EM properties. Other factors are involved too, but I don't to risk turning this into a cable thread right now
Mark,
A few comments.
The only thing that Balanced means is that both outputs have the same resistance
When in SE inserting a 49r9 resitor in the output gnd line, just as the signal line has, you already have a balanced output.
You then could insert the interlink at the other end in the balanced input for having CM rejection that’s induced in the interlink.
But only when PE is connected to both your preamp and to the Firdac.
Differential + Balanced means opposite phase and again the same output Resistance.
This signal can only be put at the other end in the balanced input.
With differential outputs, nothing of unequal opposite phase, because of gain differences, will be rejected but only the not equally amplified common mode elements from the Dac will not be optimally suppressed at the receiving side.
In all this, one important question: where is the protective earth connected to your preamp and to your Firdac.
If PE not properly connected to both sender and receiver, its obvious that you need SE, because through the gnd wire you connect both analog grounds together, just because the Dac will be floating. That’s what I guess could be your situation.
Hans
Hans,
I know the theory. I had to be convinced it was somehow wrong in practice, something I resisted very strongly for a long time. In the end I had to admit I had been blinded by theory where the implementation was virtually never like the theory. Except for PSS FFTs of course. The problem just doesn't show up very well that way. Only way I know of to find out the difference between SE and balanced is to listen without prejudice, and only then form your own personal opinion.
And its not only me either, other engineers have had to overcome theory blindness and common engineering misconceptions at times with great difficulty.
For example, one myth has to do with concept of ground. An IEEE book on the subject of grounds says ground is a 'fantasy invented by engineers' for their convenience. This is the IEEE saying that. And they're right.
Another myth frequently found among some engineers and around some audio forums is that a "threshold of audibility" is an absolute limit which no human can overcome.
In this particular case you need to check on the theory in practice using your ears. PSS FFTs are not showing you everything. It like you have to overcome preexisting bias and just listen. What would it take to get you just listen? No way?
Mark
EDIT: Just occurred to me if your system is full of opamps doing SE>Balanced and vice versa, it might already be hard to hear some low level musical details, no matter if seems you are hearing everything right now or not. I would suggest a very simple signal chain and maybe headphones if that's more simple yet sensitive. Also, my claim is only about low level musical details not about noise or distortion. If we can agree on the musical details, it will turn out we can do a lot about the distortion and or noise that sounds like distortion. Also, I would agree there is a need to be very careful about grounding for signal integrity.
I know the theory. I had to be convinced it was somehow wrong in practice, something I resisted very strongly for a long time. In the end I had to admit I had been blinded by theory where the implementation was virtually never like the theory. Except for PSS FFTs of course. The problem just doesn't show up very well that way. Only way I know of to find out the difference between SE and balanced is to listen without prejudice, and only then form your own personal opinion.
And its not only me either, other engineers have had to overcome theory blindness and common engineering misconceptions at times with great difficulty.
For example, one myth has to do with concept of ground. An IEEE book on the subject of grounds says ground is a 'fantasy invented by engineers' for their convenience. This is the IEEE saying that. And they're right.
Another myth frequently found among some engineers and around some audio forums is that a "threshold of audibility" is an absolute limit which no human can overcome.
In this particular case you need to check on the theory in practice using your ears. PSS FFTs are not showing you everything. It like you have to overcome preexisting bias and just listen. What would it take to get you just listen? No way?
Mark
EDIT: Just occurred to me if your system is full of opamps doing SE>Balanced and vice versa, it might already be hard to hear some low level musical details, no matter if seems you are hearing everything right now or not. I would suggest a very simple signal chain and maybe headphones if that's more simple yet sensitive. Also, my claim is only about low level musical details not about noise or distortion. If we can agree on the musical details, it will turn out we can do a lot about the distortion and or noise that sounds like distortion. Also, I would agree there is a need to be very careful about grounding for signal integrity.
Last edited:
Why would I have to listen for myself if a large scale double blind study had been done and had shown you were right? It would be more useful to try and find an explanation then. It's hypothetical anyway, as so far all we have are uncontrolled listening tests producing illogical results.
I see your point I guess. For me I think the answer is I want to work towards better dacs in a practical way, without the cost, complexities, and delays of doing a study when the effects I speak of are likely to be pretty obvious, that is, given the caveats already mentioned in my edited reply to Hans. Flying you out here, showing and telling you everything, trying things you might want to suggest or that you might want to see tested -- the whole thing would be so much lower cost and faster than doing a large scale blind study. Once you hear it, then its obvious. Maybe like the two files you listened to at Purify that had the same PSS FFTs spectra. They sounded different, although they measure exactly the same using that particular methodology. The fact that the FFT spectra are exactly the same would likely make some people skeptical the files could sound different. Point is you were willing to listen anyway and you heard the difference in that case. So I think you could probably hear the difference between SE and Balanced in this case. You just to be able to put aside being skeptical for a short while and listen for low level musical details without becoming distracted by hearing more perceived distortion and or noise. Those are separate issues which IME can be attenuated significantly.
Last edited:
The still mysterious use of the pronoun "we" does not make your findings any more objective. From your pictures it is easy to see that no proper A/B testing of modifications has been performed. So all you results so far are subjective and based on uncontrolled listening tests with high expectation bias.
I don't know when I will have the time to actually do it, but here is what I will do:
I'll take my old preamplifier from the attic. It is completely discrete, single-ended and class A and has a headphone amplifier built in.
I'll hook it up single-endedly (it only has single-ended inputs anyway) to the prototype RTZ shift register FIRDAC, which I will modify such that I can use my ABX switchbox to switch between large and small common-mode loop bandwidth. The FIRDAC will be driven from the digital part of my earlier solid-state DAC, which will be driven from the S/PDIF output of a DVD player. Everything is double insulated, so there will be no ground loops.
I'll do 20 trials with the digital part of the solid-state DAC in PWM8 mode (which results in a relatively clean out-of-band spectrum; lots of noise but no idle tones or other aberrations) and 20 with it in chaotic mode and report the raw results.
For what it's worth, the last time I was involved in organizing a blind test on this forum, a test related to DAC pre-echoes, there were six participants, five of whom were convinced they clearly heard a difference but had completely random results, and one who wasn't convinced at all that he heard any difference and who had everything right except for one recording - a recording he complained about, because he considered the sound quality to be insufficient. (He complained before the results were published.) He did worse on a retest, but again had the largest pre-echoes correct on all recordings except the one with insufficient quality. Unfortunately the test was too small to really draw hard conclusions. See the attachment for details.
I'll take my old preamplifier from the attic. It is completely discrete, single-ended and class A and has a headphone amplifier built in.
I'll hook it up single-endedly (it only has single-ended inputs anyway) to the prototype RTZ shift register FIRDAC, which I will modify such that I can use my ABX switchbox to switch between large and small common-mode loop bandwidth. The FIRDAC will be driven from the digital part of my earlier solid-state DAC, which will be driven from the S/PDIF output of a DVD player. Everything is double insulated, so there will be no ground loops.
I'll do 20 trials with the digital part of the solid-state DAC in PWM8 mode (which results in a relatively clean out-of-band spectrum; lots of noise but no idle tones or other aberrations) and 20 with it in chaotic mode and report the raw results.
For what it's worth, the last time I was involved in organizing a blind test on this forum, a test related to DAC pre-echoes, there were six participants, five of whom were convinced they clearly heard a difference but had completely random results, and one who wasn't convinced at all that he heard any difference and who had everything right except for one recording - a recording he complained about, because he considered the sound quality to be insufficient. (He complained before the results were published.) He did worse on a retest, but again had the largest pre-echoes correct on all recordings except the one with insufficient quality. Unfortunately the test was too small to really draw hard conclusions. See the attachment for details.
Last edited:
Hi Mark,
One last positive reaction.
1) I have no problem with your preference in specific situations for SE, just as myself being a person who always use his ears to come to conclusions without prejudice.
But I have a problem with the term SE, telling nothing but that RCA connectors are used.
It tells nothing about how the sending signal was constructed, could be made with a SE converter at the end of a fully differential amp, or from a true single ended amp from start to finish and anything in between.
On the receiving side SE could be back converted into differential or stay SE and again anything in between.
So postulating that SE sounds better is IMO like saying “Gentlemen prefer blond”.
I have no idea about your preamp receiving the DAC signal, is it internally SE and is the balanced input directly converted into SE ?
What I’m telling is that there are just so many differences between individual equipment that only by listening you can select the right connection.
Important when switching is to always keep using the same interconnect and SE and balanced to SE and SE to balanced adapters.
That way you can go from SE to SE, from SE to Balanced, from Balanced to SE and finally from Balanced ro Balanced.
My experience is that the best sounding combination of the four is strongly dependent on the equipment used, and above all, it takes away all prejudice.
Hans
P.s. with the adapters be sure to to tap the right signals
Last edited:
It has a transformer coupled input with windings designed to support balanced or unbalanced input signals. The secondary side of the transformer is coupled into an amplifier with unbalanced inputs.
When I refer to SE or balanced, I am generally referring to the output of the dac itself. Mostly I am interested in the difference in sound if either one one phase (say, the non-inverting phase), or else both phases of the switched resistor arrays are sent down a cable to the transformer inputs of the amplifier. IOW, I want to know what the difference is between any type of balancing at all, or else no balancing at all (i.e. not-fully unbalanced, or else fully unbalanced). I consider 'fully' SE = 'fully' unbalnced.
Therefore, I didn't consider Marcel's original implementation of differentially summing both resistor array phases as producing 'fully' SE outputs. Each output in that case was a differentially summed output, with the two final analog outputs being in relative inverted phase with each other. That said, only using one of the differentially summed outputs instead of both of them would result in unbalanced transmission to the preamp inputs. Therefore, that case would not constitute a 'fully' balanced interconnect.
Also, I have both RCA and XLR cables, plus adapters as needed so that SE and balanced transmission can always be correctly implemented. The pin header output of Marcels filter board always connect to XLR output connectors a few inches away. If SE RCA cables are to be used, adapters between to filter board-located XLR outputs are used as needed just before going into the RCA interconnect cables, which then carry the (not necessarily 'fully') SE signals to the amplifier.
Also, in response to another issue you raised at some point, I am very, very careful about not letting AC power grounds produce ground loops with the signal transmission grounds. To that end one of the things I can do and have tested is floating the transformer primary so there is galvanic isolation in the audio signal path. The transformer also has internal electrostatic shielding between primary and secondary to greatly attenuate any capacitive coupling between windings.
Is that sufficiently clear, or do I need to clarify more?
Last edited:
Mark, as I see it you can get away with this SE transmission with Marcel’s DAC because of the clever implementation of differentially summing the dac core signals (balanced) onto to SE output. As what Marcel puts it, using it ‘single-endedly ’. ‘Nautibuoy’ also benefits from this SE output on his electronically coupled SE input HPA. It took me quite a while to figure this out - see my queries to Marcel in earlier posts.
So, it is not really a battle between SE and BAL configuration. For another implementation of DAC without the summing function, the SE output suffers from CM issues if only one leg of the balanced output is used and there is not much an SE input of the next stage could. In such a case your transformer needs to set to tap the balanced outputs of a DAC - transformers make it easy.
Shall we now consider the ‘BALSE’ output concept to put an end to these arguments and move on …
So, it is not really a battle between SE and BAL configuration. For another implementation of DAC without the summing function, the SE output suffers from CM issues if only one leg of the balanced output is used and there is not much an SE input of the next stage could. In such a case your transformer needs to set to tap the balanced outputs of a DAC - transformers make it easy.
Shall we now consider the ‘BALSE’ output concept to put an end to these arguments and move on …
Last edited:
Acko,
I see it differently. A lot of behind the scenes work by a number of people have shown that fully-SE dacs tend to produce subjectively better sound than if balancing is used at any point. The details are NDA.
Also I know what I am suggesting goes against theory, which I just reiterate here. I also said the implementation of balancing dac outputs (virtually) never works as well in practice as theory would predict, with one semi-exception. It can look like the the theory is working well if using PSS FFTs to judge. As I have also said before, IMHO and IME believing an FFT over listening tests in this type of situation is one type of example of listening with your eyes instead of your ears. To understand how FFTs can fool you requires some deeper understanding of DFTs and spectral analysis processing; its a subject maybe better suited to its own thread. That's my considered opinion on it. I know some people are not going to agree. Some will not even agree to try listening for themselves.
As one further reminder, balancing the resistor array output results in blurring of the sound and loss of low level musical details. If nobody wants to consider that possibility, then fine, I will proceed as I see fit to get the dac sounding more up to its potential whatever it takes. When I'm done I would bet it just might give an expensive commercial dac a run for its money.
Or if you would prefer, I would be happy to return the dac to you at any time. My last understanding from you was that you wanted me to proceed as I see fit.
Mark
I see it differently. A lot of behind the scenes work by a number of people have shown that fully-SE dacs tend to produce subjectively better sound than if balancing is used at any point. The details are NDA.
Also I know what I am suggesting goes against theory, which I just reiterate here. I also said the implementation of balancing dac outputs (virtually) never works as well in practice as theory would predict, with one semi-exception. It can look like the the theory is working well if using PSS FFTs to judge. As I have also said before, IMHO and IME believing an FFT over listening tests in this type of situation is one type of example of listening with your eyes instead of your ears. To understand how FFTs can fool you requires some deeper understanding of DFTs and spectral analysis processing; its a subject maybe better suited to its own thread. That's my considered opinion on it. I know some people are not going to agree. Some will not even agree to try listening for themselves.
As one further reminder, balancing the resistor array output results in blurring of the sound and loss of low level musical details. If nobody wants to consider that possibility, then fine, I will proceed as I see fit to get the dac sounding more up to its potential whatever it takes. When I'm done I would bet it just might give an expensive commercial dac a run for its money.
Or if you would prefer, I would be happy to return the dac to you at any time. My last understanding from you was that you wanted me to proceed as I see fit.
Mark
Last edited:
The jury is still out there… please continue with this DAC. Perhaps take the fight to Holo May L3 for a start
Probably, but it requires solving a somewhat different set of problems in the overall dac design. In the end its more complicated and more expensive to do well. In return its possible to get better SQ. Again, this is IMHO and IME.